Research Dossiers

1. Data center electrical demand accelerates with hard figures hecho · alta N3

   Global data center consumption: 565 TWh in 2026 (+26.4% vs 447 TWh in 2025); power of 132 GW in 2026 vs 104 GW in 2025, towards 290 GW in 2030; IEA projects rise from 485 TWh in 2025 to 950 TWh in 2030 (reports from June and April 2026)

   → The thesis engine is quantifiable and growing double-digit; it is not narrative, though 2030 forecasts are estimates, not facts

   Gartner (junio 2026) e IEA (abril 2026)

2. GE Vernova: gas turbine backlog confirms the physical bottleneck hecho · alta N1

   2Q26 orders of $24.2B (+88% organic), total backlog of $176B, gas turbines contracted or reserved rising from 100 to 116 GW, target of at least 125 GW at 2026 close, FCF of $5.1B in the quarter and cash of $13.1B (results published 22-Jul-2026)

   → FIRM contracts plus manufacturing slot reservations; annual production rises from 20 GW in 3Q26 to 24 GW in 2028 and targets 30 GW in 2030: 5+ year revenue visibility, rare in industrials

   GE Vernova, 8-K ante la SEC, 22-jul-2026

3. The price of scarcity is already being paid: PJM capacity auctions at the ceiling three times running hecho · alta N1

   The 2028/29 auction (14-Jul-2026) closed at the ceiling of $325/MW-day, third consecutive at the cap; without the cap, PJM estimates it would have closed at $554.72/MW-day, almost $30B in total cost; of the peak increase expected for 2027/28, almost 5,100 MW of 5,250 MW is data center demand

   → Scarcity is structural but the regulatory CEILING limits further upside for generators: incremental margin from auctions is capped until the December 2026 auction

   PJM Interconnection, julio 2026

4. August regulatory blow: Pennsylvania and Texas halt the party hecho · alta N2

   Pennsylvania governor executive order (18-Aug-2026) restricting quick permits and requiring data centers to fund their own electricity; Texas order (3-Aug-2026) to audit interconnection applications; BNP analysts say IPPs are hard to own through November 2026 elections

   → Explains August correction in generators (Talen -11.3% on 18-Aug); it is delay of months, not destruction of demand, but introduces binary political risk through November

   S&P Global Market Intelligence, 19/20-ago-2026

5. Generators have already corrected sharply: what is discounted is being purged hecho · alta N3

   CEG trades more than 51% below its 52-week high despite accumulated contracts; VST at $137 on 30-Aug-2026, -27.5% in a year, with 52-week range 133-220

   → Market over-discounted in 2025 and now discounts regulatory fright: the price-thesis ratio improves just as the headline worsens, classic terrain to buy on corrections if invalidation does not trigger

   Motley Fool 26-ago-2026 e Investing.com 30-ago-2026

6. PPAs signed with hyperscalers: contractual revenue, not promise hecho · alta N2

   Microsoft-Constellation: 20-year PPA and 837 MW to restart a nuclear plant; AWS-Talen expanded to 1,920 MW through 2042; Vistra with Meta deals (January 2026) including option for a 300 MW SMR; Constellation signed ~920 MW of nuclear contracts at 18.5-year average in 2Q26

   → FIRM contracts at 15-20 years that pull capacity from the market and raise the rest's price; note: nuclear restarts (Crane 2028) have execution and schedule risk

   Comunicados de las compañías y prensa sectorial, 2024-2026

7. The inside of the data center also captures: Vertiv and Eaton with record orders hecho · alta N1

   Vertiv: 4Q25 organic orders +252%, book-to-bill ~2.9x and backlog of $15B, +109% year-over-year; Eaton 1Q26: revenues $7.45B (+17%), data center orders +240% year-over-year, backlog $22.8B

   → Power chain inside the building (UPS, liquid cooling, switchgear) is the segment with most growth and least regulatory ceiling; risk: it is the segment most correlated with hyperscaler CAPEX

   Vertiv 8-K SEC 11-feb-2026; datos de resultados Eaton mayo 2026

8. Warning of what is already discounted: the sector fell in block on 28-Jul-2026 inferencia · media N3

   On 28-Jul-2026 fell together AMD -8.1%, Vertiv -6.3%, Eaton -3.1%, and Constellation, Vistra and GE Vernova between -3.7% and -5.4%; GEV went from ~$1,175 (30-Jun) to ~$927 (25-Aug), correction greater than 20%

   → The entire electrical segment of AI trades as ONE single factor: AI CAPEX pace; diversification within the theme is less than it appears and must be sized as a single bet

   Datos de mercado julio-agosto 2026; macrotrends 25-ago-2026

Value chain

• 2026-08-03 ⚖️ Texas governor order: audit of data center interconnections in ERCOT (already happened: pressured IPPs) VST
  Could delay connection of new demand in largest liberalized power market by months
• 2026-08-18 ⚖️ Pennsylvania executive order: mandatory GRID requirements and own-funded electricity for data centers (already happened: Talen -11.3% that day, CEG and VST -4%) TLN
  Pennsylvania is epicenter of nuclear-data center deals (Susquehanna, Crane)
• 2026-09-16 🏛️ Federal Reserve rate decision
  Utilities and power infrastructure among the sectors most sensitive to long real rates
• 2026-09-30 (aproximada) ⚖️ Start of PJM bilateral contracting window before backstop auction CEG
  Window for generators to sign new contracts at scarcity prices; first test after state orders
• 2026-10-21 (aproximada) 📊 GE Vernova 3Q26 results: milestone of 20 GW/year production and progress toward 125 GW GEV
  Quarterly proof that turbine cycle keeps accelerating and is not hitting ceiling
• 2026-10-27 a 2026-11-06 📊 Hyperscaler 3Q26 results (MSFT, GOOGL, META, AMZN) with 2027 CAPEX guidance MSFT
  These four's 2027 CAPEX IS future demand for all this dossier
• 2026-10-28 (aproximada) 📊 Vertiv 3Q26 results VRT
  Most direct thermometer of hyperscaler CAPEX within building
• 2026-11-03 🏛️ U.S. legislative elections (midterms)
  Analysts deem IPPs hard to own through this event due to power bill political risk
• 2026-11-05 (aproximada) 📊 Vistra and Constellation 3Q26 results VST
  Update on PPAs signed and impact of Pennsylvania and Texas orders
• 2026-12-15 ⚖️ Close of PJM capacity auction for 2029/30, last one with agreed ceiling CEG
  If deficit persists and ceiling is renegotiated upward, generators recover pricing leverage

Assets analysed

Risks shared with other dossiers

What would change my mind

• WOULD IMPROVE the thesis: November 2026 election victory of pro-data center construction positions, upward 2027 CAPEX guidance from hyperscalers, GEV backlog above 140 GW, or withdrawal of PJM auction ceiling
• WOULD WORSEN the thesis: visible AI CAPEX cut in October-November, third large state with restrictive order, energy efficiency jump in AI models reducing GW needed per unit of compute, or turbine slot cancellations
• Would change preferred vehicle: if IPPs get blocked politically but construction continues, rotate full weight towards equipment (GEV, VRT, ETN) and contractors (PWR), who collect from CAPEX not power prices

1. The market prices in a hike, not a cut, in September hecho · alta N3

   57% probability of 25 bp hike and 42% to hold as of September 1, 2026; Polymarket assigns 55% to 25 bp hike and 44% to no change, with less than 1% to a cut

   → The assignment thesis is inverted: any portfolio built for an imminent cut starts from a premise that the market rejects with real money

   Kalshi y Polymarket, 1 de septiembre de 2026

2. Starting point of rates and committee split hecho · alta N1

   Target range 3.50%-3.75%; at the July 28-29, 2026 FOMC meeting the committee voted 9-3 to hold, and the three dissenters voted to hike 25 bp, with August 19 minutes showing broad concern about inflation

   → The committee's bias is restrictive; the bar to cut is very high while inflation remains elevated

   FOMC, 29 de julio de 2026; actas 19 de agosto de 2026

3. Jackson Hole changed the price of everything hecho · alta N3

   After Warsh's August 28 speech reaffirming the 2% PCE target, traders raised bets on an imminent hike; gold fell more than 3% Friday, its largest daily decline since June 10

   → The speech functions as a dry run for what a real hike would do: gold and duration suffer first

   Trading Economics y BullionVault, 28-31 de agosto de 2026

4. High inflation from military energy shock hecho · media N4

   General PCE near 4% and core above 3% in recent months, with supply-side pressures from Middle East energy disruptions; the US attacked Iranian drones preparing to mine the Strait of Hormuz

   → Supply-side inflation: raising rates does not fix Hormuz, but Warsh's Fed prioritizes credibility; the macro is stagflationary, the worst world for stocks and bonds at once

   Contexto de mercado Polymarket y Trading Economics, agosto de 2026

5. Employment is breaking: the other side of the binary hecho · alta N1

   July payrolls -23,000 vs +80,000 expected, and annual preliminary revision of -79,000 released August 28; consensus +45,000 for August

   → If the September 4 print comes in clearly negative, hike bets unwind and the duration trade rebounds violently; it is THE event of the fortnight

   BLS vía Trading Economics, 7 y 28 de agosto de 2026

6. The curve warns: short-end flattening hecho · alta N1

   The 10-year-2-year spread compressed to 0.39 bp on August 28 from 0.47 on August 27; the 10-year closed August at 4.72% and analysis points to 5% if the hike is confirmed

   → Typical seller flattening of a late restrictive cycle; bad for REITs and pre-earnings biotech near-term, but plants the recession that would bring the real cut in 2027

   FRED y Investing.com, 28-31 de agosto de 2026

7. The ECB is not the ECB from the script either: it is hiking hecho · alta N1

   The June 25 bp hike of 2026 was unanimous and without reservations; deposit facility at 2.25% with next decision on September 10; eurozone inflation at 2.9% in July vs 2.8% in June

   → The European leg of the original thesis (a cut helps European equities) is also inverted: the monetary wind blows against on both sides of the Atlantic

   BCE (rueda de prensa 11 de junio de 2026) y Eurostat, julio de 2026

8. Gold corrects from highs but structural support remains hecho · alta N3

   Gold at 4,446.83 dollars per ounce on August 31, 2026, far from the record of 5,602.22 dollars on January 28, 2026; still, it closes August with gains above 10%, driven by the monetary degradation trade

   → 20% correction from January but intact underlying demand; tactical risk is a Fed hike, the strategic option remains alive

   JM Bullion, APMEX y Trading Economics, 28-31 de agosto de 2026

9. Small-cap biotech prices the relief that has not yet arrived descontado · alta N3

   XBI accumulated +31.44% in 2026 (Q1 +4.76%, Q2 +24.01%) and traded at 169.27 dollars on August 25, approaching its 2021 high of 174.79

   → The rally rests on M&A and real clinical data but also on expectations of cheap financing; a Fed hike hits just its weak point, the cost of capital

   ETFreplay y Seeking Alpha, 25-31 de agosto de 2026

Value chain

• 2026-08-28 🏛️ Jackson Hole: Warsh's speech reaffirming the 2% PCE target; gold fell 70 dollars and hike probability jumped from 40% to 57% GOLDCFD
  Already passed: set the restrictive bias that arrives at September
• 2026-08-28 🏛️ Preliminary annual revision of payrolls: -79,000 jobs additionally eliminated TLT-proxy
  Already passed: confirms that the labor market is weaker than reported
• 2026-09-01 🏛️ July JOLTS job openings
  First labor market thermometer of the crucial week
• 2026-09-04 🏛️ August non-farm payrolls (consensus +45,000 after -23,000 in July) TLT-proxy
  It is the binary switch: negative dismantles the Fed hike and relaunches duration, gold and biotech
• 2026-09-10 🏛️ ECB decision with new projections; market expects possible second hike from 2.25% EUSTX50
  Defines whether Europe enters its own restrictive cycle in the middle of an energy shock
• 2026-09-10 🏛️ August PPI in the US
  Prelude to CPI; cost pressures from military energy
• 2026-09-11 🏛️ August CPI in the US XBI
  Last inflation print before the FOMC; above 4% year-over-year seals the hike
• 2026-09-16 🏛️ FOMC decision plus dot plot and projections (announcement Sep-16 at 2:00 PM ET) TLT-proxy
  The central event of the dossier: 57% priced-in hike, what moves the market will be guidance on October and December
• 2026-09-17 🏛️ Bank of England decision
  Third leg of central bank week; confirms or breaks the global restrictive coordination
• 2026-09-30 🏛️ August PCE, the Fed's preferred inflation measure GOLDCFD
  Validates or refutes Warsh's diagnosis of inflation near 4% and sets expectations for the October 28 FOMC

Assets analysed

Risks shared with other dossiers

What would change my mind

• The pivot thesis would improve (and would raise XBI, VNQ and TLT to COMPLEMENTARY or higher) with negative employment data on September 4 followed by moderate CPI on September 11: then there would indeed be a cut on the 2027 horizon and duration assets would have real upside
• It would worsen to widespread AVOIDANCE a hike on September 16 with the dot plot pointing to more: the 10-year above 5% would invalidate buying any duration until capitulation is seen
• A verifiable military de-escalation would change the entire picture: less energy inflation, less pressure to hike, and the pivot would return to live thesis instead of outdated narrative

1. Fed on hold but not dovish: the assignment's premise limps hecho · alta N2

   Rate at 3.50-3.75% after the fifth consecutive hold under Chair Kevin Warsh on July 29, 2026, with the June dot plot median at 3.8%; in the June dot plot nine members projected at least one hike in 2026, eight no change, and only one a cut; the market gives ~65% to a 25bp cut before year-end and 2-5% to a hike per meeting: the debate is between one cut and none

   → The environment is one of pause with hawkish tail risk, not guaranteed cuts; Asian carry depends on the September 16 dot plot not resurrecting the hike

   FedWatch/Schwab/primerates, 28-31 agosto 2026

2. Yuan at 3.5-year highs but the central bank is slowing it hecho · alta N2

   USD/CNY at 6.72 on September 1, 2026, with the yuan up 5.86% in 12 months; on August 25 the yuan retreated from the 3.5-year high after the PBOC signaled discomfort with the speed of gains; ING maintains a 6.67-6.92 band for the rest of 2026 and warns that positioning (very loaded in favor of the yuan) is the biggest risk

   → The Chinese currency leg is mature and with crowded positioning: residual room limited and managed by the PBOC

   Trading Economics / Reuters vía Business Recorder / ING, 25-ago a 1-sep 2026

3. China is duration and safe haven, NOT carry: negative spread of 298bp hecho · alta N2

   10-year Chinese bond at 1.68% vs. Treasury at 4.66% on August 25, 2026: spread of -2.98%; low Chinese yields amid a global yield spike reflect an economy isolated from global capital markets and in a deflationary environment

   → Buying Chinese bonds is betting on yuan appreciation and diversification, not carry; with FX hedge the position loses money vs. cash in dollars

   MacroMicro / CNBC, 19-25 agosto 2026

4. Asian FX strength is partly administered, not organic hecho · alta N2

   Japanese Minister Katayama and the US Treasury (Bessent) coordinated intervention to support the yen from historic lows, and South Korea reportedly intervened in parallel, with the Treasury acknowledging it acted to contain Asian currency risks; Japan reportedly spent about 32 billion dollars on the late-July intervention that moved the yen from ~160 to ~155

   → The narrative of strong Asian currencies is incorrect for half the region: stabilization is recent, fragile, and dependent on official intervention

   Nikkei Asia, 3-7 agosto 2026

5. Indonesia: from Asia's worst market to largest inflows since 2019 hecho · alta N1

   The rupiah touched a historic high of 18,279 per dollar in July 2026; BI held its rate at 5.75% on August 19 after raising 100bp in the year to defend the currency, in the first decision by acting governor Destry Damayanti; global funds bought 656.7 million dollars net of debt in a single day (largest figure since July 2019), with 931.7 million net in August, third consecutive month of inflows; 10-year at ~7.2% (7.198% at August 17 close, +100bp in the year)

   → It is the region's high carry with stabilization catalyst already underway; the risk is fiscal and credibility of the new BI leadership

   Bloomberg / AsianBondsOnline (ADB), 17-24 agosto 2026

6. India: structural tax reform that already moved record flows hecho · alta N1

   The ordinance of June 5, 2026 eliminated the 20% withholding on interest and 12.5% capital gains tax for foreign investors in government bonds, with retroactive effect from April 1; in June foreigners bought 418 billion rupees (4,400 million dollars) via the fully accessible route, nearly double the prior monthly record; 7,700 million dollars in the year, already above all of 2025; the rupee recovered from the historic low of ~97 to ~94.7 at end-June close; inclusion in the Bloomberg Global Aggregate could attract up to 27,000 million through 2028

   → Real and dated structural catalyst; part already priced in by the June record, but index flows from 2027-2028 are not yet in price

   Gazette oficial India / Bloomberg / CNBC / CCIL, junio-julio 2026

7. The Asian aggregate yields little: China weighs too much in the index hecho · media N3

   The Emerging Markets Asia Local Currencies index was only up 0.8% at the start of 2026 due to high weight in Chinese debt (bonds nearly flat) and Indian rupee decline

   → Buying Asia local in passive basket dilutes carry in Chinese yields of 1.2-1.7%; need to go by country or with active management

   elEconomista sobre índices JPM, febrero 2026

8. Global EM carry already ran: beware of arriving late descontado · alta N3

   EM local debt returned 19.3% in dollars in 2025, the best since 2009; in Q2 2026 the GBI-EM Global Diversified added +3.85% in dollars with index yield of 6.10% at June 30; Asian real rates remain above historic norm, which along with carry should sustain returns

   → The bulk of the rerating was 2025; what remains is carry plus local cuts, not another 19% revaluation

   Vontobel / State Street Q2 2026 / Eastspring

9. Passive flows to come: South Korea enters the WGBI hecho · media N3

   South Korean government bonds enter the FTSE Russell World Government Bond Index from April 2026 (India entered the JPM GBI-EM in 2024)

   → Index inclusion is multi-year structural demand for Asian local debt, independent of the Fed cycle

   Eastspring / FTSE Russell

10. Governance risk in Indonesia stated as unresolved hecho · media N3

    Indonesian central bank governor resignation increased monetary policy uncertainty (August 2026), although the XI Commission approved Destry Damayanti as governor 2026-2031

    → Indonesian risk premium may reopen; monitor early decisions of the new governor

    Seeking Alpha monthly / Trading Economics, agosto 2026

Value chain

• 2026-08-19 🏛️ Bank Indonesia held its rate at 5.75% in the first decision by Destry Damayanti, who promised continuity and rupiah defense
  The new governor held at 5.75% and committed to rupiah stability, a necessary condition for IDR carry
• 2026-08-25 🏛️ PBOC signaled discomfort with the speed of yuan appreciation; the yuan retreated from 3.5-year highs CBON
  Marks the tactical ceiling of the Chinese currency leg in the near term
• 2026-08-31 🏛️ Official composite PMI for China at 49.5 in August (manufacturing 49.8, services 49.0), still in contraction CBON
  Weak economy sustains low Chinese yields and pressures for more stimulus in the five-year plan
• 2026-09-04 🏛️ US employment report (August)
  Weak data cements the Fed pause and feeds carry; strong data reopens the hike
• 2026-09-10 aprox 🏛️ US CPI for August
  With July PCE at 3.7% year-over-year, another upside surprise would reactivate rate hike bets
• 2026-09-16 🏛️ FOMC with new dot plot under Warsh LEMB
  It is THE event of the quarter: decides if the environment is carry (pause/cut) or pain (upside bias)
• 2026-09-17 aprox 🏛️ Bank Indonesia rate decision (September)
  Third consecutive hold at 5.75% would validate the stabilization thesis that attracts record flows
• 2026-09-21 aprox 🏛️ Monthly LPR fixing in China CBON
  An additional cut would push Chinese yields lower (positive for price, pressure on the yuan)
• 2026-10 aprox 🏛️ China autumn plenary with new five-year plan launch and possible fiscal package CBON
  Large stimulus would raise Chinese yields (bad for bond price) but support the yuan and the rest of Asia
• 2026-10-01 aprox 🏛️ RBI decision (India) and monthly FPI flow data in bonds (CCIL/NSDL)
  Confirm that post-reform inflows sustain and were not a one-time spike

Assets analysed

Risks shared with other dossiers

What would change my mind

• WOULD IMPROVE to BUY_ON_DIPS or higher if the September 16 dot plot eliminates rate hike projections and the median points to cuts in 2027, with USD/IDR below 17,500 and third month of inflows to India
• WOULD WORSEN to AVOID across the board if the Fed regains upside bias, US PCE exceeds 4%, or Indonesia suffers two consecutive months of net outflows with the rupiah back above 18,300
• A large and EXECUTED Chinese fiscal package in October would change the mix: worse for Chinese bond price, better for the yuan and for carry in the rest of Asia

1. JGB yields at historic levels across the entire curve hecho · alta N2

   On Sep 1, 2026 the 10-year touched 3.00% for the first time since September 1996; the 5-year marked an all-time high at 2.265% and the 2-year a 31-year high at 1.795%

   → The entire curve reprices the end of zero rates: direct tailwind for banking intermediation margin, but also a signal of fiscal stress

   Reuters, 1-sep-2026

2. BoJ at 1.00% with clear bias to accelerate hecho · alta N1

   The BoJ raised to 1% on Jun 16, 2026, first time since 1995; held in July by 8-1 with Takata pushing for 1.25%; three sources say it will hike as soon as September and is studying faster pace from the current two annual hikes

   → The faster hike cycle is the engine of the banking thesis; each +25 bp adds tens of billions of yen to megabank NII

   BoJ 16-jun-2026 (comunicado oficial), CNBC 31-jul-2026, Reuters 14-ago-2026

3. September hike mostly discounted descontado · alta N2

   As of Aug 31, 2026 the market prices 74% probability of a hike at the Sep 18 meeting; Commerzbank puts it around 84%; Deputy Governor Himino said on Aug 27 it would be evaluated for every meeting, including September

   → The September hike is already priced: the real catalyst is the signal on SUBSEQUENT PACE (quarterly vs semi-annual), not the hike itself

   Pricing TONA / Commerzbank / Jiji Press, 27 a 31-ago-2026

4. MUFG: record quarterly profit and beating consensus hecho · alta N1

   Net profit of ¥809,400M in Apr-Jun 2026, +48% YoY versus ¥633,400M expected, on track for another record year; maintains ¥2.7 trillion annual target and ¥96 dividend forecast

   → The margin thesis is being fulfilled in P&L, not just in the narrative

   Japan Times / resultados MUFG, 3-ago-2026

5. SMFG: same pattern, more dependent on NII hecho · alta N1

   Attributable profit of ¥501,372M in Apr-Jun 2026, with consolidated gross profit of ¥1.40 trillion, +¥315,684M YoY, from higher interest, fee and trading income; its ratio of interest income to core revenue is 56.8%, the highest of the three megabanks

   → SMFG is the megabank with purest leverage to domestic rate hikes

   Filing 6-K SMFG, 31-jul-2026; análisis ir-tracker ago-2026

6. Demanding valuation: multiples near decade highs hecho · media N3

   MUFG P/B around 1.86-1.88, near the 10-year high of 1.9; P/E TTM ~14.9, 36% above its 10-year median (10.91); the TOPIX Banks index had accumulated +54.7% in 12 months with PBR 1.38 (Jun 2026 factsheet)

   → The market already prices the normalization scenario: the margin of safety is in corrections, not at current prices

   GuruFocus / JPX factsheet, jun-ago 2026

7. Short USDJPY as a macro-trade has failed twice hecho · alta N2

   The joint US-Japan intervention in late July was the first since 1998; the yen moved from 163+ to 155 and has already returned to above 159; earlier, in Apr-May 2026, the MOF spent a record ¥11.73 trillion (~$73,000M) after breaking 160; USDJPY closed unchanged at 159.75 on Sep 1, 2026

   → The currency expression of the macro-trade is much more fragile than the equity expression: the yield differential remains huge

   CNBC / Trading Economics / MOF vía Investing, jul-sep 2026

8. The Fed can neutralize the currency leg hecho · media N2

   After hawkish comments from Fed Chair Kevin Warsh, the market prices ~57% probability of a HIKE by the Fed in September, from ~40% a week prior

   → If the Fed hikes at the same time as the BoJ, the differential barely narrows and short USDJPY loses its driver; note: this is the implicit consensus, it can change fast

   Trading Economics, 31-ago-2026

9. Latent losses in bonds: contained in megabanks, non-trivial in the system hecho · media N3

   Major Japanese banks reported ¥7.05 trillion ($45,000M) in unrealized JGB losses in Oct-Dec 2025, with Japan Post Bank and MUFG as largest increases; for the three megabanks latent loss was below 9% of estimated annual profit; MUFG cut its JGB portfolio to ¥14.56 trillion from ¥25.07 trillion between Mar 2024 and Dec 2025

   → Balance sheet risk from duration is managed in megabanks (sold duration early); weak link is Japan Post Bank, trust banks and regionals

   Risk.net (may-2026), Morningstar (feb-2026), S&P Global MI (may-2026)

10. Takaichi fiscal risk: rising yields has dual reading inferencia · media N2

    Yield surge reflects doubts about PM Takaichi's capacity to balance fiscal responsibility with her semiconductor and AI investment plan; the bond market has been sending a warning against fiscal expansion

    → Must distinguish yields high from normalization (good for banks) from yields high from fiscal risk premium (bad for all, including banks)

    Reuters / BNP AM, 1-sep-2026

Value chain

• 2026-08-27 🏛️ Himino (BoJ) speech opening door to September
  Last official hint before the meeting; avoided committing to pace
• 2026-08-28 🏛️ Tokyo CPI rising in August and 5-year at record 2.195%; reinforced bet on September hike
  August Tokyo inflation data buttressed the case for BoJ hike the following month
• 2026-09-01 🏛️ 10-year JGB touches 3.00%, first time since 1996
  Psychological milestone mixing normalization with fiscal warning to Takaichi
• 2026-09-10 🏛️ ECB meeting
  Market discounts ~86% hike; global context of simultaneous tightening
• 2026-09-16 (aprox) 🏛️ Fed FOMC (possible first hike in 3 years)
  Fed hike sustains dollar and kills the currency leg of the trade
• 2026-09-18 🏛️ BoJ decision: hike to 1.25% and, crucially, signal on pace 8306/8316/8411
  Central event of the file: hike is discounted, quarterly pace signal is not fully priced
• 2026-10-30 🏛️ BoJ October meeting with Outlook Report
  If September stays on hold, October becomes the binary event
• 2026-11 (mediados, por confirmar) 📊 Semi-annual results (H1 FY Mar 2027) for MUFG, SMFG and Mizuho 8306/8316/8411
  Test of annual guidance and possible announcement of new buybacks/dividend raises
• 2026-12-18 🏛️ BoJ December meeting
  If hike in September, additional hike in December would confirm quarterly pace

Assets analysed

Risks shared with other dossiers

What would change my mind

• WOULD IMPROVE to COMPLEMENTARY/CORE if a 15-20% correction in banks brought MUFG P/B to ~1.4x with the BoJ cycle intact, or if the BoJ confirms quarterly pace with stable credit costs
• WOULD WORSEN to AVOID if rising yields mutate into a fiscal crisis (failed auctions, 30-year >4.5% disorderly), if the BoJ slows the cycle from market shock, or if domestic NPLs break guidance
• On the FX leg: would move short USDJPY to TACTICAL only if the Fed rules out hikes AND the BoJ signals quarterly pace in the same week

1. The 1.5% NATO exists and is official, with 2035 horizon. hecho · alta N1

   At The Hague summit in 2025, allies committed to investing 5% of annual GDP by 2035, with up to 1.5% of annual GDP for, among others, protecting critical infrastructure, defending networks, ensuring civil preparedness and innovating, with annual plans showing a credible and incremental path.

   → Real multi-year policy framework; but it is a ceiling (up to 1.5%), not a floor, and the timeframe is 2035: gradual flow, not demand shock.

   OTAN (web oficial), consultado 2026-09-01

2. The 1.5% has no standardized definition: risk of creative accounting. hecho · alta N2

   There is no standardized definition of civil investments countable within the 1.5% tranche; it is expected to include critical infrastructure protection and logistics nodes, but there are warnings that lack of standardization may encourage creative accounting, and the Bertelsmann Stiftung think tank warned that the tranche remains conceptually broad, functionally vague and open to national interpretation.

   → Part of the 1.5% will be computed with pre-existing spending (roads, bridges, intelligence); do not mechanically translate GDP into cyber-listed revenues.

   The Record / Bertelsmann Stiftung, junio-julio 2025

3. Thales: cyber is NOT the engine; it is ~6% of sales and grows little. hecho · alta N1

   H1 2026 group sales €10,949M (+6.7%); Cyber & Digital €1,797M (-2.8%), of which Cyber €681M (-2.2% reported, +1.6% organic), versus Defense €6,316M (+12.9%); adjusted EBIT of segment €195M with 10.9% margin (14.4% in H1 2025) due to Digital Identity one-offs; Cyber EBIT rose +5.5% organic with stable 14.0% margin.

   → Buying Thales as a cyber play is a purity error: you buy hard defense with a low-growth cyber appendix.

   Thales, resultados S1 2026, 2026-07-23

4. secunet: the clearest signal on the topic, orders have surged. hecho · alta N1

   Order intake H1 2026 +78.7% to €287.4M; revenues +19.2% to €204.7M; EBITDA €19.2M (+15.5%); 2026 guidance confirmed at high end of range, following 2025 revenues of €458.8M (+13%) and 2026 guidance of €460-500M; book-to-bill of 1.4x.

   → German state cybersecurity vendor with demand accelerating exactly where the 1.5% points: defense, borders and administration.

   secunet (nota de resultados), 2026-08-13 y 2026-01-29

5. Clavister: the Swedish micro-cap that already turns the topic into profit. hecho · alta N1

   Q2 2026 sales +63% to 89 MSEK, adjusted EBITDA margin record 30%, EBIT margin 17%, second consecutive quarter of net profit and backlog of 642 MSEK (+71% year-over-year); contract of 280 MSEK with the Norwegian defense material agency (NDMA) signed in January 2026, with project of some 36 months; defense revenues +300% in Q1.

   → Maximum purity of European cyber-defense topic, but tiny size: liquidity risk and working capital risk.

   Clavister (informes trimestrales y notas), enero-agosto 2026

6. Clavister warns of the ugly side of defense contracts: cash comes by milestone. hecho · alta N1

   Management noted that growth raises working capital needs, especially since large defense contracts pay by milestone; the first Norwegian milestone, somewhat over 60 MSEK, is expected in autumn.

   → Order growth is not immediate cash; monitor treasury (68.5 MSEK in March 2026) and possible expansion if it scales faster than expected.

   Clavister, conference call Q2 2026, 2026-08-20

7. Sopra Steria: raises guidance citing cyber and sovereignty, and creates a space-cyber business of €200M. hecho · alta N1

   Raised its 2026-07 annual growth target, driven by AI spending, cybersecurity and sovereign technology; Defense, Security and Space grew +10% in H1 (and +14% in Q2); in May it completed Starion and Nexova, creating a European space and cybersecurity business with over €200M in revenues and near 2,000 employees; operating margin of activity 9.6% (+0.4 pts).

   → Via services: lower purity (~13% of group in the vertical) but valuation much less stretched than weapons makers.

   Sopra Steria / Reuters, 2026-07-29

8. Indra: defense crushes it but cyber is accessory and the price is already at highs. hecho · alta N2

   H1 2026 revenues of €3,179M (+29.7%), with defense +103% and space +398%; historical high of 65.96€ intraday on August 13, with +37.6% since July; close of August 24 at 60.56€ after CJEU precautionary rejection, with average price target of 67.22€ (+11%).

   → Cyber exposure only indirect (Minsait grew +4% in Q2, the slow part of the group); the real engine is vehicles, radars and Eurofighter, and that already trades.

   Indra / Infodefensa / Benzinga España, julio-agosto 2026

9. The European cyber-defense market would grow 19% annually, but the source is a consultant. hipotesis · media N3

   European defense cybersecurity market valued at $6,663M in 2026 and projected to $16,009M in 2031, CAGR 19.2%; contrast with observed data: European cybersecurity revenues grew 10% year-over-year in April 2026 (CONTEXT panel), with identity and access management at 18%.

   → Current observed growth is 10%, not 19%; use the low number to value and treat 19% as upside scenario.

   MarketsandMarkets (proyección) y CONTEXT vía Euronews, 2026-05-29

10. What is ALREADY discounted: the re-rating of European defense. descontado · media N2

    Thales trades at 244.50€ (Aug 31, 2026) with 2026e PER of 28.9x; secunet at 2026e PER of 31.2x and 25.8x 2027e; Indra at ~26x 2026e PER per consensus gathered by Estrategias de Inversión (August 2026); the European defense sector was up only 3.4% year-to-date in July amid doubts about how much promised spending will materialize.

    → The NATO headline is already in the price of the big boys; the edge is in quarter-by-quarter order execution and in less-covered niches.

    MarketScreener / Bloomberg, julio-agosto 2026

11. Risk of spending leakage to US vendors. inferencia · media N3

    Significant portions of new cyber funding could benefit large American tech firms, despite European aspirations to develop own capabilities; in favor of the local, the Commission proposed on January 20, 2026 a package strengthening the EU's ICT supply chain against third-country risks.

    → Sovereignty regulation (revised NIS2) is the catalyst that can shield local vendor share; its process is a date to watch.

    Tripwire (análisis) / Euronews, 2025-2026

12. Atos/Eviden: the cheap cyber asset is a balance-sheet story, not a theme story. hecho · media N2

    2025 revenues slightly above €8,000M meeting target after restructuring, with headcount cut 19% and target to grow 5-7% annually and 10% margin in 2028; after selling Advanced Computing (Bull) to the French State, Eviden is left with cybersecurity, AI vision and mission-critical systems.

    → Deep value optionality with track record of value destruction and massive dilution; only for special situations profiles, not as a theme vehicle.

    Atos / Reuters / Techzine, marzo-junio 2026

Value chain

• 2026-08-13 📊 secunet published H1 2026: orders +78.7%, revenues +19.2%, guidance confirmed at high end. YSN.DE
  Is the most direct validation that European state cyber spending is converting into firm orders.
• 2026-08-20 📊 Clavister published Q2 2026: sales +63%, EBIT 17%, backlog 642 MSEK and warning of working capital pressure. CLAV.ST
  Confirms the profit inflection but introduces cash risk from milestone payments.
• 2026-08-25 ⚖️ CJEU precautionary rejection of Indra (SEAE case), European legal dispute remains open. IDR.MC
  First scratch on Indra's perfect narrative at historical highs.
• 2026-10-20 📊 Thales, 9M 2026 activity (orders and sales by segment). HO.PA
  Verify that H1 cyber recovery (+4.5%) was not a mirage.
• 2026-10-28 (aproximada) 📊 Sopra Steria, Q3 2026 revenues. SOP.PA
  First read of defense-security-space vertical with Starion and Nexova fully consolidated for the quarter.
• 2026-10-30 (aproximada) 📊 Indra, 9M 2026 results. IDR.MC
  With the stock near highs, any margin disappointment is priced dearly.
• 2026-10/11 (aproximada) 📌 Collection of first milestone of Clavister's Norwegian contract (>60 MSEK). CLAV.ST
  Clears or confirms the treasury risk of the investment case.
• 2026-11-12 📊 Clavister, Q3 2026 results. CLAV.ST
  Monitor backlog, cash and follow-on orders from Saab/BAE.
• 2026-11-12 (aproximada) 📊 secunet, Q3 2026 quarterly release. YSN.DE
  Q4 concentrates annual profit: Q3 signals early confirmation of guidance high end.
• 2026-11/12 (aproximada) ⚖️ Commission cyber package process (revised NIS2 proposed Jan 20, 2026) and 2027 defense budgets of Germany, France and Spain.
  Determines how much of the 1.5% NATO converts into contracts reserved for European vendors.

Assets analysed

Risks shared with other dossiers

What would change my mind

• Would improve the thesis: first national guidance (Germany or France) detailing concrete multi-year tranches of the 1.5% with tenders reserved for European vendors; or Thales-Cyber accelerating to double-digit organic for the whole of 2026.
• Would also improve: 15-20% correction in secunet or Thales without order deterioration (entry at better price).
• Would worsen: lasting ceasefire in Ukraine with budget path cuts; book-to-bill <1x at secunet; dilutive expansion at Clavister; or evidence that the 1.5% is mostly computed with pre-existing spending.
• Would worsen structurally: large European cyber-infrastructure awards to hyperscalers and US vendors despite sovereignty rhetoric.

1. Institutional flow to bitcoin ETFs erases year-to-date outflows hecho · alta N3

   $314M on Aug 26 2026, seventh consecutive day, August at $3.030B and net outflows for 2026 reduced to $2.260B

   → The marginal buyer has returned via the regulated channel; flow is real but recent and reversible if Fed hikes on Sep 16

   SoSoValue vía Cointelegraph, 26-ago-2026

2. BTC rebounds sharply but still far from highs hecho · alta N1

   ~$78,100 on Aug 31 2026; +26.5% in a month and -27.7% year-over-year; price moved from $87,549 on Dec 31 2025 to $58,745 on Jun 30 2026 (Grayscale 10-Q, N1)

   → August rebound already discounts much of the flow relief; buying today is paying the rally, not anticipating it

   CoinMarketCap/Crypto.com 31-ago-2026 y 10-Q GBTC 30-jun-2026

3. Dual binary on September 15-16: CLARITY and FOMC hecho · alta N2

   CLARITY Act cloture on Sep 15 2026 at 14:15 ET, requires 60 votes and failure could kill the bill per Reuters; Polymarket prices it at 13-14% and Kalshi the broad contract at 20-21%. After Warsh's Jackson Hole (Aug 28) the probability of 25bp hike in September is 57% over current 3.50%-3.75% range

   → Two events with uncertain outcome in 48 hours; prediction market says CLARITY likely fails, so asymmetric surprise would be if it passes

   Latham/Reuters/Kalshi/CME, 27-29 ago 2026

4. The 2026 Fed debates HIKING, not cutting: macro regime inverted for crypto hecho · alta N2

   Kalshi moved from ~70% pause to 48% hike after Warsh's hawkish speech committed to fighting inflation; JP Morgan WM already has +25bp as base case from supply shocks linked to Iran

   → Rising real rates are headwind for zero-cash-flow assets; but raise reserve income for Circle: same event hits differently inside the sector

   CNBC 28-ago-2026; J.P. Morgan WM ago-2026

5. Changing correlation: BTC trades like gold-with-beta, capped by dollar inferencia · media N4

   Investors buy BTC and gold together as hedge against weak dollar and persistent inflation, while dollar strength with broken yen above 160 is the force that limits crypto (Aug 31 2026)

   → Hedge narrative coexists with risk asset reality: in a rate scare BTC would fall with equities, not with gold

   KuCoin/Odaily 29-ago y Coinbase live 31-ago-2026

6. Circle: the real business of the theme grows, but is a single-rate bet hecho · alta N1

   USDC in circulation of $73.3B (+19% year-over-year) and on-chain volume of $14.8T (+151%) in Q2 2026; net income of $48M and adjusted EBITDA of $143M (+8%); 95% of revenue still comes from reserves ($668M) and total figure of $701M came below estimates

   → Circulation grows despite crypto market down 40%; a hiking Fed improves reserve income short-term, opposite to the rest of the sector

   Nota de resultados Circle en SEC, 05-ago-2026

7. MiCA gave Europe to Circle hecho · alta N3

   Jul 1 2026 MiCA's last transition period ended; Tether withdrew USDT from regulated European market and Circle says USDC is the only top-10 stablecoin that complies; Brussels equivalence consultation closes Aug 31 2026 with possible extension to Sep 30

   → Tangible regulatory advantage for USDC in EU; risk is MiCA revision via equivalence reopens door to Tether

   Prensa especializada UE y CNMV, jul-ago 2026

8. MicroStrategy returns to buying, but model no longer carries premium hecho · alta N2

   Bought 4,603 BTC for $369.7M ($80,318 average) between Aug 24-30 financed with sale of 4.53M shares; holds 845,050 BTC at $75,412 average cost and trades at mNAV ~1.01x; in June 2026 mNAV fell below 1.0x

   → Without NAV premium the engine of issuing-expensive-to-buy-BTC doesn't create value; purchase was financed with pure dilution at ~1x

   8-K vía Investing.com, 31-ago/01-sep-2026

9. Miners are no longer crypto: they are AI data centers with firm contracts hecho · alta N1

   IREN raised cloud AI ARR target to over $4.0B by year-end with ~85% already under contract (Jul 20 2026); liquidates mined bitcoin daily and had zero BTC on balance at Jun 30 2026 (10-K, N1); Hut 8 signed a $9.8B 15-year lease

   → Exposure to BTC via miners has died as a thesis; what remains is an AI CAPEX bet that duplicates the AI infrastructure brief

   10-K IREN 30-jun-2026 y CoinDesk 20-jul-2026

10. Flow to ether is the new data point of the month hecho · alta N3

    $225.8M on Aug 27, best day since Oct 2025, and $1.420B cumulative since Aug 17; ETH still 55-58% below its ~$4,950 August 2025 high

    → ETH recovery beta is greater than BTC's, but with higher volatility and reversal risk

    Farside/SoSoValue vía CryptoBriefing, 29-ago-2026

Value chain

• 2026-08-28 🏛️ Jackson Hole: hawkish speech from Fed Chair Kevin Warsh that spiked hike probability to ~50-57% (already passed)
  Changed September event price from clear pause to coin flip with hike bias
• 2026-08-31 ⚖️ Closing of EU consultation on MiCA review and equivalence regime (possible extension to Sep 30) CRCL
  Decides whether Tether can return to Europe and erode USDC's regulatory monopoly
• 2026-08-31 📌 MicroStrategy announcement: resumes buying with 4,603 BTC after ten-week pause (already passed) MSTR
  Largest corporate buyer returns to market, though financed with dilution at mNAV ~1x
• 2026-09-04 🏛️ US employment report (with JOLTS and productivity earlier in week)
  Data that could tilt FOMC: strong employment reinforces hike and penalizes crypto
• 2026-09-15 ⚖️ CLARITY Act cloture vote in Senate (14:15 ET; 60 votes; markets price it at 13-21%) COIN, CRCL, IBIT
  Failure could kill US regulatory framework until after elections; success would be month's bullish surprise
• 2026-09-16 🏛️ FOMC decision (current range 3.50-3.75%; ~55% probability of +25bp)
  Hike raises real rates and hits crypto beta hard just one day after CLARITY vote
• 2026-09-30 ⚖️ Possible MiCA consultation closing extension CRCL
  Sets definitive course for European stablecoin market
• aproximada 2026-09-11 🏛️ US August CPI
  Last inflation reading before FOMC with Iran energy shocks on the table
• aproximada principios de noviembre 2026 📊 Circle Q3 2026 results CRCL
  First read on effect of possible rate hike on reserve income and Arc ramp
• aproximada septiembre 2026 🚀 Arc mainnet launch, Circle's blockchain CRCL
  First serious lever to diversify 95% of revenue today dependent on reserves

Assets analysed

Risks shared with other dossiers

What would change my mind

• IMPROVES if CLARITY Act cloture exceeds 60 votes on Sep 15 (market gives it 13-21%: would be the asymmetric surprise) and Fed pauses on Sep 16: IBIT would move to COMPLEMENTARY and CRCL conviction would rise to 7
• IMPROVES if EU maintains European bank reserve requirement in MiCA review, consolidating USDC's moat
• WORSENS if Fed hikes and ETF flow turns negative two weeks later: would return to AVOID all directional exposure and maintain only CRCL
• WORSENS if MicroStrategy sells BTC again to pay preferreds: signal of structural stress in proxies that would taint sentiment of the entire theme

1. Joint US-Japan intervention quantified and overrun hecho · alta N1

   $96.4bn aggregated between 30-Jul and 26-Aug 2026, after record ¥11.73 trillion (~$73bn) in Apr-May; USDJPY closed August at 159.8-160, returning almost all

   → The 160-164 ceiling is politically defended but ammunition spent without results erodes credibility; the 164 level is the real line of fire

   MOF Japon/Tesoro EEUU confirmacion 3-ago; FXStreet 31-ago; Investing 27-jul

2. BoJ 17-18 Sep: hike to 1.25% mostly priced in hecho · alta N2

   OIS ~85% (27-Aug), Doton Research 81% (24-Aug); 8 of 14 Yomiuri economists expect hike; Reuters (14-Aug, 3 sources) talks of accelerating pace above 2 hikes/year

   → The hike itself no longer pays; what pays is the subsequent pace guidance (January as next stop) or disappointment (yen collapses)

   Japan Times 27-ago; Reuters 14-ago; Yomiuri 25-ago

3. Fed 15-16 Sep: real binary at 47% hecho · alta N2

   Polymarket 47% hike 25bp after Warsh speech on 28-Aug; headline PCE ~3.7-4% and core >3%; 3 pro-hike dissenters in July; JPM expects hike, Goldman expects hold all 2026

   → It is the pivot variable for the entire dossier: USDJPY differential is more sensitive to Fed than BoJ; full size forbidden before 16-Sep

   Polymarket 28-ago-1-sep; JPM WM ago-2026; Goldman via Yahoo Finance ~18-ago

4. The carry trade came out reinforced from intervention hecho · alta N2

   Japanese investors bought >5 trillion yen in foreign assets after intervention (CNBC 20-Aug); pensions and managers kept selling yen

   → While Japanese money costs 1% against 3.6-3.75% in the US, every yen rally is sold; USDJPY short pays negative carry (~2.6 points annualized) if it stalls

   CNBC/Monex/State Street 20-21 ago

5. ECB 10-Sep becomes a live meeting hecho · media N2

   Refi at 2.40% after hiking on 11-Jun-2026; market ~50/50 for another hike; eurozone inflation fell to 2.8% with energy decelerating; meeting with new projections

   → Hawkish surprise pushes EURUSD toward 1.17-1.18; hold with soft tone confirms range ceiling; do not anticipate, react

   Calendario decisiones Investing; Cambridge Currencies jul-2026; ECB calendario oficial

6. PBoC actively brakes yuan appreciation hecho · alta N1

   Fixing 6.7841 on 24-Aug, 593 pips weaker than Reuters estimate (almost record deviation); 663 pips on 27-Aug; spot 6.72, 3-year high

   → Managed currency in direction opposite momentum: do not chase yuan appreciation, the central bank sets the pace

   PBoC via TradingEconomics/Reuters 24-27 ago

7. BoE and SNB complete the map: hawkish pause and zero anchor hecho · alta N1

   BoE 3.75% with 6-3 vote (30-Jul), ~25% hike priced for 17-Sep, inflation 2.9% in July; SNB 0% (18-Jun) with increased disposition to intervene and inflation forecast 0.6% in 2026

   → GBP with light carry support within G5; CHF strong but with institutional ceiling — both are range trades, not trend trades

   BoE 30-jul; SNB comunicado oficial 18-jun

8. The dollar as a cycle: flat for the year, no trend hecho · alta N2

   DXY 99.4 on 1-Sep; opened 2026 at 98.4, high 101.5 (27-Jul), low 96.2 (27-Jan); down 10.2% in 2025

   → No directional dollar cycle to exploit: 2026 is a 96-101.5 range year where meeting binaries rule, not trend

   ICE/TradingEconomics 1-sep; Convextrade agosto

9. Narrative to watch: EURUSD 1.25 consensus target narrativa · baja N4

   Goldman and Deutsche Bank target 1.25 year-end conditional on Fed not hiking; the pair has ranged 1.12-1.19 for 12 months

   → Consensus euro bullish and very populated; if Fed hikes on 16-Sep that positioning unravels violently toward 1.13

   Roboforex citando GS/DB, 27-ago

Value chain

• 2026-08-28 🏛️ Warsh speech: current data, 2% target, short rate priority — raised hike probability to 47% USDJPY, EURUSD
  Already occurred: recalibrated Sept binary toward coin flip
• 2026-09-04 🏛️ US August non-farm payrolls EURUSD, USDJPY
  Last large employment data before FOMC; weak NFP kills Fed hike
• 2026-09-10 🏛️ ECB: decision with new projections (meeting at the Bundesbank) EURUSD
  50/50 hike: Europe's binary of the month
• 2026-09-16 🏛️ FOMC: Fed decision with Warsh, 3.50-3.75% current, 47% hike probability todos
  The pivot variable for the entire dossier: decides yen more than BoJ
• 2026-09-17 🏛️ BoE: decision with inflation at 2.9% and prior 6-3 vote GBPUSD
  Fourth dissent or surprise hike fires the pound
• 2026-09-18 🏛️ BoJ: decision, hike to 1.25% priced at 80-85%, pace guidance is key USDJPY
  Disappointing after $170bn intervention would be a Tokyo credibility crisis
• 2026-09-24 🏛️ SNB: quarterly decision, 0% expected USDCHF
  Will confirm if disposition to intervene against strong franc remains elevated
• 2026-10-28 🏛️ BoJ October: second hike window if September disappoints USDJPY
  Market prices ~85% cumulative hike by this date
• 2026-11-03 📌 US legislative elections (midterms) USDJPY, DXY
  US-Japan currency cooperation has a political clock: after midterms Washington's appetite may shift
• aprox 2026-09-11 🏛️ US August CPI EURUSD, USDJPY
  With PCE near 4%, hot CPI seals Fed hike

Assets analysed

Risks shared with other dossiers

What would change my mind

• Would improve the USDJPY short thesis: weak NFP on 4-Sep + cold CPI on 11-Sep that lower Fed hike to <25%, with BoJ intact >80% — then upgrade from tactical to complementary and scale to full size after 16-Sep
• Would worsen: hot CPI that puts Fed hike >65% — close USDJPY short before FOMC, not after
• On EURUSD: I only turn directional if ECB hikes on 10-Sep AND Fed holds on 16-Sep; one leg is not enough, the annual range proves it
• On yuan: five consecutive fixings aligned with consensus would move me from AVOID to TACTICAL short on USDCNH

1. NATO 5% framework remains intact with calendar hecho · alta N1

   At The Hague 2025 summit allies committed to invest 5% of GDP in defense by 2035, minimum 3.5% in core defense, credible annual pathway plans and progress review in 2029; Spain exempted. In 2026 Lithuania spends 5.33% of GDP, Estonia 5.1% and Latvia 4.92%

   → Multi-year demand contracted by political treaty, not by cycle; the eastern flank already executing above target

   OTAN (actualizado 29-jun-2026), CSIS (jul-2026), Statista (2026)

2. SAFE exhausted by excess demand and EU preparing successor hecho · media N2

   SAFE lends up to EUR 150 billion backed by EU budget for joint purchases; first tranche approved released some EUR 38 billion with first disbursements in March 2026; per Bloomberg the program was already depleting in spring 2026 due to exceptionally high demand and successor proposals expected by late spring 2026

   → Bottleneck is not demand but financial instrument; SAFE successor and MFF 2028-2035 are the key regulatory catalyst for coming months

   Consejo UE, Comisión Europea, Bloomberg vía Militarnyi (may-2026)

3. Rheinmetall: explosive growth but cash and guidance disappoint hecho · alta N1

   Q2 2026 sales EUR 3.289 billion (+69.8%) and portfolio EUR 80.4 billion, but stock fell 5.67% due to cash flow weakness, sales outlook cut and inventory increase; H1 FCF negative at EUR 1.6 billion; 2026 guidance cut to EUR 13.7-14.2 billion due to EUR 300 million impact in naval division (F126 cancellation); H1 new orders EUR 16.238 billion (+141%)

   → Confirms the lesson from previous dossier: market no longer pays for orders without cash; 5-Nov-2026 publication is the arbiter

   Resultados Rheinmetall 06-ago-2026 (Infodefensa, Investing)

4. BAE: most balanced combination of growth+cash+return hecho · alta N1

   H1 2026 sales GBP 15.772 billion (+9%), underlying EBIT +11%, EPS +13% and FCF GBP 1.791 billion with raised annual guidance; portfolio GBP 84 billion; buybacks 10% higher year-on-year and interim dividend 15 pence payable 2-Dec-2026

   → Only large European prime combining growth, strong positive cash and direct shareholder return; natural core of the basket

   BAE Systems, resultados semestrales 30-jul-2026

5. Leonardo: orders soaring, but revenue line not rising hecho · alta N1

   H1 orders EUR 16.3 billion (+39%) and portfolio EUR 58.6 billion (+17%), with book-to-bill 1.6x; order guidance raised to around EUR 28.2 billion and EBITA EUR 2.210 billion, but revenues maintained at around EUR 22.1 billion including nine months of IDV; net debt EUR 3.2 billion (+49%) from IDV purchase

   → Portfolio provides 2.6-year visibility but execution (converting orders to revenues) is the bottleneck; monitor industrial capacity in 2027

   Leonardo, resultados 30-jul-2026; Janes 28-ago-2026

6. Thales: the positive cash surprise of the semester hecho · alta N1

   H1 sales EUR 10.9 billion (+7.8% organic), adjusted EBIT +10% and free operating cash flow EUR 1.8 billion, almost triple year-on-year, with partially raised guidance; defense orders +28%, already 59% of new bookings; defense invoices EUR 6.316 billion (+13.1% organic) with EBIT EUR 875 million (+22%)

   → Combined defense+cyber+space exposure with best cash conversion of semester in Europe; medium defense purity (~58%) but high quality

   Thales, resultados 23-jul-2026

7. US: record portfolios and raised guidance in LMT and RTX hecho · alta N1

   Lockheed: Q2 sales +11% to USD 20.1 billion, FCF USD 2.9 billion and record portfolio USD 230 billion including multi-year THAAD interceptor contract; 2026 EPS raised to USD 29.95-30.65; RTX grew revenues 14.5% with record portfolio USD 289 billion (+22%)

   → The US ammunition and air defense cycle accompanies the European one; additionally Trump requested USD 1.5 trillion defense budget for 2027, a 50% increase (proposal, not law: not a budgetary fact)

   Lockheed Martin (nota de prensa y 8-K SEC, 23-jul-2026); resultados RTX jul-2026

8. What the market discounts: stagnation despite results descontado · media N2

   Following the rally started in 2022, the European sector stalled in H1 2026 despite solid results, due to caution on high valuations; on 11-Aug-2026 RBC initiated buy coverage on Rheinmetall, Thales, Leonardo, BAE and Saab, and Barclays raised Saab and initiated Kongsberg, Babcock and Renk on overweight; analysts see 2026 as consolidation year with very selective investors

   → Sector beta already in price; alpha is in cash selection and entries on correction

   Bloomberg vía Oninvest (11-ago-2026), CNBC (30-may-2026)

9. Peace in Ukraine: unlikely in 2026, but it is the tail risk inferencia · media N3

   The Kremlin announced a deep pause in peace negotiations while continuing its offensive in the east (late Aug-2026); Ukrainian military intelligence stated 18-Aug that no agreement will be signed this year; trilateral negotiations suspended since February 2026 after Iran war escalation; prediction markets: 18.5% probability of ceasefire before 31-Dec-2026

   → Peace headline risk that hurt the sector in 2025 is today dampened; moreover European rearmament no longer depends only on Ukraine (Greenland, Iran, hybrid war)

   RBC-Ukraine/NV (18-ago-2026), Kyiv Post (ago-2026), mercados de predicción

10. Drones and anti-drone: from announcement to executable budget hecho · media N2

    Poland launched January 2026 anti-drone wall of around EUR 3.8 billion with PGZ, Kongsberg and Advanced Protection Systems; France announced EUR 8.5 billion to quadruple drone and missile stocks by 2030 and Germany EUR 10 billion in military drones; European anti-drone defense initiative expects initial operating capabilities by end of 2026; on 05-Aug-2026 a drone with explosives was found at Leipzig airport

    → Anti-drone/electronic warfare segment is fastest-growing item; benefits Thales, Leonardo, Hensoldt, Kongsberg and Rheinmetall (effectors)

    ECFR (ago-2026), Euronews (14-abr-2026), Comisión Europea

11. Germany: structural rearmament beyond Ukraine hecho · alta N2

    On 22-Apr-2026 Germany presented its first independent military strategy since WWII, with active personnel growth from 185,000 to 260,000 by mid-2030s and reserve of at least 200,000

    → Anchor of domestic demand for Rheinmetall and Hensoldt over a decade; risk is annual Bundestag budget execution

    Ministerio de Defensa alemán (22-abr-2026)

12. Safran and Rolls-Royce: the expensive part of the story is civil, not defense hecho · alta N1

    Safran H1: revenues EUR 17.571 billion (+19%), record 18.4% margin and FCF EUR 2.616 billion with raised annual guidance, but driven by strong civil engine spares demand; Rolls-Royce H1: underlying operating profit +46% to GBP 2.5 billion with 22.5% margin and FCF GBP 2 billion, with electrical power for data centers as engine

    → Both are excellent companies but their profit engine is civil aftermarket and data centers, not rearmament: indirect/medium thematic purity; do not confuse the dossier

    Safran (28-jul-2026), Rolls-Royce (30-jul-2026)

Value chain

• 2026-08-30 (aproximada) 🏛️ Kremlin announces deep pause in peace negotiations while maintaining offensive in east (already occurred: reduces short-term ceasefire headline risk)
  Sector main downside risk (ceasefire) losing probability near-term
• 2026-10-20 (aproximada) 📊 Lockheed Martin Q3 2026 results LMT
  Confirmation of raised USD 30 EPS guidance and ammunition/THAAD pace
• 2026-10-21 (aproximada) 📊 RTX Q3 2026 results RTX
  Record USD 289 billion portfolio must begin translating to margin
• 2026-11-05 📊 Rheinmetall Q3 2026 results (date confirmed by company) RHM.DE
  Thesis arbiter: operating cash and confirmation of ~EUR 135 billion portfolio target
• diciembre 2026 (aproximada) 🚀 European Anti-Drone Defense Initiative (EDDI) initial operating capacity
  First verifiable anti-drone segment milestone; adjudications to Thales, Leonardo, Hensoldt and Kongsberg
• finales de octubre 2026 (aproximada) 📊 Thales Q3 2026 orders and sales publication HO.PA
  Verify +28% defense order growth not one-time spike
• mediados de septiembre 2026 (aproximada) ⚖️ European Commission State of Union speech with predictable announcements on SAFE successor and MFF 2028-2035
  Defines European rearmament funding post-2027; catalyst for entire European basket
• noviembre 2026 (aproximada) 📊 BAE Systems trading update BA.L
  Tracking raised guidance and H2 order entry
• octubre-diciembre 2026 ⚖️ US FY2027 defense budget processing and European national 2027 budgets (Germany, Poland, France)
  Converts NATO targets into appropriated money; US strong increase proposal still not law
• principios de noviembre 2026 (aproximada) 📊 Leonardo 9M 2026 results LDO.MI
  First full reading of consolidated IDV and portfolio-to-revenue conversion

Assets analysed

Risks shared with other dossiers

What would change my mind

• Would improve the thesis: Rheinmetall positive operating cash in Q3 (5-Nov) plus a SAFE successor endowed with more than EUR 150 billion; elevate Rheinmetall to COMPLEMENTARY and global conviction to 9
• Would worsen the thesis: verifiable ceasefire in Ukraine with troop withdrawal, or a second guidance cut in any European prime due to execution; downgrade basket to TACTICAL and activate hedges via sectors benefiting from peace (European construction)
• Composition change: if Leonardo demonstrates in 9M that revenues accelerate toward portfolio pace, would rival BAE as core by valuation

1. Managers launch ultra-long JGB funds for retail hecho · alta N2

   Mitsubishi UFJ AM joins Daiwa AM and Amova selling ultra-long JGB funds; the 30-year yields near 4%, above German 30-year bund (3.6%) and near US Treasury 30-year (5.2% on 08-12-2026)

   → New product line with recurring commission for managers; signal that the shift broadens from equities to fixed income

   Reuters, 12-08-2026

2. But retail money in these funds is still minimal hecho · alta N2

   Each fund does not exceed 3 billion yen (around $18.8 million) and Amova's 30-year fund had 554 million yen at end of June, growth lower than expected; retail fears yields may keep rising

   → The Reuters headline is multi-year thesis, not material income today; don't pay multiple for this yet

   Reuters, 12-08-2026

3. Record domestic savings and deposits losing weight hecho · alta N1

   Household assets at 2.386 trillion yen at end of March 2026, +7.1% year-on-year; by end of September 2025 cash and deposits accounted for 49.1% of total, below 50% for first time in 18 years, due to shift towards investment via NISA

   → Structural tailwind of the theme: each percentage point exiting deposits is ~24 trillion yen of commissionable assets

   BOJ flow of funds via Bloomberg (25-06-2026) y Japan Today (dic-2025)

4. NISA as political-fiscal engine of the shift hecho · alta N1

   NISA accounts +5 million since end of 2023; one of every four citizens over 18 has an account; government target of 34 million accounts by end of 2027; 87.2% of new NISA users record gains (QUICK survey Oct-2025)

   → Multi-year State policy; positive user experience feeds flows back, though a sharp market fall would test it

   FSA (21-05-2025) y QUICK (encuesta oct-2025)

5. SBI surpasses Nomura in quarterly profit hecho · alta N1

   SBI beat Nomura in pre-tax profit (225,800 vs 211,500 million yen) and attributable profit (148,100 vs 145,600 million) in Apr-Jun 2026; first Japanese broker to exceed 16 million accounts

   → Low-cost digital platform wins volume battle; hybrid bank+broker model (sweep account) is the structural winner of the theme

   Presentacion resultados SBI, 31-07-2026

6. Nomura and Daiwa migrate to recurring income, which is what to watch hecho · alta N1

   Nomura: record recurring income of 59,200 million yen, record net inflows of 539,600 million and record recurring assets of 31.7 trillion, 17 consecutive quarters of net inflows. Daiwa: record balance-linked income of 35,800 million with fixed cost coverage of 125.7%: recurring commissions already cover entire fixed cost base

   → Mix shifts from cyclical (brokerage) to stable (balance commission); deserves gradual re-rating, but part already granted by market

   Resultados Q1 FY2026 (jul-ago 2026)

7. The BOJ withdraws and needs someone to buy JGBs hecho · alta N2

   BOJ expected to reduce JGB holdings by 48 trillion yen this fiscal year and maintain that pace; analysts expect government to raise JGB issuance by 15 trillion yen this year

   → State has incentive to facilitate retail JGB channel for years: regulatory tailwind for platforms

   Reuters, 12-08-2026

8. Market discounts BOJ hike in September descontado · media N3

   Japanese 10-year at 2.95% on 31-08-2026; market assigns ~87% probability of BOJ 25bp hike in September; official rate at 1.0% after June hike, highest since September 1995

   → September hike almost fully priced; asymmetric risk is it does NOT hike (weak yen, brokers ok) or it hikes and hardens message (hit to duration and stock market)

   Trading Economics, 31-08-2026

9. Rakuten fintech reorganization with date hecho · alta N1

   Rakuten Bank to become parent of Rakuten Card and Rakuten Securities HD, effective October 1, 2026 (agreement signed 05-20-2026)

   → Creates third quoted bank+broker+card pole; possible Rakuten Bank re-rating, but with holding complexity and dilution from third-party issuance

   Nota de prensa Rakuten Group, 20-05-2026

Value chain

• 2026-09, aproximada 🚀 Mitsubishi UFJ AM 20-year JGB fund launch 8306.T
  First test of whether retail buys long duration with 30-year yields at highs
• 2026-09-01 🏛️ BOJ bond market survey (August)
  Thermometer of real JGB demand after 10-year at 2.95%
• 2026-09-16 🏛️ Fed rate decision
  Spread with US conditions yen, JGBs and risk appetite in Japan
• 2026-09-17 🏛️ BOJ monetary policy meeting
  Meeting on September 17 and 18 with ~87% probability of hike to 1.25% already priced in; risk is in the surprise
• 2026-09-25, aproximada 🏛️ BOJ flow of funds (Q2 2026): household assets and weight of deposits
  Confirms or denies that deposit-to-product shift keeps accelerating
• 2026-09-30 📌 Payment of third-party allocation of Rakuten Securities HD 5838.T
  Defines real dilution of reorganization
• 2026-10-01 📌 Effective date of Rakuten fintech reorganization 5838.T
  Birth of Japan's third quoted bank+broker+card pole
• 2026-10-27, estimada 📊 Nomura Q2 FY2026 Results 8604.T
  Test of whether 17-quarter streak of recurring net inflows continues
• 2026-10-29 🏛️ BOJ October meeting with outlook report
  If September brings no hike, this is the next window
• 2026-10-30, aproximada 📊 SBI and Daiwa Q2 FY2026 Results 8473.T
  Verify if April-June profit peak was sustainable or cyclical ceiling

Assets analysed

Risks shared with other dossiers

What would change my mind

• WOULD IMPROVE the thesis: 15-20% correction in SBI/Nomura without flow deterioration (entry point), September flow of funds showing acceleration of shift, retail JGB funds capturing real money (more than 100 billion yen aggregate)
• WOULD WORSEN: BOJ forced to halt normalization (rate leg fails), Nikkei correction above 20% with net NISA outflows (flow leg fails), or Japanese fiscal credibility crisis with 30-year above 4.5% disorderly

1. September hike almost fully priced descontado · alta N2

   Traders assigned ~80% probability to a hike at the 18-Sep-2026 meeting (24-Aug-2026); Reuters (14-Aug-2026) cited three BOJ sources contemplating hiking in September and accelerating pace from two hikes per year

   → The event itself offers little alpha; what drives the sector is the pace signal (guidance toward 1.5%+ in 2027) in Ueda's press conference

   Bloomberg/Japan Times 25-ago-2026; Reuters 14-ago-2026

2. Starting point: 1.00% since June, pause in July with hawk dissent hecho · alta N2

   The BOJ hiked to 1.00% on 16-Jun-2026 (7-1, highest since 1995) and held on 31-Jul by 8-1, with Takata voting to hike to 1.25%; the bank warns core inflation clearly above 2%

   → Normalization cycle intact and biased to accelerate; the short end reprices the regional asset (loans tied to prime and reserve remuneration)

   CNBC 16-jun-2026 y 31-jul-2026; Trading Economics jul-2026

3. Profits already reflect the cycle: +36.9% in the April-June 2026 quarter hecho · alta N3

   The 73 listed regional banks earned 597,900 M JPY in Apr-Jun-2026 (+36.9% year-over-year); 61 of 73 improved profit and only 3 posted losses; results published through 14-Aug-2026

   → Operating leverage to short rates is real and measurable, but consensus that already knows it: beware paying peak multiple on peak profit

   Agregado de resultados trimestrales (via BigGo/prensa japonesa) 17-ago-2026

4. Deposit beta remains low but rises step by step hecho · media N3

   After the hike to 1.00% (Jun-2026), megabanks and large regionals (Bank of Fukuoka from 3-Aug-2026) raised ordinary deposit rates from 0.30% to 0.40%; average deposit rate was 0.33% in Jul-2026 with policy rate at 1.00%; online banks pay up to 1.0% (Aozora) and 1.1% conditionally

   → Each 25 bp hike adds margin while beta stays ~40%; if online competition forces betas >60%, the thesis degrades quickly for banks with less loyal deposits

   Nota sectorial japonesa jun-2026; BOJ via Trading Economics jul-2026; Japan Times 20-dic-2025

5. The deposit filter is essential: a third of regional banks lose deposits hecho · alta N1

   BOJ Financial System Report (Apr-2026): ~30% of regional banks and ~50% of shinkin show year-over-year deposit decline (data to Feb-2026); regional deposit share continues declining due to aging, inheritances migrating to large banks and NISA (~27 M accounts); loan-deposit gap at 108 bn JPY in Dec-2025, four-year low

   → Dispersion will be huge: those losing deposits must pay for them or shrink balance sheet; buying the sector as indiscriminate basket is the wrong way to play the theme

   Bank of Japan FSR 21-abr-2026 (N1); The Diplomat feb-2026

6. Latent losses in bonds: the hidden liability of the cycle hecho · alta N2

   Record latent losses of regional banks in domestic bonds: 21,300 M USD at 30-Sep-2025 (+260% since Mar-2024); large banks: 7.05 bn JPY in JGB at Dec-2025; Bloomberg (24-May-2026) notes that yield rebound widens valuation gap between regionals with weak and strong portfolios

   → The same hike that fattens margin depresses book value for those who bought duration late; require short portfolio or latent losses <15% of capital as purchase criterion

   Risk.net abr-2026; Bloomberg 24-may-2026; datos agregados sep-2025

7. The revaluation already occurred closes the gap with megabanks hecho · media N3

   PBR at Jun-2026: Chiba 1.52x, Shizuoka FG 1.41x, Yokohama FG 1.41x, Fukuoka FG 1.22x, Mebuki 1.20x, versus MUFG ~1.7x; gains of 57-100% in 12 months (Shizuoka +101%, Concordia/Yokohama +86%, Chiba +77%, Fukuoka +70% at Feb-2026); TOPIX Banks +54.7% at 1 year (JPX factsheet Jun-2026)

   → The easy part of the re-rating (from 0.4-0.6x to 1.2-1.5x book) has already passed; the mandate's argument that they are less revalued than megabanks is weak today in leaders and only survives in the small tail, which has the worst deposit quality

   Nota sectorial japonesa jun-2026 (N4, contrastada con Yahoo/JPX N2-N3); Trading Economics feb-2026

8. Consolidation: the free option of the sector inferencia · media N2

   Bloomberg (24-Nov-2025): rate hikes and competition for deposits push mergers among the 73 listed regionals; activist funds (Ariake) seeking targets; Aomori-Michinoku merger with ~80% market share in their prefecture; regulators favorable

   → In small banks with PBR <1 the M&A optionality exists, but buying bad banks expecting a takeover is speculation; only as satellite position and with stable deposits

   Bloomberg 24-nov-2025; The Diplomat feb-2026

9. Credit growing above deposits: structural funding pressure hecho · alta N2

   Bank loans +5.4% year-over-year in Jul-2026 (BOJ) versus deposits +1.8% (Mar-2026); regionals lending +4.3%

   → Good volume for margin today, but the gap forces competing for liabilities tomorrow: reinforces the criterion of buying only deposit franchises dominant in their prefecture

   BOJ via Trading Economics jul-2026; S&P Global Market Intelligence may-2026

Value chain

• 2026-08-14 📊 Published Apr-Jun results of the 73 regionals: aggregate profit 597,900 M JPY, +36.9% sector
  Confirm leverage to rates and set the bar high for the half
• 2026-08-27 🏛️ Himino speech (BOJ Deputy Governor) without braking September hike expectations
  Last official communication before blackout: keeps hike alive
• 2026-09-18 🏛️ BOJ meeting: decision on hike to 1.25% and pace signal in Ueda's press conference sector
  It is THE event of the dossier: with ~80% priced, guidance drives it, not the move
• 2026-09-18 aprox 🏛️ Japan's national CPI for August
  Publishes in same window as meeting; core >2% cements the hike and following ones
• 2026-09-19 a 2026-10-05 🚀 Announcements of ordinary deposit rate hike (0.40% → 0.50%?) by megabanks and large regionals sector
  Will measure deposit beta in real time, the variable that decides who truly gains margin
• 2026-09-29 📌 Semiannual ex-dividend for Fukuoka FG and much of the sector 8354.T
  Technical flows and confirmation of shareholder return policies
• 2026-10-30 🏛️ BOJ meeting with Outlook Report (new projections) sector
  If September was a pause, October is the safety net; market already priced >85% cumulative hike by this date
• 2026-11-06 aprox 📊 Semiannual results (Apr-Sep) of Mebuki, Chiba, Kyoto FG 7167.T, 8331.T, 5844.T
  First complete read of June hike effect and possible guidance improvement and buybacks
• 2026-11-13 aprox 📊 Fukuoka FG semiannual results and close of regional earnings season 8354.T
  Will reveal dispersion in deposits and latent losses: the quality filter in numbers
• 2026-12-18 🏛️ BOJ December meeting sector
  Market prices high probability of being above 1.25% year-end; defines perceived terminal rate

Assets analysed

Risks shared with other dossiers

What would change my mind

• WOULD IMPROVE the thesis: hike on 18-Sep accompanied by explicit signal of faster pace (Ueda validating January-2027) with contained deposit beta <45%; or sectoral correction >12% without deposit deterioration, which would restore margin of safety
• WOULD WORSEN the thesis: pause in September AND October (truncated cycle); deposit beta >60% after next hike; two quarters of deposit decline in chosen names; rise in bankruptcies pushing credit cost >30 bp; sharp yen appreciation <140/USD that freezes the BOJ

1. 2026 agricultural income is NOT high by market, but by subsidies hecho · alta N1

   2026 net agricultural income forecast at 153.4-153.6 billion dollars, -2.6% real vs 2025; direct government payments +45.2% to 44.3 billion; remains 24% below 2022 record (February 2026 forecast)

   → The dossier's starting premise is weak as formulated; income improvement depends on recent grain rally, not published official data

   USDA ERS via DTN y Farm Bureau, 05-02-2026

2. August 2026 grain rally: the new and real detonator hecho · alta N2

   Corn at 511.95 cents/bushel on 31-08-2026, +13.96% in month and +28.79% year-over-year, highest since July 2023; wheat +48% year-over-year; Pro Farmer tour projected corn well below USDA and futures exceeded 5 dollars for first time in 18 months

   → If sustained through harvest, 2026-27 income will exceed February forecast and activate machinery orders toward mid-2027 (the late derivative of thesis)

   Trading Economics / Reuters via Pro Farmer, 21 a 31-08-2026

3. China buys American soybeans again: structural support for rally hecho · media N2

   As of 20-08-2026, pending soybean sales to China for 2026-27 of 6.79 million tonnes, 27% of announced annual target of 25 million tonnes over three years

   → Gives rally room beyond harvest scare; the three-year Chinese commitment, if met, is best catalyst for 2027 agricultural income

   Farm Progress con datos de export sales USDA, 28-08-2026

4. Machinery demand still at cycle bottom hecho · alta N1

   US tractors -10.9% and combines -5.3% year-over-year in July 2026; year-to-date -13.1%; in 2025 tractors fell 9.86% to 195,857 units and combines 35.58% to 3,579

   → Confirms NO recovery yet in retail sales; anyone buying today buys the expectation, not the data

   AEM (asociacion de fabricantes), informe 11-08-2026; farmdoc Illinois 02-2026

5. Quantifiable floor signals: used inventory drained and dealers optimistic hecho · alta N2

   Deere: high-horsepower tractor used inventory model years 2023-24 nearly -40% year-over-year (August 2026); utility tractors 100-174 HP -29% year-over-year in used inventory; dealer sentiment at 3-year high, talking grain price rallies late 2026 and 2027

   → The channel has cleared: when income improves, new production will respond quickly with operating leverage; best evidence in thesis's favor

   Call de resultados Deere 20-08-2026; Tractor Zoom Pro 03-2026; Sandhills/TractorHouse 17-08-2026

6. Deere: beat-and-raise, but from construction and small farm, not large farm agriculture hecho · alta N1

   EPS 5.10 dollars vs 4.69 expected and revenue 12.61 billion; annual guidance raised to 4.75-5.0 billion net income; but still expects large agricultural machinery in US and Canada down 15-20% in fiscal 2026 and management notes fiscal 2026 likely marks cycle bottom (20-08-2026)

   → Excellent and diversified business, but segment embodying thesis (large precision ag) still falling; beat came from elsewhere

   Resultados Deere Q3 fiscal 2026, 20-08-2026

7. Deere trades at extreme multiple vs history: recovery is discounted descontado · alta N3

   P/E ~35x and P/S ~3.7x, both above historical average; historical P/E: average 19.77 and median 19.55; premium to machinery industry (27.6x) and peers (29.8x)

   → Buying DE today pays upfront for 2027-28 bull cycle; error margin is minimal and fiscal 2027 guidance disappointment would hit hard

   Fullratio 10-08-2026; Yahoo/Motley Fool 24-08-2026; SimplyWallSt 11-08-2026

8. AGCO: pure player punished, with North America strong and Europe/Brazil weak hecho · alta N1

   Cut 2026 guidance citing Western Europe and Brazil weakness, but North America up ~20% at constant currency with market share gains in high-horsepower tractors; 2026 adjusted EPS guidance 5.50-5.75; negative free cash flow of 347 million in H1 (30-07-2026)

   → Pure exposure to theme at ~19-20x adjusted guidance after value fell; more leverage to recovery than DE, with more execution and Europe risk

   Resultados AGCO Q2 2026, 30-07-2026 (nota de prensa y transcripcion)

9. CNH: margins at lows and explicit L-shaped recovery expectation hecho · alta N1

   Adjusted EBIT margin in agriculture fell to 5.2% from 8.1% in Q2 2026; management expects L-shaped recovery with 2027 retail demand virtually flat, though sees constructive cycle indicators: dealer inventory normalization, aging fleets and new/used price balance (01-08-2026)

   → The sector itself does NOT discount strong 2027; if grain rally holds, possible positive surprise on low expectations, especially in cheap names

   Resultados CNH Q2 2026 (8-K en SEC), 01-08-2026

10. Tariffs: margin headwind for entire sector hecho · alta N1

    At AGCO, lower volumes, higher input costs and tariff pressure compressed adjusted operating margin 170 basis points to 6.6%; at CNH agriculture margin fall due to unfavorable mix and tariff costs; Deere instead benefited from tariff rebates in quarter

    → Tariffs are common unresolved risk that can eat part of recovery operating leverage

    Resultados Q2 2026 de AGCO y CNH; Q3 fiscal Deere, julio-agosto 2026

Value chain

Assets analysed

Risks shared with other dossiers

What would change my mind

• WOULD IMPROVE thesis: corn yield cut in September WASDE, China exceeding 25 Mt soybean pace, or AEM turning positive before Q1 2027 (would accelerate entire orders curve)
• WOULD WORSEN: record harvest confirmed (Pro Farmer already projects soybeans ABOVE USDA record, 4.572 billion bushels) deflating grain; subsidy cut; or Deere guiding 2027 without growth, dragging entire sector by its reference weight

1. The myth of the captive customer died in 2025 hecho · alta N3

   The Department of Defense declared $4.0B in savings by canceling consulting contracts with Accenture, Booz Allen Hamilton, and Deloitte (April 2025); the GSA instructed agencies to review programs from highest-spending consultants, including Accenture Federal, Booz Allen, Deloitte, Guidehouse, GDIT, IBM, Leidos, and SAIC

   → The explorer's premise (government = captive customer) is false for management consulting; only mission-critical technical integration is hard to cancel

   Nextgov/FCW abril 2025; FedSavvy 2025

2. Leidos accelerates and raises guidance with the wind of the defense budget hecho · alta N1

   T2 2026 revenue of $4.6B (+7%), adjusted EBITDA $631M with 13.8% margin, non-GAAP free cash flow of $761M (August 4, 2026); the Defense segment achieved 2.2x book-to-bill in the quarter and 1.9x over twelve months; backlog of $48.7B with funded backlog +44%

   → It is the asset where committed spending (firm backlog) grows faster than stock price

   8-K de Leidos en SEC, 4 de agosto de 2026

3. Booz Allen: cheap but shrinking hecho · alta N1

   Quarter ended June 30, 2026 revenue of $2.8B, -4.2% year-over-year; net income $198M, -26.9%; adjusted EPS $1.81, +22.3%; 1.5x book-to-bill; $39.0B backlog (published July 24, 2026); FY2027 guidance of $11.2B-$11.7B, 0%-4% growth

   → The low multiple reflects declining revenue; margins rise from headcount cuts, not demand

   8-K de Booz Allen en SEC, 24 de julio de 2026

4. Accenture: the problem is not the government, it is AI hecho · alta N1

   T3 fiscal 2026 (ended May 31, 2026): revenue $18.7B (+6%), bookings $19.3B (vs $19.7B a year prior), diluted EPS +9% to $3.80; cut annual guidance: 3%-4% growth in constant currency, or 4%-5% excluding an estimated 1% impact from US federal business; on June 18, 2026 the stock fell nearly 20%, its worst session on record, after cutting forecasts and reporting -2% bookings

   → The federal drag is quantifiable (~1% of growth); what the market fears is commoditization by AI of the hours-billed model, which is hypothesis, not fact

   8-K de Accenture en SEC y prensa, 18 de junio de 2026

5. Accenture buying its pivot to OT security hecho · alta N2

   Announced ~$4.18B for a majority stake in cybersecurity firm Dragos and acquisitions of runZero and NetRise, its largest bet in operational technology (power grids, factories) (June 18, 2026)

   → Reinforces purity in security, but paying $4.18B while organic growth cools raises execution risk

   Anuncio de Accenture, 18 de junio de 2026

6. CACI, the best operator in the group in 2026 hecho · alta N2

   FY2026 results (August 6, 2026): revenue $9.57B, +10.9% with 7.2% organic; free cash flow per share +68% to $735M; stock rose 10.2% on results; funded backlog of $5.4B, +28.6% year-over-year

   → Confirms that demand exists for those selling mission technology, not PowerPoints; already partially rewarded by the market

   Nota de resultados de CACI e Investing.com, 6 de agosto de 2026

7. FY2027 budget favors defense-tech and punishes civil hecho · media N3

   Pentagon RDT&E request for FY2027 stands at $210.4B vs $144.9B in FY2026; requested federal civil IT spending rises to $75.7B (+$7.8B), but the proposal cuts ~10% of civil agency discretionary spending and CISA faces a $707M cut in FY27

   → It is a REQUEST, not approved law: the mix favors LDOS/CACI (defense and intelligence) and penalizes civil portfolios (BAH civil, Leidos health)

   GovCIO/Winvale/DLT sobre la petición presupuestaria de la Casa Blanca, abril-junio 2026

8. The valuation discount is already extreme at BAH descontado · alta N3

   BAH traded at $76.36 on August 20, 2026, within a 52-week range of $59.50-$112.22; all-time closing high of $177.35 on October 28, 2024; Goldman Sachs maintains sell rating (July 27, 2026)

   → Market already discounts prolonged stagnation; the debate is whether it is a floor or a trap

   GuruFocus/Macrotrends/TipRanks, agosto 2026

Value chain

Assets analysed

Risks shared with other dossiers

What would change my mind

• Would improve the thesis: FY2027 budget approved without long continuing resolution; BAH stringing together two quarters of positive revenue growth (moves from WATCH to BUY_ON_CORRECTIONS); stabilization of ACN bookings in September
• Would worsen the thesis: shutdown in October 2026; Pentagon insourcing plan executing with concrete figures; Leidos loses the MDE contract; evidence of price deflation per hour attributable to AI in FY2026 10-Ks

1. Transformer lead times measured in years, not months hecho · alta N3

   <cite index="8-1,8-2">Average lead time of 128 weeks for standard power transformers in the US in 2026 and about 144 weeks for generator step-up units</cite>; <cite index="8-5">a GSU order placed mid-2026 receives a delivery date in 2029</cite>

   → The manufacturer has structural pricing power: whoever controls the production slot controls the entire project schedule

   Wood Mackenzie via prensa sectorial, mayo-junio 2026

2. Structural demand, not cyclical peak hecho · alta N3

   <cite index="1-11">Since 2019, generator step-up demand has grown 274% and substation transformer demand 116%</cite>; <cite index="1-1">Wood Mackenzie models for 2025 a 30% deficit in power transformers and 10% in distribution units</cite>; <cite index="1-7">more than half of the US distribution transformers, about 40 million units, already exceed their expected life</cite>

   → There is replacement demand in addition to growth demand: the cycle has two engines, reducing the risk that an AI slowdown kills it outright

   Wood Mackenzie / POWER Magazine, enero 2026

3. Crane case: grid connection is worth more than the reactor hecho · alta N1

   <cite index="28-4,28-10">FERC approved on June 1, 2026 the waiver allowing Constellation to transfer 760 MW of Capacity Interconnection Rights from Eddystone to Crane</cite>; <cite index="20-2">without the transfer, Crane's full deliverability was at risk until at least December 31, 2030, three years after the 2027 startup target</cite>; <cite index="25-5,25-6">PJM had identified multiple prior transmission projects, including hundreds of miles of new 765 kV and 500 kV lines</cite>

   → Confirms central thesis: the scarce asset is the connection, not generation; existing interconnection rights have become monetizable assets and transmission lines into multi-year orders for PWR and equipment for GEV/ETN/HUBB

   FERC orden 195 FERC 61,162 del 1 de junio de 2026; Utility Dive 3 de junio de 2026

4. GEV: gas backlog of 100 to 116 GW in one quarter hecho · alta N1

   <cite index="10-1">Orders of $24.2 billion (+88% organic), sequential backlog +$13 billion, gas equipment and slot reserves backlog of 100-116 GW with target of at least 125 GW by end-2026, free cash flow of $5.1 billion in the quarter and cash of $13.1 billion</cite> (2Q26 results published July 22, 2026); <cite index="14-2">already taking orders for 2031 deliveries</cite>

   → Earnings visibility through 2031 with <cite index="18-15">equipment orders in H1 2026 priced 20% above Q4 2025</cite>: real pricing power, not narrative

   8-K de GE Vernova ante la SEC, 22 de julio de 2026

5. PJM: third consecutive record capacity price hecho · alta N2

   <cite index="65-13,65-14">PJM capacity auction announced December 17, 2025 hit the cap of $333.44/MW-day, record third consecutive auction, and came up 6,625 MW short of the 20% reserve margin target</cite>; <cite index="64-6">more than 46,000 MW already approved remain unbuilt due to permitting, supply chain and project economics</cite>

   → The pricing signal to build generation and grid is maximal; the regulator itself acknowledges supply chain is the brake, extending the order cycle for all four stocks

   PJM y Utility Dive, 17-18 de diciembre de 2025

6. The bottleneck within the bottleneck: GOES steel and labor hecho · media N4

   <cite index="5-9,5-10">Cores require grain-oriented electrical steel and Cleveland-Cliffs is the only domestic producer</cite>; <cite index="8-11,8-12">its $195 million investment in Butler Works, with up to $75 million DOE subsidy, won't be operational until July 2028</cite>; <cite index="8-4">the binding constraint is a US workforce of about 15,000 people hand-winding copper</cite>

   → The shortage is not resolved before 2028 even if capital pours in: incumbents' high margins have at least two more years

   Manufacturing Mag / IndustrialSage, mayo-junio 2026

7. ETN and HUBB confirm cycle breadth beyond gas hecho · alta N1

   ETN 2Q26 (July 31, 2026): <cite index="35-1">sales +21% with 14% organic growth, 12-month average orders +41% in Electrical Americas and Electrical sector backlog +43% year-over-year</cite>; HUBB 2Q26 (July 28, 2026): <cite index="46-2,46-4">Utility Solutions sales +10% to $1.026 billion, with Grid Infrastructure +12%</cite> and <cite index="51-1">1.2x book-to-bill in Utility Solutions</cite>

   → The cycle is not just turbines: grid components, switchgear and substations grow double-digit with record margins; the thesis is full-chain

   8-K de Eaton y de Hubbell ante la SEC, julio de 2026

8. Already discounted: multiples in demanding zone descontado · media N4

   <cite index="54-1">GEV trades at 51.5x forward P/E on August 26, 2026</cite> (other sources place it at 43x); <cite index="76-14,76-18,76-19">PWR trades at 75x trailing P/E versus a 10-year average of 34, a 119% premium</cite>; <cite index="85-2">ETN at about 31x forward P/E</cite>; <cite index="82-2">HUBB at 30.9x trailing and 25.4x forward P/E</cite>

   → The market already pays for several years of growth: asymmetry at these prices is poor and demands entry discipline; HUBB is the only one with defensible multiple

   GuruFocus, stockanalysis.com, fullratio, mayo-agosto 2026

Value chain

Assets analysed

Risks shared with other dossiers

What would change my mind

• Would improve the thesis: 15-20% market correction in GEV/PWR without order deterioration (invalidation criteria intact would turn WATCH into BUY and BUY_ON_CORRECTIONS into CORE); new PJM capacity auction with still-short procured volume
• Would worsen the thesis: two consecutive quarters of book-to-bill below 1.2x in GEV Power; hyperscaler capex cuts; evidence that new transformer capacity from 2028 arrives earlier and at lower prices; extension of state data center moratoriums to Virginia or Texas
• Would change the vehicle, not the thesis: if US multiples keep expanding without earnings improvement, rotate exposure to Siemens Energy/Hitachi with same demand driver

1. Japanese curve at 30-year highs hecho · alta N2

   JGB 10y ~2.94% on 31-08-2026 (30-year high); 30y touched 3.89% record on 20-01-2026 and traded 3.9-4% in June 2026; 40y record 4.23% in January 2026

   → Reinvestment of portfolios at 4% versus old guarantees of 1-2% generates structural positive spread for Japanese life insurers

   Trading Economics / Reuters / Nikkei, agosto 2026

2. BOJ about to raise again hecho · alta N2

   Market discounts ~80-87% probability of 25bp rise to 1.25% in September 2026 (was 23% before July meeting); former adviser Adachi sees another rise in January 2027

   → The rate tailwind is still alive, but at 80%+ already discounted a rise is NOT a catalyst; a pause would be a hit

   Trading Economics citando precios de mercado, 25/27-08-2026

3. Dai-ichi: quarterly profit +113% hecho · alta N1

   Group adjusted profit ¥158,100M in April-June 2026 (+113% year-on-year, 28% of annual target); ordinary profit +195.6% to ¥251,000M; FY2025 closed at ¥552,000M (+25%), third record year, adjusted ROE 12.7% meeting FY2026 target ahead of schedule

   → The positive domestic spread already translates to real cash, not promise; note: they didn't raise annual guidance despite the beat

   Daiichi Life Group, resultados publicados agosto 2026 y mayo 2026

4. The market HAS ALREADY paid much of it hecho · alta N3

   Dai-ichi +67.2% in 12 months to August 2026, trading ~1,846 yen versus 52-week range of 1,052-1,903; L&G +24% after August 2026 results; NN and Aegon near 52-week highs

   → Invalidates the explorer's premise that 'silent winners off the radar': discovery already happened; what's buyable is persistence, not surprise

   Investing.com / ABC Money, agosto 2026

5. Record unrealized bond losses hecho · media N3

   The 4 major Japanese mutuals (Nippon, Dai-ichi Life insurer, Sumitomo, Meiji Yasuda) accumulate ~¥15.13 trillion (~96,000M USD) of unrealized losses on domestic bonds as of 30-06-2026, +7% in the quarter; FSA accelerated balance sheet review; stress threshold cited in market: 30y JGB at 4.5%

   → The same engine that creates spread destroys book value of old stock; if rates rise TOO FAST, the gift becomes a capital and surrender problem

   Prensa financiera citando informes de las companias, agosto 2026; Insurance Business/Nikkei junio 2026

6. The insurers themselves buy the thesis hecho · alta N2

   Meiji Yasuda doubled its JGB purchase plan for FY2026 to over ¥2 trillion seeing the 30y at 4% as a buy opportunity (30-06-2026); half of the major life insurers will expand domestic bonds this fiscal year (Nikkei, 28-04-2026)

   → Confirmation from inside the sector that current rate levels are economically attractive for the life business

   Nikkei Asia, abril y junio 2026

7. New Japanese capital regime without accidents hecho · alta N3

   J-ICS (solvency at economic value, equivalent to Solvency II) entered into force for the fiscal year ended 31-03-2026; AM Best maintains stable outlook and highlights improvement in reinvestment yields

   → The binary regulatory risk has passed: transition to the new standard has not forced capital raises in quoted names

   AM Best via Insurance Business, junio 2026

8. Europe: solvency rising and reform that frees capital hecho · alta N3

   Aggregate SCR ratio of 20 major European groups rose to 217.7% in 2025 from 208.4%, supported by steepening of the risk-free curve; Fitch estimates that Solvency II reform (transposition by 30-01-2027) provides 5-7% average capital benefit, more in life

   → Already high buffers + extra capital in 2027 = ammunition for dividends, buybacks and pension transfer growth

   Mapfre Economics / Fitch via Insurance Business, julio 2026; Directiva UE 2025/2

9. European downside nuance: margin doesn't expand as much inferencia · alta N3

   S&P (December 2025) does not foresee material margin expansion in European life despite high rates, due to higher weight of unit-linked and profit-sharing rules with the insured

   → In Europe the thesis is more about capital return and solvency than pure rate leverage; Japan has more beta to spread

   S&P Global via Insurance Journal, 17-12-2025

10. Solid H1 2026 European results with rising capital hecho · alta N1

    NN Group: OCG €1,073M (+5%), net income €1,066M (+172%), SII 224% from 213%, dividend on account +12% (06-08-2026). L&G: core operating profit £918M (+7%), EPS +11%, SII 201% (209% pro forma), £1,200M buyback in progress (August 2026). Aegon: OCG +27% to €416M, SII 169%, 2026 buybacks raised to €577M, dividend +11% (20-08-2026)

    → The sector returns cash in rising and verifiable fashion; the question is not whether the business is doing well but how much is left to pay in price

    Comunicados de resultados de NN, L&G y Aegon, agosto 2026

11. Credit risk is not zero hecho · media N2

    L&G declares over 99% of its bond portfolio in investment grade but with private credit skewed to UK (August 2026); Morgan Stanley published estimates of Japanese insurers' private credit exposure (August 2026); S&P notes that European private credit exposure is much lower than US peers

    → The premise 'no banking credit risk' is only partially true: annuity businesses (L&G, Protective of Dai-ichi, Aegon US) carry corporate and private credit on balance

    Transcripcion resultados L&G agosto 2026; Investing.com; S&P diciembre 2025

Value chain

Assets analysed

Risks shared with other dossiers

What would change my mind

• WOULD IMPROVE the thesis: 15-20% correction in Japanese names without fundamental change (would spike conviction to 8-9 and decision to CORE), or evidence in November 2026 that surrenders do NOT rise despite high rates
• WOULD WORSEN the thesis: BOJ pause + 10y JGB fall below 2.3%, signals of mass surrenders, or a US private credit event that spreads to Protective/Transamerica; any of the three would lead me to cut to MONITOR
• STRUCTURAL CHANGE: if the January 2027 Solvency II reform arrives diluted or the Japanese FSA toughens J-ICS after its review, the freed capital argument disappears

1. Truce with expiration date: November 10, 2026. hecho · alta N3

   November 10, 2026 marks expiration of the 12-month suspension of broad Chinese rare earth export controls; what happens will determine if the world economy's most concentrated dependency becomes active crisis or remains managed risk.

   → Binary catalyst with absolute date affecting the entire complex; the market knows it but unclear which scenario is priced.

   EBC/acuerdo Trump-Xi, abril 2026

2. Hard April 2025 controls still in force despite truce. hecho · alta N2

   April 2025 controls on seven rare earths and permanent magnets remain in effect; China requires case-by-case license for dysprosium, terbium, and samarium, and US companies report continued access difficulty.

   → The real bottleneck is heavy elements (Dy/Tb); the truce is cosmetic for defense.

   EBC/MOFCOM, abril 2026; confirmado por Xinhua 20-05-2026

3. Projected structural NdPr deficit. inferencia · media N3

   Bloomberg Intelligence projects 36% global NdPr deficit by 2030 even with $10 billion in public financing expected in 2026; new mines take 10–15 years.

   → Medium-term price support; but it is projection, not fact.

   Bloomberg Intelligence, 2026

4. NdPr prices: violent, volatile rally in 2026. hecho · media N3

   NdPr alloy reached $126.16/kg in China on April 1, 2026, +138% from ~$53/kg at year-start; then corrected and by June 2026 NdPr benchmark recovered 10% to $109.55/kg.

   → Price hovering around the $110/kg floor agreed with MP and Lynas: government floors are ceasing to be subsidy and becoming market reference.

   SMM vía rare-earth-mining.com, 01-04 y 01-06-2026

5. MP Materials: strong growth but still without net profit. hecho · alta N1

   2Q 2026 revenue of $108.5 million, +89% year-over-year, with NdPr production of 840 tonnes (+41%), sales of 1,006 tonnes (+127%), and $17.6 million income from price protection agreement; however adjusted EPS was –$0.01, below expected $0.01 profit.

   → The thesis works in volume and EBITDA but earnings per share remain at zero: what is being paid is the future magnet business.

   Resultados MP Materials, 06-08-2026

6. MP's state floor: firm contract, not promise. hecho · alta N1

   The DoD signed a 10-year agreement with $110/kg floor for MP's NdPr; includes $400 million equity and $150 million loan, and PPA income contributed $42.3 million in 1Q 2026, 32% of combined revenue; as condition, MP restricted from selling to non-US customers.

   → Downside cash flow assured, but one-third of current revenue is public transfer: political dependence, not market.

   MP Materials/DoD, julio 2025; 10-K/10-Q 2026

7. Lynas: record year and first producer of heavy elements outside China. hecho · alta N1

   FY26 record revenue of A$977.9 million and net profit of A$222.4 million versus A$8.0 million prior year; became first supplier of dysprosium and terbium outside China; firm 12-year agreement with JARE for 5,000 tonnes/year NdPr with $110/kg floor.

   → Today the only profitable asset with cash in the theme; but stock fell post-results due to operational issues, signaling demanding expectations.

   Resultados FY26 Lynas, 26-08-2026 (ASX)

8. Heavy elements still symbolic in volume outside China. hecho · alta N2

   Global dysprosium and terbium demand measured in thousands of tonnes annually; in 1Q 2026 Lynas produced 8 combined tonnes.

   → The Western independence narrative in heavy elements is currently a drop in the ocean: years of Chinese advantage.

   Bloomberg Graphics, mayo 2026

9. Energy Fuels crosses heavy elements threshold in the US. hecho · alta N1

   First kilogram of 99.9% terbium oxide in Utah from US ore, after nearly 30 kg of dysprosium oxide; commercial expansion under construction with ~$104 million capex largely supported by public grants and loans; its terbium qualified on August 19, 2026 by one of Japan's largest magnet makers without additional validation.

   → Real upside optionality in heavy elements with solid balance; but from kilos to tonnes is years away.

   Notas de prensa Energy Fuels, 25-03, 29-07 y 19-08-2026

10. Project Vault: US strategic mineral reserve. hecho · media N2

    Trump plans a strategic critical minerals reserve of at least $12 billion, combining $1.67 billion private capital with a $10 billion Ex-Im Bank loan (announced February 2026).

    → Incremental public demand and political price floor; the loan still required bank approval: monitor execution.

    Bloomberg vía El Correo Gallego, 02-02-2026

11. Gallium and germanium: two markets split by embargo. hecho · media N3

    Since December 2024 China bans gallium, germanium, and antimony exports to the US; gallium FOB export price ~$400/kg versus domestic China price, confirming export license is the binding constraint; germanium in Western storage trades at $6,350/kg, a 104.6% premium to China.

    → West-China price bifurcation is structural; benefits Western recyclers of byproducts, though no pure-play listed vehicle.

    SMM/USGS vía rare-earth-mining.com, 03-08-2026

Value chain

Assets analysed

Risks shared with other dossiers

What would change my mind

• WOULD IMPROVE the thesis: effective reimposition of Chinese controls on November 10–11, 2026, firm contracts (not letters of intent) from Project Vault buying from domestic producers, or Lynas Dy/Tb production exceeding 50 tonnes annually.
• WOULD WORSEN the thesis: Chinese general licenses normalizing trade in verifiable form, a broad US-China trade deal with specific minerals chapter, DoD PPA to MP cut or renegotiated, or breakthroughs in low-heavy-element magnets reducing Dy/Tb demand.

1. Near-total closure of Black Sea grain corridor. hecho · alta N2

   More than 97% of Russia and Ukraine grain export capacity in the Azov-Black Sea basin is halted; the region exported 7.2 Mt/month average in prior season (20-Aug-2026).

   → Real supply shock, not noise: the world's largest wheat exporter (Russia) cannot ship in peak season Aug-Dec; structural price support while it lasts.

   Reuters via Business Recorder, 20-ago-2026

2. Wheat at 3-year highs with war premium. hecho · alta N2

   Chicago Dec-2026 closed at 7.61 USD/bu (+61 cents in week to 28-Aug-2026); December position +12% weekly to ~279 USD/t, highest since February 2023; Dec corn at 5.34 USD/bu.

   → The Black Sea halt is substantially discounted in wheat futures; buying wheat here is betting on further escalation, with binary downside risk if there is a truce.

   Grain Journal y Barchart, 28-ago-2026

3. Nitrogen has already corrected: urea falling from 700 to 420 USD/t. hecho · media N3

   Urea at 420 USD/t on 27-Aug-2026 (-3.45% in month, +9% YTD) after exceeding 700 USD/t on 3-Aug; reason: negotiating progress on Hormuz, China easing exports and end of planting season.

   → The nitrogen profit peak may be behind us; Q3 results from CF and NTR will capture high realized prices with lag, but market looks ahead.

   Trading Economics, 27-ago-2026

4. The persistent bottleneck is sulfur and phosphate, not nitrogen. hecho · media N3

   Sulfur +261% year-on-year with Russian export veto through end 2026; China restricts phosphate exports (DAP/MAP/NPK) covering 50-80% of volumes (week of 17-Aug-2026); U.S. retail DAP at 917 USD/t.

   → Divides the sector in two: those buying sulfur/ammonia (Mosaic) suffer negative margins; those selling Canadian potash or nitrogen with cheap American gas (NTR, CF) win.

   iGrow News con datos Trading Economics, 17-ago-2026; DTN, 19-ago-2026

5. Nutrien: strong potash, marginal phosphate, valuation lagging. hecho · alta N1

   Q2-2026 (released 5-Aug-2026): net income 1,220 M USD, adjusted EBITDA 2,430 M, sales 10,812 M; potash EBITDA +15% to 1,240 M with net prices +13%; adjusted EPS missed (2.61 vs 2.89 expected) due to nitrogen; manufactured phosphate gross margin negative (-31 USD/t); raised potash volume guidance and cut capex.

   → The potash leg (Canada, no Hormuz exposure) and largest agricultural retail in North America give exposure to the theme with less binary risk than pure nitrogen.

   Nutrien press release y earnings call, 5/6-ago-2026

6. CF Industries: the winner already rewarded by market. hecho · alta N1

   Q2-2026 sales of 2,222 M USD vs 1,890 M one year prior (8-K, 5-Aug-2026); +49.2% YTD to 10-Aug but at 7.6x forward earnings with 11% FCF yield; buybacks reduced shares ~10% in one year; consensus neutral with mean price target 126 USD.

   → Low multiple but on cyclically inflated earnings: market already pays much of war premium; 28-Jul truce pause cost it -3.5% in one day, sign of how sensitive it is.

   SEC 8-K 5-ago-2026; Investing.com 10-ago-2026; stockanalysis.com jul-2026

7. Mosaic: the sulfur victim, with contrarian optionality. hecho · alta N1

   Q2-2026: net loss of 272.8 M USD (vs +410.7 M in Q2-2025), adjusted EBITDA 407 M; sulfur guided to 700-710 USD/t in Q3; Faustina halted, Bartow at 40%; potash holds (EBITDA 278 M, cash cost 84 USD/t vs price 275); long-term debt rises to 4,830 M and cash hole of ~500 M in year.

   → It is the inverse bet: if sulfur normalizes (end of Russian veto or peace), Mosaic has largest leverage to recovery; today it is bearing CAPEX and not capturing margin.

   10-Q y earnings call Mosaic, 4/11-ago-2026

8. Grain traders capture volatility: ADM raises guidance for second time. hecho · alta N2

   ADM raised 2026 guidance to 5.15-5.60 USD/share from 4.15-4.70 (4-Aug-2026); ag services and oilseeds operating profit +129% year-on-year; Bunge also raised guidance prior week.

   → Intermediaries win with volatility and flow redirection (Middle East/Africa importers seeking origin in Americas and Australia), with less directional risk than fertilizer.

   Reuters, 4-ago-2026

9. FAO food index at 3-year high but far from 2022. hecho · alta N1

   131.1 points in July 2026 (released 7-Aug-2026), maximum since January 2023; cereals +3.4% monthly, wheat +5.8%; index still ~20% below March 2022 peak.

   → There is tension but not food crisis like 2022 yet; upside exists if Black Sea halt extends to winter when Azov ice prevents recovering shipments.

   FAO, 7-ago-2026

10. Ukrainian 2027 planting threatened and storage damaged. hipotesis · baja N4

    Projection of Ukrainian cultivated area falling from 20 M ha to 13-15 M ha; Ukraine claims 90% destruction of storage capacity (late August 2026).

    → If confirmed, shock shifts from logistic (reversible with truce) to production (lasts 1-2 campaigns); is the variable converting the theme to structural. Treat as unverified.

    Blasina y Asociados citando informes ucranianos, 29-ago-2026; NO verificado en fuente primaria independiente

11. El Niño boosts fertilizer demand. inferencia · media N3

    NOAA assigns 81% probability to very strong El Niño through end 2026, threatening yields in Asia and South America.

    → Weak harvests typically raise fertilizer application to protect yield: second tailwind for NTR/CF beyond supply shock.

    Investing.com citando NOAA, 28-jul-2026; conviene validar en fuente NOAA directa

Value chain

Assets analysed

Risks shared with other dossiers

What would change my mind

• WOULD IMPROVE thesis: primary confirmation (USDA/FAO) of structural damage to Ukrainian productive capacity; extension of Russian sulfur veto to 2027; El Niño materializing in crop losses in Asia/South America; winter with Azov frozen and unrecoverable Russian export backlog.
• WOULD WORSEN thesis: verifiable naval truce in Black Sea (would drop WHEATCFD to AVOID and NTR/CF one notch); China reopening phosphate and urea simultaneously; urea below 350 USD/t; signs of fertilizer demand destruction by farmer insolvency in 2027 campaign.
• ASSET SWITCH: if sulfur normalizes before Q1 2027, MOS moves from MONITOR to COMPLEMENTARY for its leverage to phosphate recovery.

1. Global security spending revised upward twice in one quarter hecho · alta N3

   Gartner Q2'26 forecast (Jun-25-2026): $248.9 billion in 2026 (+12.7% constant currency) and $372.6 billion in 2030; the 2026 figure moved from 244.2 to 246.2 then to 248.9 billion, two upward revisions in one quarter

   → The spending tailwind is real and accelerating; validates the explorer's framework that cybersecurity is last to be cut

   Gartner, Forecast Information Security Worldwide 2Q26, 25-jun-2026

2. AI and cloud security, the fastest-growing segments hecho · media N3

   Gartner for first time counts spending on securing AI itself: it is the only growth curve accelerating, captures more new dollars than any other category and will be the largest security line item by 2029; cloud security posture management grows 27.6% annually from $4.7 to $16.1 billion

   → Whoever has security product for AI agents and cloud captures marginal growth; favors CRWD, ZS, PANW and penalizes pure hardware

   Gartner 2Q26 via Software Strategies Blog, 6-jul-2026

3. State attacks up but +7.5%, not exponential hecho · media N3

   158 APT incidents from North Korea, Russia and China in H1'26, +7.5% vs 147 in prior half (S2W TALON, Aug-12-2026); Russia +30% (26 incidents); China -17.5% but pivoting to stealthy long-term espionage; North Korea stole $2.02 billion in crypto in 2025, +51%, annual record (Chainalysis)

   → The narrative that 215 geopolitical events are driving cyberattacks is directionally true but hard data (+7.5%) is more modest than the story; the real driver of buying is budget and regulation, not weekly headlines

   S2W TALON H1 2026 APT report, 12-ago-2026; Chainalysis 2025

4. Iran 2026 conflict: cyber front exists but is opportunistic hecho · media N3

   Israeli National Cyber Director confirmed on Mar-6-2026 a wave of Iranian wipers that deleted data from over 50 small Israeli companies, exploiting pre-existing weaknesses; well-protected companies were unaffected

   → Direct commercial argument for vendors: having modern protection made the difference between being victim or not; reinforces platform demand in SMB and mid-market

   Israel National Cyber Directorate via EclecticIQ, mar-2026

5. CrowdStrike: record quarter and reacceleration confirmed hecho · alta N1

   Q2 fiscal 2027 (closed Jul-31-2026, reported Aug-26-2026): record net new ARR of $333 million (+51%), annual ARR net new guidance raised 630 bps to +34%, record operating cash of $530 million and FCF of $377 million; revenue +26% to $1.47 billion and total ARR $5.84 billion (+25%); stock up over 20% in one session after results

   → Best-in-sector business reaccelerates; but result drove +20% in one day: most good news already priced

   8-K de CrowdStrike en SEC, 26-ago-2026

6. CRWD valuation in extreme zone hecho · media N4

   At Aug-19-2026 CRWD traded at around 175x forward non-GAAP earnings and over 38x forward sales, among most-expensive large tech in market; after post-results rally at roughly 30x estimated sales for fiscal year ending January 2028, enormous multiple for 25% growth

   → Good technology is not good investment: at this price margin for error is nearly nil; one merely-good quarter can cost -20/-30%

   TechTimes 19-ago-2026; Motley Fool 29/30-ago-2026

7. Palo Alto closes CyberArk and publishes full fiscal 2026 year Sep-1-2026 hecho · alta N1

   CyberArk acquisition closed for ~$25 billion (Feb-11-2026) with mixed payment of $45 cash plus 2.2005 PANW shares per share (material dilution); for Q4 fiscal 26 (publishes Sep-1-2026) consensus and guidance point to +32% revenue interannual with next-generation security ARR +59-60%, first full quarter with CyberArk and Chronosphere

   → Immediate, verifiable catalyst tomorrow Sep-1-2026; note: part of +32% is inorganic, demand organic breakdown

   8-K PANW en SEC 11-feb-2026; Hudson Labs ago-2026

8. Fortinet: sector's cheap value accelerates hecho · alta N1

   Q2'26 (Jul-29-2026): revenue +26% to $2.05 billion, product +52%, billings +33%, non-GAAP operating margin 38%, FCF $965.6 million (47.2% margin); annual guidance raised to $8.02-8.18 billion, +19%; Moody's upgraded senior debt to A3, best rating among pure-play cybersecurity listed peers

   → Growth comparable to CRWD with impeccable balance sheet and much lower multiple; risk: part of push is firewall refresh cycle (hardware), which exhausts

   Nota de resultados Fortinet, 29-jul-2026

9. Zscaler: warning of what happens when expectation fails hecho · alta N1

   Stock fell over 30% on May-27-2026, worst day on record, after guiding only 16-17% ARR growth for fiscal 2027, below consensus, losing two top sales leaders despite Q3'26 showing +25% revenue to $850.5 million and ARR of $3.525 billion (+25%)

   → Empirical proof of common sector risk: high multiple + soft guidance = -31% in one day; simultaneously, already-suffered derating makes ZS a buy candidate if commercial execution stabilizes

   CNBC 27-may-2026; 8-K Zscaler en SEC 26-may-2026

10. Sector consolidation underway hecho · alta N1

    CyberArk buy is second-largest acquisition in cybersecurity history; Zscaler bought Red Canary ($651.4 million), SPLXAI ($40.6) and SquareX ($113) to expand AI and Zero Trust capabilities; CrowdStrike announced XM Cyber buy in 2026

    → Large platforms buy growth: be careful reading reported ARR without adjusting for inorganic (ZS grows 21% organic vs 25% reported)

    Filings y notas de prensa de las compañias, feb-ago 2026

Value chain

Assets analysed

Risks shared with other dossiers

What would change my mind

• WOULD IMPROVE thesis: sector correction of 20-25% with no deterioration of fundamentals (businesses grow 20-25% with rising cash; only multiple excess); or evidence that AI agent security generates audited material ARR (>$500 million at a vendor) not just announcements
• WOULD HURT thesis: two or more sector leaders guiding below 15% growth for 2027 (would signal Gartner's 12.7% is ceiling, not floor); global IT spend cut from recession proving cybersecurity does get cut; or major breach from security product breaking category trust

1. Crane (ex Three Mile Island 1) clears its final two regulatory hurdles hecho · alta N1

   FERC granted exemption to transfer interconnection rights to Crane and NRC approved fuel license amendment, major milestone toward 2027 restart; NRC approval is Amendment 307, dated July 1, 2026

   → The catalyst we were watching (license by May 2027) has been brought forward in the key regulatory part; the plant is positioned to return to service in second half of 2027, with full 835 MW for Microsoft on 20-year term

   NRC y nota de resultados 2T26 de Constellation, agosto de 2026

2. Westinghouse formally launches its IPO process hecho · alta N1

   Westinghouse confidentially filed a draft S-1 registration with the SEC for an IPO, communicated by Cameco on July 31, 2026 alongside its second quarter results; Cameco documents that 57% of the world's 417-reactor fleet uses Westinghouse technology and a pipeline of up to 91 potential AP1000s, roughly 105 GWe

   → The second catalyst from the previous dossier has materialized in filing; share count and price are not fixed, so valuation (speculation ranges well above the $8 billion enterprise value paid in 2023) is still hypothesis, not fact. Crystallizes value for CCJ (49%)

   Cameco resultados 2T26 / World Nuclear News, 31 de julio - agosto de 2026

3. Forward uranium marks 18-year highs; spot converges hecho · media N3

   Long-term price reached $94 per pound at end of June 2026, highest since 2008, while spot consolidated at 85-87 after exceeding $100 in early year; as of August 28, 2026 uranium futures trade around $90.4 per pound

   → The utilities' contracting cycle keeps pushing the price at which they commit future pounds; benefits producers with long contracts (CCJ). Note: Bank of America's $135 target for 2026 is bank estimate, not fact

   Crux Investor / TradeTech / NYMEX vía agregadores, junio-agosto de 2026

4. Oklo: real technical execution, but massive dilution and higher cash burn hecho · alta N1

   First criticality of Groves isotope reactor in roughly 11 months from start of construction; closed quarter with $3 billion in cash and securities, including $1.9 billion raised via ATM programs during 2026; raised operational cash use in 2026 to $120-150 million and capex to $400-500 million; 2Q26 revenue of only $1.2 million, from recent acquisitions

   → Execution improves (criticality, DOE preliminary safety analysis approval for Aurora-INL with target start in 2028), but shareholders pay the bill: the $1.9 billion ATM capital raise confirms our warning that future revenue does not equal shareholder value today

   Resultados 2T26 de Oklo, 6-7 de agosto de 2026

5. Vistra: Meta PPA not yet in guidance and ERCOT cools hecho · alta N1

   Adjusted 2Q26 EBITDA of $1.767 billion, +30% year-over-year, and launch of Helix Digital Infrastructure with KKR, NVIDIA and Kuwait Investment Authority, with up to $1 billion committed by Vistra; ERCOT weakness drives it toward the low end of 2027 range, though Cogentrix and Meta PPAs could add roughly $700 million not included

   → The contracted nuclear business is fine, but the merchant business in Texas reminds that these companies are not just AI theme: ERCOT suffers low prices from early batteries and data center loading delays, an early signal that data demand may arrive later than promised

   Resultados 2T26 de Vistra y transcripción de la llamada, 7-14 de agosto de 2026

6. Skeptics quantify the gap between announcements and real capex interpretacion · media N3

   In their February 2026 earnings calls, Amazon and Google said they expect to spend up to $200 billion and $185 billion respectively on data centers this year; their nuclear commitments are a fraction of one percent of that spend

   → Healthy counterweight to narrative: tech companies' nuclear money is still small versus total capex; the thesis holds on firm PPAs with existing operators, not on nuclear feeding most AI this decade

   Bulletin of the Atomic Scientists, 20 de julio de 2026

7. Deals keep arriving, but you must separate binding from non-binding hecho · alta N2

   Meta announced in January 2026 agreements with TerraPower, Oklo and Vistra supporting up to 6.6 GW of new and existing clean energy through 2035; in contrast, NANO Nuclear signed in August 2026 a non-binding commercial framework with Tillman for up to 6 GW in 2040, subject to definitive agreements, customers, financing and permits

   → The hard rule of the dossier holds: signed PPA with operating plant (Vistra-Meta, Microsoft-Crane, Amazon-Talen) is contracted revenue; gigawatt frameworks to 2040 from unlicensed startups are narrative that should not be paid for

   AP/CBS enero de 2026; Interesting Engineering agosto de 2026

8. The American state finances the nuclear chain hecho · alta N1

   In November 2025 the DOE loaned $1 billion to Constellation and in June 2026 announced $17.5 billion in loans for the American nuclear supply chain to accelerate deployment of 10 large commercial reactors

   → Reduces the sector's cost of capital and supports Westinghouse's AP1000 pipeline; the symmetric risk is dependence on federal policy

   DOE vía Bulletin of the Atomic Scientists, junio-julio de 2026

Value chain

Assets analysed

Risks shared with other dossiers

What would change my mind

• Would improve the thesis: fuel loading at Crane and grid synchronization before end of 2027; Westinghouse IPO valued with firm AP1000 orders behind; conversion of Oklo-Ohio MOU or TerraPower-data project into binding contracts with calendar and funding; uranium spot sustained above $100 with term rising
• Would worsen the thesis: capex guidance cuts from two or more hyperscalers in fall 2026 earnings calls; Crane delayed to 2028; second wave of massive dilution in startups without milestones; adverse PJM/FERC rule on co-location; any relevant nuclear safety incident; ERCOT and PJM weakening together, signal that data demand is not arriving

1. Fed hawkish shift: from expecting cuts to fearing hikes hecho · alta N3

   The 10-year US Treasury trades around 4.7% (31-Aug-2026) and the market assigns ~57% probability to a 25 bp hike in September after Jackson Hole, vs ~40% a week earlier; at early 2026 the market was pricing rate cuts and now it is pricing several hikes this year

   → It is the regime shift that invalidates the leg of the previous thesis betting on rate cuts in 2027 as a free option on Colonial and German residential; now that option has the opposite sign

   Trading Economics 31-ago-2026 / iShares-BlackRock jul-2026

2. REITs hold up despite rates: beat equities in 1H 2026 hecho · alta N3

   REITs outperformed broad equities by a considerable margin through mid-2026, reversing the 2025 trend and proving resilient in a high and rising rate environment

   → The mechanical relationship rates-up-REITs-down is not playing out in 2026 because operational growth (AI, supply scarcity) compensates; but an actual hike in September will put this resilience to the test

   Nareit, actualización de mitad de año, ago-2026

3. Equinix: third consecutive quarter of double-digit growth and raised guidance hecho · alta N1

   In 2Q 2026 (published 29-Jul-2026) recurring monthly revenue grew 11% year-over-year, gross bookings annualized 23% (second largest volume in its history) with record backlog, and raised 2026 guidance and long-term targets; lifts 2026 revenue growth to 11-12% and AFFO per share to 10-12%, and through 2029 expects revenue +10-13% annually

   → Growth is real and contracted, not narrative; the question is how much of this the price already reflects after the sector rerating

   Equinix, 8-K ante la SEC y nota de resultados, 29-jul-2026

4. Digital Realty: record backlog and FFO guidance raised, but with extraordinary items hecho · alta N1

   In 2Q 2026 it generated 1,900 M$ revenue (+29% year-over-year), closed with record backlog of 1,900 M$ annualized rent and raised Core FFO per share guidance (ex promote) to 8.15-8.20$; note: the quarter includes 188 M$ of promote from its joint venture and 94 M$ net from an insurance agreement

   → Clean Core FFO (2.13$/share ex promote in 2Q) is the figure to use for valuation; the extraordinary items inflate the quarterly picture

   Digital Realty, nota de resultados, 23-jul-2026

5. Merlin: data centers are already the engine — segment revenue x3.4 in one year hecho · alta N1

   Data center revenue rose from 13.5 to 46.4 M€ in one year and nearly 80% of half-year value creation came from this business (1H 2026 results reported to CNMV on 28-Jul-2026); total revenue of 308 M€ (+11.7%) and EBITDA of 229 M€ (+11.7%); closed June with 160 MW leased or preleased from the Mega Plan vs an annual target of ~200 MW that it assures it will exceed by far

   → The MRL-as-cheap-European-access-to-AI-axis thesis confirms in the numbers; additionally raises FFO per share guidance from 0.53 to 0.55 €, although the price already captures part of the success

   Merlin Properties vía CNMV, 28-jul-2026

6. Prologis-SEGRO: from hostile bid to FIRM agreement recommended hecho · alta N1

   On 4-Aug-2026 Prologis announced an agreement with SEGRO's board for a recommended acquisition valuing 100% at ~18,800 M$; at 1,031.7 pence per share, ~14,000 M£ in total; closing expected in first half 2027, subject to SEGRO shareholders, court and regulators

   → Validates that European logistics land with power capacity is a strategic asset; SEGRO moves from fundamental thesis to merger arbitrage (0.0920 PLD swap + partial cash alternative), with regulatory and market risk until 1H 2027

   Prologis, comunicado y 8-K, 4-ago-2026

7. Colonial: defensive that delivers, with AI peeking through in rents hecho · alta N1

   Closed 1H 2026 with recurring profit of 111 M€ (+4%), EPS of 17.18 cents, 94% occupancy, and AI and technology representing 30% of new rents signed; LTV of 36.7% in 1Q with over 70% of the 500 M€ divestment program executed; entry into Germany within 2026-2028 Strategic Plan

   → Clean balance sheet and correct execution, but without growth engine comparable to data centers; its rerating depended on rate cuts that are now receding

   Colonial vía CNMV y Renta 4, jul-ago 2026

8. Vonovia: confirms guidance but CUTS rental growth guidance, with LTV rising hecho · alta N1

   In 1H 2026 it confirmed guidance (adjusted EBITDA 2,950-3,050 M€) but cut rental growth forecast; the portfolio revalued +1.1% like-for-like (+848 M€), EPRA NTA per share came flat at 46.22 € and LTV rose to 46.0% from 45.4% with net debt/EBITDA at 14.0x; the stock was trading near 20.10 € on 20-Aug-2026

   → Discount to NTA exceeding 50%, but with high leverage (S&P flags 15.5x adjusted debt/EBITDA at Jun-2026) and no rate catalyst: the discount may persist; is potential value trap if Fed hikes and Bund follows

   Vonovia 1S 2026 (ago-2026) e Investing.com

9. China: relief in the 4 tier-1 cities, but hard data still in free fall hecho · alta N2

   By Aug-2026 the four tier-1 cities (Beijing, Shanghai, Shenzhen and Guangzhou) have eased purchase restrictions; Shanghai launched another round of measures on 20-Aug-2026 after July acceleration worsened; but real estate investment fell 19.2% year-over-year between January and July 2026 and construction starts down 24% and new home prices falling at a faster pace per data released 11-Aug-2026

   → DEMAND stimulus EXECUTED but incremental and municipal, not the big fiscal package; JP Morgan warns that the marginal effect of each round of easing is declining

   Bloomberg, Reuters, China Daily, ago-2026 / Oficina Nacional de Estadística china

10. Vanke: record losses and refinancing at the edge — the bellwether still ailing hecho · alta N2

    Vanke reported a loss of 2,200 M$ in 1H 2026 (26-Aug-2026) with debt pressure mounting, following a record annual loss of 89,000 M yuan (13,000 M$) in 2025, accumulating over 130,000 M yuan of losses in two years; additionally a Shenzhen court sentenced Evergrande founder Hui Ka Yan to life imprisonment on 20-Aug-2026

    → The AVOID from the previous report stands: with no land in fundamentals, the largest semi-state survivor burning capital and default risk still alive

    Bloomberg 26-ago-2026 y 20-ago-2026

Value chain

Assets analysed

Risks shared with other dossiers

What would change my mind

• WOULD IMPROVE the overall thesis if the Fed doesn't hike on 16-Sep-2026 and the 10Y returns below 4.3%: would reactivate defensive legs (COL, VNA) and extend the REIT rally beyond the AI axis
• WOULD DETERIORATE sharply if there is a hike in September AND language points to more: would move MRL and EQIX from buy-on-dips to watch, and VNA to avoid
• On the AI axis: a visible cut in CAPEX budgets from two or more hyperscalers for 2027 would invalidate half the report; data center contract renewals with increases below 10% would be the first warning
• In China: would move from AVOID to TACTICAL only with quantified CENTRAL fiscal stimulus aimed at buying inventory, plus three months of price stabilization; the municipal relief executed so far does not meet the threshold

1. Fed on pause but with explicit hawkish bias in minutes hecho · alta N1

   Rates held at 3.50-3.75% on July 29, 2026; minutes published August 19: many participants assessed that tightening would likely be warranted if inflation, currently above 3%, did not decline soon, and three FOMC members dissented in July arguing that rates should rise due to elevated inflation

   → The floor on US rates is in place; the directional risk is to the upside, not downside. Bad for long duration and demanding technology multiples

   Actas FOMC de julio vía prensa, 19 de agosto de 2026

2. Warsh at Jackson Hole: hawk without forward guidance hecho · alta N1

   On August 28, 2026, Warsh said inflation remains too high and suggested the central bank might need to raise rates in coming months; he stated that although summer readings were better than expected, they do not indicate to him that underlying trends have improved significantly, and after the speech Treasury yields rose sharply

   → Warsh deliberately removes forward guidance: higher risk premia structurally at each meeting because the market has no anchor

   Federal Reserve / CNBC / PBS, 28 de agosto de 2026

3. September is a coin flip and that is NOT in equity prices hecho · alta N2

   After Warsh's speech, Kalshi shows 48% to a 25 bp hike, federal funds futures near 56%, when before the speech hold had almost 70%; the swing was extreme: 67% probability of a hike on July 31, 44% on August 7 after the weak employment report

   → With probabilities at 50%, any outcome on September 16 moves market; asymmetry punishes long duration assets more if it hikes

   CME FedWatch, Kalshi, Polymarket, 28-30 de agosto de 2026

4. US inflation: improvement at the margin, far from target hecho · alta N1

   July CPI +0.1% monthly, 3.4% year-over-year headline and 2.5% core, both one-tenth less than June; the Fed itself raised its June 2026 PCE forecast to 3.6% headline and 3.3% core

   → Disinflation exists but is slow and energy-driven; it does not justify cuts and keeps the hiking option alive

   BLS vía CNBC 12 de agosto de 2026; SEP de la Fed 17 de junio de 2026

5. The long US bond remains broken as a safe haven hecho · alta N2

   Curve as of August 28, 2026: 2-year 4.36%, 10-year 4.72%, 30-year 5.21%; in August the 30-year auction was awarded at 5.22%, the highest yield in that maturity since 2001, and the long end cheapened due to supply, not data; also July deficit was $432.3 billion, the largest monthly since March 2021

   → Confirms the previous dossier decision: out of TLT-proxy; the problem is fiscal and term premium, not just Fed

   Tesoro de EE. UU. / Mariemont Capital / StreetStats, 12-28 de agosto de 2026

6. ECB: first hike in 3 years and pause that is not end of cycle hecho · alta N1

   On June 11, 2026 the ECB hiked 25 bp: deposit 2.25%, refinancing 2.40%, marginal 2.65% with inflation projected at 3.0% for 2026; on July 23 it held, and minutes stress that July's pause should not be interpreted as the end of the tightening cycle, with another hike likely absent clear improvement in inflation outlook. APP and PEPP balance sheets continue to shrink without reinvestment

   → Hawkish bias with QT ongoing: support for euro and pressure on peripheral debt; the September 10, 2026 meeting is the catalyst

   BCE (comunicados 11 de junio y 23 de julio de 2026; actas vía Trading Economics 27 de agosto de 2026)

7. BoE: aggressive hold with 6-3 fracture hecho · alta N1

   On July 30, 2026 the BoE held at 3.75% for the fifth consecutive meeting, with a 6-3 vote; Pill, Greene and Mann voted to hike to 4%; inflation rose to 2.9% in July due to energy and the market is pricing in two or even three quarter-point hikes in 12 months, with an implied rate of 4.33%

   → Contrast: nearly 90% of surveyed economists expect unchanged rates for the rest of the year; the economist-market gap is itself a source of pound volatility on September 17

   Bank of England 30 de julio de 2026; Fidelity International 21 de agosto de 2026

8. BoJ: hike to 1.00% executed and another almost discounted for September 18 hecho · alta N1

   The BoJ hiked 25 bp on June 16, 2026 to 1.0%, after holding 0.75% since December 2025, stating it will continue raising the official rate and adjusting the degree of monetary accommodation; as of August 24 traders assigned around 80% probability to a hike at the September 18 decision

   → At 80% probability, the hike itself is already mostly in the yen price; residual value lies in the subsequent path (another hike as soon as January 2027) and in Japanese banking margins

   Banco de Japón 16 de junio de 2026; Bloomberg 24 de agosto de 2026

9. PBOC: January promises unexecuted; tactical liquidity, not stimulus hecho · media N2

   China held its reference rates unchanged for the fifteenth consecutive month in August (1-year LPR 3.00%, 5-year 3.50%); in August the MLF provided 500 billion yuan against 600 billion in maturities, a net withdrawal of 100 billion, offset by 386 billion in 7-day injections and overnight operations of up to 600 billion daily between August 27 and September 1; in January Pan Gongsheng promised moderately lax policy in 2026 with RRR and rate cuts that remain nowhere in size

   → Confirms low conviction in China: ANNOUNCED stimulus is not EXECUTED stimulus; without a detonator, there is no entry

   PBOC vía Investing/análisis especializado, 20-27 de agosto de 2026; Xinhua 22 de enero de 2026

10. SNB: the only one clearly comfortable, with the franc as a red line hecho · alta N1

    Rate held at 0% on June 18, 2026 with inflation projected at 0.6% for 2026 and 2027 and 0.7% for 2028; the bank states it has greater willingness to intervene in the foreign exchange market to check rapid and excessive franc appreciation

    → The franc is safe haven with an artificial ceiling: the SNB will sell francs if it appreciates too much; we prefer gold and yen as safe havens rather than long CHF

    SNB, 18 de junio de 2026

11. Warsh proposes reducing meetings and removes guidance: change in communication regime hecho · alta N1

    July minutes reveal Warsh proposed moving to six scheduled meetings per year to allow information to accumulate between meetings; in June the statement completely removed guidance on future hikes

    → Less guidance = structurally higher risk premia at each meeting; favors gold and volatility strategies, penalizes duration leverage

    Actas FOMC julio 2026; comunicado FOMC 17 de junio de 2026

12. Narrative to watch: Warsh links AI to higher potential growth narrativa · media N2

    At Jackson Hole (August 28, 2026) Warsh signaled that AI could accelerate the economy, with growing capital flowing into AI infrastructure, and that the Fed is monitoring its impact

    → If the Fed believes AI raises potential growth, it tolerates higher rates for longer: reinforces AI CAPEX as a theme but makes its financing more expensive

    Yahoo Finance, 28 de agosto de 2026

Value chain

Assets analysed

Risks shared with other dossiers

What would change my mind

• The pro-gold and pro-yen thesis would improve if the Fed does NOT hike on September 16 with PCE near 3.7%: measurable loss of credibility in breakevens to the upside
• The Japanese banking thesis would worsen if the BoJ holds on September 18 AND softens guidance: it would close the USDJPY short and freeze entry into MUFG
• I would reassess long-term US bonds (from AVOID to TACTICAL) only with two consecutive months of job destruction and CPI <2.5%, or a credible fiscal plan
• I would activate the China dossier only with an executed RRR or LPR cut of at least 25-50 bp, not with forum promises
• A verifiable de-escalation with Iran that crashes crude >20% would dismantle half the case for ECB and BoE hikes and would make me rotate from gold toward short-term European duration

1. Record retail flow confirmed with first-half data hecho · alta N3

   European ETF industry recorded net inflows of 221.7 billion euros in the first half of 2026 and assets grew from 2.58 to 3.10 trillion euros between December 2025 and June 2026; ETFGI registers 45 consecutive months of net inflows

   → The shift from deposits to market is no longer a thesis but a statistical series; resilience was proven because quarterly ETF inflows exceeded 100 billion euros for the first time in the first quarter of 2026 despite the Iran conflict outbreak in late February

   LSEG Lipper (julio 2026) y ETFGI (13 de julio de 2026)

2. CORRECTION to previous dossier: the German law was not signed; the real milestone is from August 12, 2026 hecho · alta N1

   The federal cabinet approved on August 12, 2026 the Frühstartrente bill presented by the Finance Ministry; it is not yet law, must pass through Bundestag and Bundesrat, with target to close in 2026 to come into force January 1, 2027 with retroactive effect to January 2026 for the first cohort

   → The German catalyst remains alive but with residual parliamentary risk in the fourth quarter of 2026; I declare unverified the alleged May 2026 signature from the previous dossier

   Bundesfinanzministerium y Deutsche Rentenversicherung, 12-13 de agosto de 2026

3. The size of the German catalyst is small in direct money hecho · alta N1

   The subsidy is 10 euros per month, 120 euros per year, totaling 1,440 euros between ages 6 and 18; starts only with the 2020-born cohort; only certified standard deposits are subsidized with effective cost cap of 1.0%

   → Approximately 90-100 million euros annually of public money per cohort: irrelevant as flow, relevant as machine for opening accounts and creating habit; the 1% cost cap pressures margins and favors low-cost operators

   Bundesfinanzministerium FAQ y R+V/Sparkasse, julio-agosto de 2026

4. Toll collectors publish record quarters citing the theme hecho · alta N1

   flatexDEGIRO: revenue +26% to 167 million and net profit +55% to 61 million in the second quarter of 2026, and expects German retirement savings reforms to create additional growth opportunities from 2027; Euronext: first quarter 2026 record with revenue +15.3% to 528.5 million and retail investor in Euronext's book grew 50% versus prior year; Amundi: record net inflows of 32 billion in the first quarter and its ETF platform exceeded 400 billion under management by end of June

   → The toll collector thesis is confirmed in income statement, not just in industry flows

   Resultados de las compañías, abril-julio de 2026

5. Fatigue signal in customer acquisition by flatexDEGIRO hecho · alta N1

   flatexDEGIRO gross customer additions in the second quarter of 2026 (89,000) fell 13% year-over-year and 27% quarter-over-quarter, although assets under custody rose 30% to 108.3 billion euros

   → Current profit growth comes from activity and interest, not new customers; if additions keep falling, the market may question the multiple before the 2027 catalyst

   Presentación de resultados flatexDEGIRO, 22-23 de julio de 2026

6. The SIA remains a non-binding recommendation and monitoring is in 2027 hecho · alta N1

   The Commission recommendation of September 30, 2025 is not binding on member states; the Commission will monitor implementation via the European Semester and the interim review of the RIS in 2027

   → Each country is a separate mini-catalyst; there is no obligation or common timeline, so the risk of disappointment due to slowness is real

   Comisión Europea vía KPMG/PwC, octubre-noviembre de 2025

7. Spain: open consultation but no date or design hecho · media N2

   The Spanish Government opened public consultation for the creation of the savings and investment account, with deadline until January 30, 2026; there is currently no scheduled date to know if there will be an account in Spain or how it will be

   → Spain lags behind Germany; without defined tax incentive there is no domestic catalyst in 2026

   Ministerio de Economía vía finReg360/INESE, diciembre de 2025

8. Retail Investment Strategy: politically agreed but will not apply until 2029 hecho · alta N2

   Political agreement on December 18, 2025, technical trilogues until second quarter 2026 and Council endorsement in June 2026; member states will have 24 months to transpose the directive (likely until late 2028) and most provisions will apply six months later (mid-2029)

   → The bulk of pan-European regulation is 2028-2029 history, not 2026; moreover the industry itself (EFAMA) criticizes that the resulting regime is more complex than the current one, which tempers enthusiasm

   Consejo de la UE y Arthur Cox/Matheson, junio-agosto de 2026

9. Market integration package advances in parallel hecho · alta N1

   The Commission announced in early December 2025 a major legislative package to integrate EU financial markets, which will require broad negotiation in Parliament and Council during 2026; additionally the Council agreed in June 2026 its negotiating position on the PEPP review

   → Multi-year regulatory tailwind for pan-European exchanges and asset managers, but no profit impact before 2027-2028

   Comisión Europea, ESMA y Consejo, diciembre de 2025 - junio de 2026

Value chain

Assets analysed

Risks shared with other dossiers

What would change my mind

• WOULD IMPROVE the thesis: clean approval of the German law in the fourth quarter of 2026, a generous Spanish or Italian tax design, or evidence that broker customer additions reaccelerate without needing volatility
• WOULD WORSEN the thesis: market decline exceeding 20% before January 2027 launch, dilution of the German incentive in Parliament, or interim review of the RIS in 2027 showing near-zero national adoption
• CORRECTION already incorporated: the claim in the previous dossier that the German law was signed in May 2026 is not verified; the real milestone is the draft approved by Cabinet on August 12, 2026, still pending Bundestag and Bundesrat

1. Shanghai closes August near the 4,000 ceiling hecho · alta N3

   Close 3,986 on 31-08-2026, +4.02% for the month; 52-week range 3,702-4,258; year high (4,185) was in March, so the index remains in correction since then

   → August rebound led by state banks, not technology: it is a defensive-liquidity market, not macro conviction; the 4,000 level is resistance that only breaks on a catalyst

   Trading Economics / datos de cierre, 31-08-2026

2. PBOC parks cuts: key change vs previous dossier hecho · alta N2

   LPR rates unchanged 15 consecutive months through August 2026; PBOC eliminated explicit references to RRR and rate cuts in 1Q report; more than half of 33 economists (Bloomberg, June 2026) no longer expect cuts throughout 2026 and Goldman and Nomura abandoned their easing calls

   → The broad monetary stimulus leg of the previous thesis is invalidated in the short term; market support shifts to fiscal-regulatory and domestic liquidity, not rates

   Trading Economics 20-08-2026; FSM/informe monetario PBOC 1T26; Bloomberg 24-06-2026

3. Real estate reform of 28-08-2026: structural, not a check hecho · alta N2

   PBOC and NFRA require releasing mortgages only after work completion, extend mortgages to 40 years, prioritize finished housing sales and CSRC will support mergers of listed developers; Pinpoint AM rated it stronger than expected; developers fell due to financing fears and prices continue falling 0.2-0.3% monthly

   → Fixes delivery risk (buyer confidence) but not demand, and in short term tightens leveraged private developers; state banks and state developers win share, not a reason to buy the sector

   SCMP, Nikkei, Investing.com, 28/31-08-2026

4. Weak macro confirmed: no floor yet hecho · alta N1

   2Q26 GDP +4.3% (worst since 4Q22, below target 4.5-5%); July: retail sales +0.6%, unemployment rising; August composite PMI 49.5 with manufacturing 49.8 and services 49; real estate investment -18% in 1H26 and total fixed investment -5.7%

   → Stock market rises against a slowing economy: support is liquidity, technological substitution and policy expectations, a fragile balance if the September catalyst fails

   NBS vía CNBC 15-07 y 17-08-2026; PMI oficial 31-08-2026

5. Xi-Trump summit with firm date: 24-09-2026 hecho · alta N2

   State visit confirmed after Trump's trip to Beijing in May (agreements: 200 Boeing planes, USD 17 billion annually in agricultural purchases through 2028, oil purchases); prepared by Bessent-He Lifeng call on 30-07-2026; open frictions: US plan to ban Chinese robots and power investors, rare earths, possible Chinese arms to Iran

   → Dated binary catalyst: success consolidates truce in managed trade framework; failure reactivates sanctions with market having almost no cushion in expensive tech

   CNBC 15/18-05-2026; Axios 30-07-2026; Japan Times 31-07-2026

6. Alibaba: AI capex eats profit but cloud accelerates hecho · alta N1

   Quarter to Jun-2026 (published 20-08-2026): revenues +9% (RMB 268.9 billion), cloud +45% with 12% EBITA margin and EBITA +133%, AI ARR >RMB 16 billion, capex +75% (RMB 67.7 billion), net profit -76%, non-GAAP -38%; stock fell ~8.6% post-earnings; already spent half of its RMB 380 billion plan through 2029

   → The underlying thesis (China builds its own AI cycle with verifiable revenues) is reinforced, but market punishes margin compression: same pattern as US hyperscalers with much lower multiple

   6-K en SEC y nota de resultados, 20-08-2026

7. Nvidia substitution: real but with physical bottleneck hecho · media N3

   Cambricon targets ~500,000 accelerators in 2026 (from 116,000 in 2025) with reported ~20% yields on SMIC's N+2 node; its flagship chip performs like Nvidia's 2020 A100; Beijing included it alongside Huawei in first government AI chip purchase list; its stock reached being China's most expensive (~CNY 1,680)

   → Trading volume is real (ByteDance more than half of orders) but valuation discounts perfect execution on a foundry with 4 of every 5 chips defective: binary risk profile that we confirm AVOID

   Bloomberg vía TrendForce/Tom's Hardware dic-2025; The Substrate jun-2026

8. Hong Kong tech has corrected much more than A-shares hecho · alta N3

   Hang Seng Tech ~4,700 on 20-08-2026, -15% in 12 months, with 52-week range 4,230-6,715 (a -30% from highs); Shanghai +2.9% year-on-year and Hang Seng at 25,585

   → Asymmetry has shifted: risk-reward is today in corrected Hong Kong tech, not in domestic A-market holding up on liquidity and banking

   Yahoo Finance / Investing.com, 20/28-08-2026

9. Regulators slow leverage: they want slow bull hecho · alta N2

   In January 2026 exchanges raised margin requirement from 80% to 100% after record volume of RMB 3.6 trillion daily and margin balance of RMB 2.65 trillion; net foreign inflows exceeded USD 50 billion in prior months

   → Beijing actively manages market pace: limits explosive potential but also reduces 2015-style purge risk; supports thesis of broad index exposure not retail momentum

   Xinhua 14-01-2026; CNBC 20-01-2026; Bloomberg 13-01-2026

Value chain

Assets analysed

Risks shared with other dossiers

What would change my mind

• Would improve thesis: 24-09 summit with verifiable agreement (signed reduced tariff, chips framework), surprise PBOC cut in 4Q, or two consecutive quarters of BABA/Tencent cloud margin improvement
• Would worsen thesis: sanctions on Chinese banks or cloud after summit, real estate prices accelerating declines despite reform, PMI below 49 for three months without fiscal response, or ByteDance withdrawal as anchor client for domestic chips
• Complete framework invalidation: Iran escalation with direct Chinese involvement reactivating maximum sanctions regime

1. SpaceX trades and has already lived through its first euphoria-hangover cycle hecho · alta N2

   SpaceX went public on June 11-12, 2026 selling 556 million shares at 135 dollars, the largest IPO in history; as of August 30, 2026 it trades at 141.50 dollars with a 52-week range of 104.83 to 225.64

   → The 37% correction from highs eliminates part of the euphoria premium; current price is only 5% above IPO, improving entry point versus prior dossier

   CNN 12-jun-2026; Investing.com 30-ago-2026

2. SPCX beats earnings but valuation remains extreme hecho · alta N2

   Q2 2026 results (Aug 4, 2026): revenue of 7.81 billion vs 6.82 expected, EPS -0.09 vs -0.21 expected; at ~2 trillion market cap implies ~64x annualized sales. Morningstar notes the valuation implies waiting decades for earnings to justify it

   → Growing faster than expected but the multiple still discounts total dominance; also announced Aug 25, 2026 a 100 billion Louisiana spaceport: giant CAPEX delaying free cash

   Google Finance / Morningstar, agosto 2026

3. RKLB: record results and transformation via M&A hecho · alta N1

   Q2 2026 (Aug 10, 2026): record revenue of 234 million (+62% year-over-year), record backlog of 2.36 billion (+137%) and over 1 billion in new contracts already signed in Q3; Q3 guidance of 250-265 million with adjusted EBITDA loss of 17-23 million

   → Growth is real and contracted but the company remains unprofitable: the thesis depends on backlog converting to margin

   8-K de Rocket Lab en SEC, 10-ago-2026

4. RKLB buys Iridium: leverage for recurring cash hecho · alta N1

   Deal of June 29, 2026 at 54 dollars/share in cash and stock, 8 billion enterprise value, 24% premium; Iridium earned 871.7 million in 2025 with 114.4 million net income versus 601.8 million revenue and 198.2 million loss for Rocket Lab in 2025; will require over 3 billion in cash plus ~1.8 billion to refinance Iridium debt, expected close in 2027

   → Buys cash and spectrum (shortcut to Starlink model) but with a 3.6 billion bridge loan from Deutsche Bank and Wells Fargo and stock issuance: dilution and debt are the price

   10-Q y nota de prensa en SEC; SpaceNews 29-jun-2026

5. Neutron still hasn't flown: the binary catalyst narrows hecho · alta N2

   The first-stage tank rupture during hydrostatic testing on Jan 21, 2026 pushed mid-year debut to late 2026; exact timing depends on first-stage qualification and other critical tests in late 2026

   → RKLB's key binary event remains pending; a failure or new delay to 2027 would hit the multiple and credibility just as it digests Iridium

   Spaceflight Now, 10-ago-2026 (call de resultados)

6. Golden Dome moves from announcement to verifiable contracts hecho · alta N1

   The FY2026 defense budget law approved Feb 3, 2026 included 13.4 billion for Golden Dome; Space Force announced Apr 24, 2026 20 contracts worth up to 3.2 billion across 12 companies for space interceptors; SDA awarded Jul 13, 2026 up to 955 million to L3Harris and up to 798 million to Sierra Space for 18 missile-tracking satellites each; SpaceX received 1.6 billion for 18 launches through end 2027

   → Budgeted and awarded money, not promises; caveat: the space interceptor OTAs are prototypes with demonstration required in 2028 of a capability not yet proven, not production runs

   Congreso/Space Force/SDA vía Air & Space Forces Magazine y GovConWire, feb-jul 2026

7. LUNR mutates from lunar bet to defense contractor hecho · alta N1

   Q2 2026: revenue of 206.17 million (vs 50.31 a year prior), net loss expanded to 46.45 million; record backlog of 1.8 billion (Aug 13, 2026); selected by L3Harris to deliver 18 IM-300 platforms for an anti-missile defense mission

   → Thesis changes: no longer binary lunar lottery but manufacturer with defense backlog; still burning cash and margin remains to be proven

   Nota de resultados GlobeNewswire 13-ago-2026; Simply Wall St

8. ASTS executes launch calendar but money arrives in 2027 hecho · alta N2

   BlueBird 11, 12 and 13 launched Aug 5, 2026 on a Falcon 9 with production advancing to satellite 42; 2026 guidance reaffirmed of 150-200 million revenue despite missing Q1 consensus (14.73 million); significant commercial service and government revenue expected for 2027

   → Real factory execution but pays its launches to its direct rival (SpaceX) and analysts are cutting price targets: BofA down to 80 and UBS to 78, both Neutral

   Nota de prensa ASTS 28-jul-2026; stockanalysis.com ago-2026

9. Orbital data centers: demos yes, business no hecho · media N3

   Starcloud-1 (launched Nov 2, 2025) ran an Nvidia H100 and trained a small language model in orbit; Starcloud raised 170 million in March 2026 at 1.1 billion valuation and asked FCC for a constellation of 88,000 satellites; Axiom launched its first two orbital data center nodes on Jan 11, 2026; Google Suncatcher will launch two prototype satellites in early 2027

   → Confirms prior thesis: it's demo CAPEX without a paying customer; the investable route remains launchers who charge to lift mass (SPCX, RKLB) not promises of orbital gigawatts

   Quartz jul-2026, GeekWire mar-2026, Introl ene/feb-2026, Fierce Network

10. FLY stabilizes after collapse but carries IPO litigation hecho · alta N2

    Q2 (Aug 17, 2026): EPS loss of 0.42 vs 0.57 expected, reaffirms 2026 revenue guidance of 420-450 million; closes at 26.54 on Aug 12, 2026 versus August 2025 high of 60.35; the company, executives and underwriters seek to dismiss securities lawsuits over the IPO (Aug 2026); won a 144 million NASA CLPS contract for Blue Ghost

    → Modest operational improvement with legal risk open; remains aggressive small position, no urgency to enter

    CNBC/TipRanks y Macrotrends, agosto 2026

Value chain

Assets analysed

Risks shared with other dossiers

What would change my mind

• I would upgrade RKLB thesis to conviction 8-9 if Neutron reaches orbit on first attempt before end 2026 and Iridium close advances without repricing
• I would upgrade SPCX to COMPLEMENTARY with aggressive entry if it falls to 105-120 zone without operational deterioration, or if free cash appears earlier than expected despite Louisiana
• The entire dossier worsens if Golden Dome suffers FY2027 budget cuts or if a media launch accident freezes licensing
• LUNR and ASTS would worsen if they announce capital raises exceeding 10% before showing sustained positive gross margin
• Orbital data centers would shift from narrative to investable thesis if Starcloud-2 (October 2026) or Suncatcher (early 2027) demonstrate competitive per-hour compute cost and the first commercial contract with a number appears

1. Unitree: first pure humanoid listed, with debut euphoria hecho · alta N2

   IPO set for August 6, 2026 at 150.80 yuan/share, ~9 billion dollar valuation, equivalent to 219x 2025 earnings and ~36x sales; debut on August 19 with +629% intraday and +487% close (~53 billion), retail oversubscription of 8,000x

   → The public market already priced the theme and it is pure euphoria; buying the pure at 1,300x earnings is paying a perfect 10-year scenario at today's prices

   Caixin, prospecto STAR Market, trackers verificados contra CNBC/Bloomberg, agosto 2026

2. Unitree's growth decelerates just as it goes public hecho · media N3

   H1 2026 forecast: revenue of 1,052-1,128 million RMB (+35.6% to +45.4%) but excluding-extraordinary profit declining between 6.4% and 22%, vs ~335% growth in 2025

   → Margins falling while valuation explodes: the asymmetry is clearly negative in the Chinese pure play

   Prospecto de Unitree (30 julio 2026) vía análisis del folleto

3. US blocks foreign robots: regulatory bifurcation of the market hecho · alta N1

   On July 28, 2026 the FCC added advanced robotic devices of foreign production to the Covered List: new models cannot obtain the necessary authorization to be imported and sold except conditional approval; defined as mobile robots, humanoids and quadrupeds; it is the third category expansion in 8 months (drones in Dec-2025, routers in Mar-2026)

   → Negative for Unitree and Chinese in the US (US was ~13-18% of their revenue); structural positive for American, Korean and Japanese manufacturers and components

   FCC Public Notice DA 26-786, 28 julio 2026, y análisis de Sidley/Skadden/Baker McKenzie

4. Tesla Optimus: the red flag is confirmed hecho · alta N2

   As of July 23, 2026 formal production at Fremont has NOT started: the July 22 Q2 update confirms lines under installation, production 'anticipated for year-end', no date, no units, no price; first units will go to internal Optimus Academy, not customers

   → The catalyst we were monitoring resolves negatively: Tesla has never published a single Optimus unit number; the 1 million/year narrative is design capacity, not production

   Tesla Q2 2026 shareholder update (22 julio 2026) vía trackers independientes

5. XPeng crystallizes value in robotics but continues burning cash hecho · alta N2

   Robotics unit raises over 900 million dollars at 6.3 billion valuation (IDG, Tencent, Alibaba), record for Chinese embodied AI; spins off as subsidiary maintaining XPeng ~82%; target of over 1,000 IRON/month, mass production by end of 2026, commercial deliveries in 2027; but net loss in Q2 nearly tripled to 1.34 billion RMB

   → The bullish catalyst we were monitoring advances: the external 6.3 billion valuation gives floor to optionality, but no timeline for profitability or volume guidance for robotics

   XPeng resultados Q2 2026 (24 agosto) y anuncio de ronda; Electrek/CnEVPost

6. China concentrates volume; human remains better hecho · media N2

   Chinese manufacturers ~97% of global H1 2026 shipments (~19,000-22,000 units); AgiBot exceeded 15,000 cumulative units on July 1; but Unitree founder himself admitted in Beijing that robots are not yet as efficient as humans and learn slowly and MERICS notes that UBTech described its Walker S2 as half as efficient as a human worker in early 2026

   → Brooks is still right: volume yes, productivity not yet; the gap between shipments and real economic value is the central risk of the theme

   Trackers sectoriales, CNBC (21 agosto 2026), MERICS

7. Figure AI: real deployment at BMW and factory scaling, demanding private valuation hecho · media N3

   Figure 03 in production at 1 robot/hour, over 1,000 units exceeded; Figure 02 completed 11 months at BMW Spartanburg contributing to over 30,000 vehicles with 1,250+ hours; private valuation of 39 billion with secondary trading below, signal that market wants to see revenue before paying more

   → The most solid Western commercial proof, but not directly investable and with private ticker that secondary no longer validates

   Figure AI/BMW comunicados, Sacra, trackers de secundario, agosto 2026

8. NVDA: the toll remains intact although robotics is immaterial in its P&L hecho · alta N1

   Q2 fiscal 2027 (closed July 26, 2026): revenue of 96.2 billion, +106% year-over-year, gross margin 75%; Jetson Thor available from August 10 with Agility, Amazon Robotics, Boston Dynamics, Figure and Meta as early adopters; the Edge segment (robotics+automotive) contributed 6.4 billion and robotics is not broken out

   → NVDA charges toll whoever sells the robot, but whoever buys it for robotics reasons gets it wrong: today it is a free option inside a data center thesis

   Nota de resultados NVIDIA (SEC 8-K, 26 agosto 2026) y newsroom

9. Symbotic: robotic logistics crosses to GAAP profit hecho · alta N1

   FQ3 2026: revenue 721 million (+22%), net profit of 55 million vs loss of 21 million a year prior; adjusted EBITDA 95 million, more than double; contracted backlog of 22.5 billion; but negative free cash flow of 164.6 million in the quarter due to collection timing and stock fell 26.4% in the last year

   → Robotics that already invoices improves its profit quality while market punishes it: buying profile on correction with multi-year backlog

   Symbotic press release y 10-Q (5 agosto 2026)

10. Teradyne: robotics grows but the engine is semi testing hecho · alta N1

    Q2 2026: total revenue 1.329 billion (+104%), of which 100 million is robotics; fifth consecutive quarter of robotics growth, +33% year-over-year; robotics is only 8% of total, down from 12% a year ago

    → Bought for the AI test cycle with robotics as afterthought; robotics thematic purity is today lower than a year ago

    Teradyne press release (28 julio 2026)

11. Consensus: 2026 as year of production doubling inferencia · media N3

    Citi survey of Chinese chain: global humanoid production entering exponential growth in 2026, at least doubling vs 2025, with demand concentrated in 3D vision, reducers and spindles

    → Reinforces components thesis, but is bank opinion on small base: doubling from ~20,000 units remains a tiny market in euros

    Citi Research, encuesta a 5 empresas de la cadena china, 2026

Value chain

Assets analysed

Risks shared with other dossiers

What would change my mind

• I would improve the pure-play thesis if Unitree corrects more than 70% and re-accelerates revenue, or if a Western deployment (Figure/BMW, Apptronik) publishes robot productivity metrics comparable to human labor cost
• I would improve XPEV to COMPLEMENTARY if IRON delivers thousands of units to external customers with published price and margin in 2027
• Everything would worsen if a serious humanoid accident triggers regulatory moratorium, or if hyperscaler AI capex cuts and dries financing across the entire chain
• SYM would worsen if negative FCF repeats two quarters without calendar explanation or if Walmart renegotiates

1. Waymo operates at real commercial scale and is the undisputed leader

   As of August 2026: 11 US metropolitan areas, approximately 3,500 vehicles, ~500,000 paid rides per week, over 20 million cumulative rides and target of 1 million weekly by end of 2026. $16 billion funding round valuing it at $126 billion. London and Tokyo in preparation.

   → The robotaxi is no longer a promise: it is a business that charges half a million rides per week and grows. GOOGL has inside an asset valued at $126 billion that the market barely discounts.

   thechargeport.com/robotaxi-tracker, Bloomberg Tech via Investing, Engadget

2. Peer-reviewed science validates Waymo's safety

   Kusano et al. study in Traffic Injury Prevention on 56.7 million driverless miles: statistically significant reduction in injury accidents versus humans; up to 96% fewer injury accidents at intersections. At 7.1 million miles: 0.6 injury incidents per million miles versus 2.80 for humans (-80%).

   → The safety argument is the regulatory moat: with such scientific data, regulators have basis to continue authorizing expansion. Reduces risk of a complete sector halt.

   Traffic Injury Prevention / PubMed / arxiv.org/pdf/2312.12675 / waymo.com/safety/impact

3. The regulator tightens: two Waymo recalls in 2026 and NHTSA directive

   May 2026: withdrawal of ~3,800 robotaxis for entering flooded areas. June 2026: ~3,900 for entering highway construction zones (13 incidents), with service pause on highways until late July. NHTSA required correction plans by end of July 2026 for interference with emergency scenes, and there is open investigation into an incident with a school bus and a child.

   → The number one risk to the sector is not technology but a high-profile accident or fine. Software recalls are fixed quickly, but they limit expansion speed and can move the stock short-term.

   NHTSA via CNBC, Fox Business, TechCrunch, thechargeport.com

4. Tesla advances for real but its real scale is tiny versus the narrative

   Unsupervised service in 6-meter radius (Austin, Dallas, Houston, Miami, Orlando, Tampa), expanded hours to 6-22h from August 2026, but only ~85 active vehicles according to independent trackers; in California still with safety driver due to lack of permit. Cybercab in production at Giga Texas for employees only. In favor: 56% cheaper than Waymo on comparable trip. Against: 15 accidents reported to NHTSA and crash rate worse than humans according to preliminary data from critical sources.

   → TSLA prices the dream of millions of robotaxis but operates fewer than 100. Pure optionality: if FSD v15 rewrite (late 2026) works, it soars; if not, there is much air in the price. Small position or none.

   driveteslacanada.ca, thechargeport.com/robotaxi-tracker, electrek.co, basenor.com

5. Uber confirms itself as the sector toll without burning capital making vehicles

   Q2 2026 record: gross bookings +22% to $58 billion, free cash flow of $10 billion in 12 months for first time. Commits over $10 billion to autonomous vehicles, target of 15 cities with autonomous service by end of 2026 and multiple partners (Waymo, Pony, Wayve, Waabi, Nuro, Lucid, WeRide). Key: where robotaxis outside its platform exist (Phoenix, LA, SF), its rides grew faster, not slower.

   → The data that Uber accelerates precisely in cities with robotaxis dismantles the fear that Waymo kills it. Uber wins with or without a driver: it is the safest way to play the theme.

   investor.uber.com, CNBC, transcripcion Q2 2026

6. Zoox (Amazon) achieves first federal exemption and already charges

   On July 30, 2026 NHTSA granted Zoox the first commercial exemption from regulations requiring steering wheel and pedals; on August 10 it started charging for rides in Las Vegas. Accumulates over 3 million miles and close to 1 million passengers in 4 cities.

   → Huge regulatory precedent: vehicles designed without controls can now charge in the US. Benefits the entire sector (including Tesla's Cybercab) and gives AMZN another leg of growth little discounted.

   CNBC, Reuters, Axios, fox5vegas.com

7. Pony AI: the most explosive percentage growth, via Uber in Europe

   Q2 2026: robotaxi revenue +691% to $12.1 million (charged fares +849%), fleet of ~2,000 vehicles with target of 3,500 by year-end and 20 cities. Commitments for over 4,000 vehicles with Uber and other partners, including 2,000+ robotaxis in five European cities. But net loss of $59.8 million in the quarter.

   → Real growth but tiny base and strong cash burn: PONY is high-risk speculative bet. What matters is that it validates Uber's model as global distributor of third-party robotaxis.

   globenewswire.com (resultados PONY), Motley Fool transcript, stocktitan.net

8. Lidar supplier Hesai already profitable: picks and shovels work

   First half 2026: revenue of $227 million (+25%), ADAS lidar shipments +87% and robotics +165%, fifth consecutive GAAP profitable quarter. Design wins with Volkswagen (China), Great Wall, Mercedes L3, and its ETX lidar in negotiation (RFQ) with robotaxi operators.

   → Supply chain already makes money today, without waiting for robotaxi to be profitable. HSAI is the picks and shovels of the sector, although with China-US geopolitical risk.

   SEC 6-K de Hesai, investor.hesaitech.com, automotiveworld.com

Value chain

• Robotaxi service at scale → GOOGL via Waymo: 500,000 paid rides/week, valued at $126 billion within Alphabet
• Distribution platform and toll → UBER: charges commission from Waymo, Pony, Wayve and WeRide without making anything, and accelerates where robotaxis exist
• Aggressive all-or-nothing bet → TSLA: ~85 cars without supervision today, Cybercab in production, everything depends on FSD v15
• Proprietary design robotaxi with federal exemption → AMZN via Zoox: charging in Las Vegas since August
• Chinese operator expanding via Uber → PONY: +691% in robotaxi revenue but burning cash
• Eyes of the car (lidar) → HSAI: profitable, market share leader in China, in negotiation with robotaxi operators; OUST/LAZR alternatives much weaker
• Brain of the car (chips) → NVDA (Drive/Thor platform) and QCOM (Snapdragon Ride); MBLY as pure player of driving systems
• Chip manufacturing → TSM, wins whoever wins
• Base vehicles for fleets → Hyundai (Ioniq 5 for Waymo) and Zeekr; do not price the theme well in the US

1. First clinical validation of an AI-discovered drug already in Phase III

   Rentosertib (Insilico Medicine), TNIK inhibitor for pulmonary fibrosis discovered with generative AI: Phase IIa published in Nature Medicine with +98.4 mL of lung function versus -20.3 mL with placebo; Phase III started July 2026

   → AI drug discovery is no longer a promise: there is clinical proof published in the most serious medical journal. If Phase III succeeds, it would be the first AI drug approved and would validate the business model of the entire subsector

   Nature Medicine (jun 2025) + comunicado Insilico/PRNewswire (jul 2026)

2. Large pharmas pay billions to access AI drug design

   Lilly signed with Insilico an agreement worth up to 2.75 billion USD (115 million USD upfront); Isomorphic Labs (Alphabet) closed Series B of 2.1 billion USD in May 2026 and adds partnerships with J&J (January 2026), Lilly and Novartis for almost 3 billion USD committed

   → Big money is already flowing with signed contracts, not just demos. The listed way to play Isomorphic is GOOGL (owner); the pharma way is LLY and JNJ that keep the resulting drugs

   Comunicados Isomorphic Labs, J&J Innovation, Goodwin Law, curaesalud.substack.com

3. No AI-designed drug yet exists approved by FDA

   As of February 2026, zero AI-designed drugs with FDA approval; Insilico has 28 candidates and half in trials, but none approved

   → The gap between enthusiasm and real cash remains open: trials take years. Pure AI-drug bets are long-term and high-risk; near-term benefit is with whoever sells data, diagnostics and picks and shovels

   Sermo (informe feb 2026), curaesalud.substack.com

4. MASAI trial (Lund, published in The Lancet 2026) confirms AI in cancer screening

   First randomized AI trial in cancer screening with over 105,000 women: 29% more cancers detected without more false positives, 44% less radiologist workload and fewer aggressive cancers among screened

   → Maximum quality evidence (not marketing) pushing European healthcare systems to deploy AI in screening. Winners are image equipment and AI manufacturers: GE HealthCare, Siemens Healthineers, Philips

   The Lancet 407(10527):505-514 (2026), Universidad de Lund

5. Tempus AI: AI oncology diagnostics already profitable

   Q2 2026: revenue 382.5 million USD (+22%), first net profit (5.6 million USD versus -42.8 million USD a year ago), annual guidance ~1.6 billion USD, Personalis acquisition for ~1.5 billion USD and first FDA approval of companion diagnostic tumor-only and tumor-normal

   → TEM crossed the profitability line: it is the most direct and already verifiable way to invest in AI applied to oncology, although it trades expensive and Personalis integration adds risk

   Nota de resultados Tempus AI 30-jul-2026 (investors.tempus.com) y 8-K en SEC

6. FDA: over 1,500 AI medical devices authorized, but weak evidence

   1,524 AI devices authorized through Q1 2026 (76% radiology); GE HealthCare leads with 120 authorizations; but only 1.6% of 691 reviewed cited randomized trial and 5.8% was withdrawn, almost always for software failures

   → Regulatory volume grows fast (real growth), but weak evidence opens risk of withdrawals and regulatory tightening (FDA guidelines 2026 and EU AI Act from August 2026)

   Lista oficial FDA de dispositivos IA + análisis IntuitionLabs (jul 2026)

7. The sector's cash engine: obesity, now also in pill form

   Lilly Q2 2026: revenue 22.974 billion USD (+48%), annual guidance raised to 85-87 billion USD; Mounjaro 9.943 billion USD (+91%). The orforglipron pill (Foundayo) launched in Q2 2026 invoiced 98 million USD in its first quarter; Novo's Wegovy oral made ~855 million USD in the semester

   → LLY confirms its dominance and pills open a market even larger than injections. Watch out: FDA requested extra hepatic and cardiovascular safety data on orforglipron and Novo sued Lilly over advertising; noise, but the duopoly remains intact

   Resultados 2T 2026 de Eli Lilly y Novo Nordisk (via Vozpopuli, Yahoo Finance, BigGo)

8. Recursion: platform validation but falling revenue

   Q2 2026: Genentech exercised option on first neuroscience target from collaboration (external validation), cash until 2028, but revenue of only 7.7 million USD (-60%) and stock -23% on the year

   → Pure AI biotechs (RXRX) are optionality: growing scientific validation but no business yet. Only for a small speculative position, not as a pillar

   8-K de Recursion en SEC (5-ago-2026), Quiver Quantitative

9. Robotic surgery: real and recurring growth

   Intuitive Q2 2026: revenue 2.89 billion USD (+19%), procedures +16% (Ion +36%), 468 da Vinci systems placed versus 395 a year ago, 85% of recurring revenue

   → ISRG remains the mature business with a moat in the theme: grows double-digit with recurring revenue, although trades at high premium (forward PER in the 30s) and U.S. moderates

   Resultados 2T 2026 Intuitive Surgical (16-jul-2026)

Value chain

• AI models to design drugs → Isomorphic Labs, playable via GOOGL (owner) and via its pharma partners JNJ, LLY, NVS; pure and speculative bet: RXRX
• Large pharmas that buy innovation and keep the drug → LLY (agreements with Insilico and Isomorphic, obesity cash engine), JNJ, NVO, MRK
• Oncology data and AI-powered precision diagnostics → TEM (Tempus), already profitable and acquiring Personalis for recurrence detection (MRD)
• Imaging equipment with FDA-approved AI → GE HealthCare (GEHC), Siemens Healthineers and Philips (PHG), leaders in authorizations; the MASAI trial gives them tailwind in Europe
• Robotic surgery already invoicing → ISRG, near-monopoly with 85% recurring revenue
• The chips training all these models → NVDA, collects toll whoever sells the drug or the diagnostic