THE THESIS
What we believe — and exactly what would prove us wrong.
26 falsifiable claims1 linked nameas of September 29, 2026
Energy
Energy & Grid
2 claims · 0 strained
- Energy and grid capacity binds AI capex: dispatchable additions (~5-8 GW/yr) cannot match demand (NERC +69%/+224GW peak revision) on the 2026-2027 timescale. The constraint is thermodynamic, not capital-allocation (Smil's 50-70yr transition iron law + Jevons), and hyperscaler nuclear PPAs (~1.8 GW by 2030) cover ~2% of self-created demand.high
- As AI inference commoditises and platform-layer margins compress, value migrates to the physical-input layer — energy, materials, grid — where picks-and-shovels suppliers earn the revenue the software layer fails to translate. Load-bearing signal: PJM's capacity auction cleared at the FERC-approved cap three years running ($329.17 -> $333.44 -> $325.00/MW-day, 2026/27-2028/29), with the RTO short of its reliability requirement; DeepSeek's sub-$6M training is the commoditisation trigger.high
Materials
Materials & Critical Minerals
0 claims · 0 strained
Chips
Chips & Compute
1 claim · 0 strained
Eldercare
Eldercare & Pharma
0 claims · 0 strained
Credit
Private Credit
5 claims · 0 strained
- USD reserve-currency dominance is structurally durable on a 5-10yr horizon despite fiscal strain (C-01); BRICS+ alternatives lack the institutional depth and dealer-of-last-resort plumbing to displace USD within the window C-01 stress matures. The GENIUS-Act stablecoin regime now deepens this: statutory Treasury-buy obligations make stablecoin issuers a new structural (mechanical, not informed) marginal Treasury buyer.medium
- Passive flows + stablecoin Treasury demand + insurance balance sheets are the marginal price-setters in equities and Treasuries; the informed value investor is no longer the price-setter. The same demographic cohort that built the passive bid mechanically reverses it via RMD-driven decumulation (C-13) — symmetric, no catalyst required.medium
- The AI-capex bubble and private credit are fused through the same end-investor: hyperscaler capex is increasingly debt-financed (>$100B new issuance in 2025) via GPU-collateralised facilities (CoreWeave-style) into life-insurance balance sheets (~1/3 of US life insurance assets now in private credit). AI-capex stress therefore transmits through private credit into annuity obligations — landing the loss on retirees, not just equity holders.medium
- US private credit's true default rate is ~6% when PIK-by-amendment is treated as distressed exchange — roughly 3x KBRA's ~2.1% headline — with PIK-by-amendment rising from 2.6% (2021) to 6.1% (Q3 2025) and Fitch TTM defaults at ~5.8% (Jan 2026). Opacity routes the loss through life-insurer balance sheets to annuity holders, not informed credit investors.high
- Office CMBS delinquency reached a record ~12.3% (Jan 2026), above the 2008 peak; the 2026 CRE maturity wall is ~$875B (MBA, Feb 2026), down from $957B in 2025; separately ~$300B of private credit sits in semi-liquid retail-redemption structures against illiquid underlying. Both are queued fragilities in the same liquidity-coupling system, with depleted absorption capacity because regional banks (>70% of US CRE) have carried three years of mark-to-model deferral.high
—
Cross-Cutting & Macro
18 claims · 0 strained
- US sovereign fiscal arithmetic is unsustainable; net interest exceeds defence spending; structural deficit ~$2T/yr forces a choice between inflation, cuts, or financial repression.high
- Aggregate US equity valuation is inconsistent with positive forward real returns; multiple compression toward historical mean implies ~50% real price decline over a multi-year horizon. Record operating margins (~21%) are the accounting-identity mirror of the fiscal deficit + low household savings, so they compress mechanically when those flows reverse — and the gap can close through TIME (flat nominal prices as earnings grow, 1966–82 analogue), no discrete catalyst required.high
- The US household balance sheet is the binding political-buffer floor: ~53% of Americans cannot cover a $1,000 emergency (Bankrate 2026), revolving APR >21%, subprime auto 60-day delinquency at a multi-decade record (~6.9% Jan 2026 vs ~0.4% prime), bottom-50% holds ~2.5% of wealth — leaving weeks, not months, of shock-absorption and little political space for fiscal adjustment.high
- AI removes entry-level white-collar roles first (Brynjolfsson: ~13% relative employment decline for ages 22-25 in the most AI-exposed occupations since late 2022; Goldman: ~-16k jobs/month net), severing the wage/payroll-tax pipeline that replenishes both the consumer base and federal revenues — feeding directly back into the fiscal arithmetic (C-01) and breaking the C-04 productivity AND consumer-revenue bull cases simultaneously.medium
- The post-1980 era of disinflation, low real rates, and rising asset prices was a one-time labour-supply event (China + Eastern Europe integration + female participation surge), not central-bank achievement; that tailwind has fully reversed (China working-age -68M since its 2014 peak). The next 2-3 decades default to upward goods inflation, upward wage share, downward profit share, downward financial-asset prices.high
- Japan's 34-year post-bubble case — the most aggressive monetary+fiscal policy in modern history (BoJ owns ~49% of JGBs, balance sheet ~100% of GDP, debt ~260% of GDP) — produced ~zero real equity recovery (Nikkei real ~0.24%/yr 1989-2023) and real residential property -35.9% from the 1991 peak. The mechanism is institution-invariant: same direction in Korea, China, Italy, Germany.high
- All projected US working-age growth through 2050 depends on net immigration (Census 2023); native-born working-age growth is negative on trend. Post-2025 policy has sharply reversed inflows (preliminary 2026 estimates: net-negative migration for 2025, first since the early 1930s). The US does not escape the demographic channel — it trades asset-deflation for an immigration-politics problem, making working-age trajectory the most volatile political variable in the system.high
- The cohort that built the passive equity bid mechanically reverses it through RMD-driven net selling: retirees net-sell financial assets to fund consumption above pension income, no catalyst required. This is symmetric to the accumulation that drove C-06, and the Peak-65 cohort (largest in US history) concentrates the selling pressure into the late-2020s-2030s.medium
- Ferguson's Law (net interest > defence spending) was breached in FY2024 ($881B vs $874B) for the first time in ~a century, with FY2026 annualising above $1T interest. The historical record of states crossing this line (Bourbon France, Ottoman, Edwardian Britain, Habsburg Spain) shows consistent structural reorganisation, not return to status quo.high
- Turchin's Political Stress Index (a composite of wage stagnation, intra-elite competition, and state fiscal distress) rose from ~7 in 2000 to ~37 in 2012 — comparable to its 1850s pre-Civil-War reading; his 2010 Nature prediction of 2010-2020 instability was empirically validated (PLOS ONE 2020 retrospective: instability rose 'an order of magnitude').medium
- Average G7 sovereign debt-to-GDP crossed 100% in 2020, matching the post-WWII peak — but the post-WWII debt was resolved mainly through demographic-driven nominal-GDP denominator expansion. The current configuration lacks that mechanism (working-age peaks already passed in Japan, Germany, Italy, China, Korea; US growth immigration-dependent), so the prior resolution path is precluded.medium
- China's current-account surplus (~$700-900B in 2024; goods-trade surplus >$1T) is the mechanical result of domestic income distribution under-paying households relative to productivity (household consumption ~38% of GDP, ~20pp below the US). It is not resolvable by trade policy because the consumption-enabling institutions have not been built; the surplus is forced onto global markets.high
- USD share of allocated global FX reserves fell from ~73% (2001) to ~56% (2025), absorbed mostly by gold and a basket of smaller currencies — not a rival (RMB ~2.5% despite a decade of internationalisation policy) — while dollar FX trading volumes hold ~88%. De-dollarisation is a real long-run vector but, on the empirical record, not a near-term investment thesis.high
- Broad geographic EM allocation via headline ETFs reproduces dollar-system exposure in less-liquid wrappers rather than substituting for it: ~70% of EM external sovereign issuance is dollar-denominated, the China weight inside EM (~25-30%) makes 'ex-China EM' a materially different trade, and effective diversification requires institutional-quality + current-account screening at the country level.medium
- The financial system is a coupled-fragility / phase-transition system (Bar-Yam co-movement, Taleb second-derivative fragility, Sornette LPPL) that resolves via discontinuity, not smooth correction; Howell's Global Liquidity Index is the coupling vector, and the 2026-2028 refinancing calendar is the trigger window: ~$40T on Howell/CrossBorder's own estimate (no institutional source publishes it); IIF (>$9T EM, 2026) and S&P (~$11T rated, 4.5 yrs) are smaller because narrower. The mechanical-calendar argument holds at any of the three.medium
- The Fed cannot simultaneously absorb all five 2026-2027 fragilities — private credit, CRE, JGBs (BoJ ~49% of JGBs; repatriation sells USTs), consumer stress, and the AI-capex plateau — because Hunt's velocity collapse (M2V ~1.30 vs 2.20 peak) breaks the easing-to-growth transmission, so the 2020 intervention template (single-vector shock + balance-sheet room + functional velocity) cannot repeat in a multi-vector, velocity-constrained configuration.medium
- Long-duration nominal sovereign bonds are not a reliable hedge in this regime: the two post-1980 tailwinds (declining real rates, demographic bid) are both reversed. As of Sept 2026 the fiscal stress is priced through the term premium, not through repression (30y auction 5.308% on 2026-09-10, highest since 2001; 30y TIPS ~3%; Fed hiked 12-0 on 2026-09-16), so long nominals lose in price; the 1939-1979 financial-repression analogue (positive nominal, deeply negative real returns by policy design) is a possible later-stage response, not the present condition. 30y USTs returned ~-4%/yr nominal and ~-7.5%/yr real over 2020-2025. At 5%+ starting yields the deflation-tail case for some duration is stronger than in 2020-21, so the claim rests on the regime argument, not the starting yield. Japan is the completed test case.high
- The regime structurally displaces capital from the post-1980 default (cap-weighted passive, long-duration bonds, DM consumer-AI mega-cap) into real assets, commodity producers, energy infrastructure, equal-weight, and institutional-quality EM. The spine is Druckenmiller's Q4 2025 13F ($4.49B, 43% turnover; exits NVDA/META/BA/BLK; enters Brazil ~$247M, Alcoa ~$73M, XLF ~$301M, RSP ~$225M, inverse-Treasury 15-20% notional, Barrick).high