“Capital is moving away from the promise of intelligence and toward the demonstrated productivity of intelligence.”
The global financial system is confronting its first major collision between the extraordinary growth of Artificial Intelligence and the financial capacity required to sustain it. Record semiconductor profits, accelerating AI revenues, deteriorating infrastructure credit and persistent inflation are no longer separate developments — together they reveal a new capital regime in which demand for intelligence keeps expanding while the market withdraws its unconditional support for financing it.
The AI Supercycle is not experiencing a collapse in technological demand. It is experiencing a reassessment of who can finance that demand profitably. That distinction changes the investment thesis: capital is beginning to discriminate between chipmakers capturing extraordinary margins, infrastructure operators borrowing heavily to deploy capacity, developers monetizing intelligence, and industrial companies seeking measurable productivity. The next rotation depends on it.
The Five Signals
October 15's diagnosis
-
The U.S. PPI rose 1.3% month on month and 6.3% year on year. Gasoline climbed 20.3% in September; diesel rose 21% on the month and 115.2% year on year. Retail sales gained 0.8%, but much reflected higher fuel spending rather than stronger consumption — reinforcing a K-shaped regime. Yet the 10-year Treasury yield fell toward 5.1% during the equity selloff. signal: A bond rally during an equity selloff may signal financial-system risk, not recovery.
-
CoreWeave carries roughly $50 billion in debt. Its 2031 bonds traded near 79 cents on the dollar, yielding close to 16%, with default insurance near 1,000 basis points — a model-implied five-year default probability approaching 63%. Proposed listings from AirTrunk, Firmus and SB Energy were paused, even as Asian data-center M&A reached ~$21.3 billion in 2026, more than double all of 2025. signal: The constraint is no longer access to intelligence — it is access to capital cheap enough to profit from it.
-
TSMC reported Q3 net profit of ~$23.6 billion, up 69%, on ~$46.7 billion revenue, guiding to ~44% full-year growth and $60–64 billion of 2026 capex. Yet the KOSPI fell 7.57%, Samsung ~9.5%, SK Hynix ~11.1%, the Nikkei nearly 5%, and the Philadelphia Semiconductor Index has erased more than $1.6 trillion over six sessions. signal: The market is separating the value of technological leadership from the price it will pay for it.
-
Healthcare gained ~2.6% and staples ~1% as technology stayed under pressure and Treasury yields fell. Money is moving toward perceived resilience even as inflation stays elevated. signal: Markets now distinguish inflation risk — a higher cost of capital — from credit risk, the inability to raise it at all.
-
The STOXX 600 closed +0.38% at 640.16, but ASML fell ~4.2%, Infineon ~5.2%, Soitec ~7.7%. Global smartphone shipments dropped 22% year on year to ~255.3 million units as memory costs rose ~350%, partly on AI demand. signal: AI investment now competes with existing industries for components, power and capital — redistributing profitability.
Where Global Capital Is Moving
formation, deployment, productivity
The emerging regime separates three things: whether a company can raise funding, whether funding becomes operational capacity, and whether capacity earns returns above its risk-adjusted cost. U.S. hyperscalers are expected to spend more than $700 billion on AI infrastructure in 2026; Gartner estimates total AI spending could approach $1 trillion this year. But financing architecture is becoming more demanding, with equipment depreciating faster than the debt funding it.
The other side is monetization. Anthropic reported ~$16.8 billion in Q3 revenue — up nearly 50% sequentially — at an ~11% adjusted operating margin and roughly $1.70 of revenue per dollar of compute (company-reported figures). The next rotation may occur within AI rather than away from it: the question is no longer whether intelligence creates demand, but who captures the profits.
Physical infrastructure tells the same story. Europe's ~$40 billion of battery investment profits from volatility — UK daily power spreads widened ~80% to £123/MWh — yet more capacity will eventually narrow the premium. Alstom's £800 million ($1.06 billion), 15-year maintenance extension shows value migrating from building assets to running them intelligently over their lifecycle.
the new hierarchy
- Distinguish technological demand from investment profitability: a fast-growing market can still deliver poor returns for capital-intensive players.
- Watch credit markets alongside equities — yields, CDS and delayed issuance warn earlier than earnings.
- Prioritize companies that convert intelligence into cash flow: strong balance sheets, durable margins, measurable productivity.
- Subject AI expenditure to capital-productivity discipline, incorporating financing cost, depreciation and utilization.
- Europe should connect AI directly to its industrial strengths rather than replicate hyperscaler capital intensity.
The AI Supercycle is entering its Great Financial Selection: capital will increasingly move away from the promise of intelligence and toward the demonstrated productivity of intelligence — and lasting economic power will belong to those who can both finance intelligence and make it productive.