“The AI supercycle is not weakening. Its financial architecture is being repriced.”
The most consequential development overnight is not the technology selloff. It is the widening divergence between AI's ability to generate revenues and the financial system's willingness to finance its expansion.
The AI economy is separating into businesses increasingly capable of monetizing intelligence and businesses whose expansion requires progressively more expensive capital. Asian semiconductor equities are correcting sharply and credit protection costs for major AI borrowers are widening, yet advanced chip makers report exceptional demand. The AI supercycle is not necessarily weakening — its financial architecture is being repriced.
This is the next evolution of The Great Split. The first divide concerned access to intelligence; the second, ownership of accelerated-computing infrastructure. The third will concern the difference between the returns intelligence generates and the cost of capital required to produce it. That difference — the AI Return Spread — could become the defining investment variable of the next market cycle.
The Five Overnight Signals
What the morning tape reveals
-
September U.S. consumer prices rose 0.6% on the month, taking the annual rate to 3.7%; core inflation rose 0.3%, above consensus. Technology shares fell sharply — CoreWeave about 16%, Oracle 9%, Nebius and Dell 11% each — before defensive buying pushed the 10-year yield back toward 5.11%. Lower sovereign yields do not mean cheaper financing: if corporate spreads widen faster than yields fall, effective costs keep rising. signal: The next divide is between firms that can finance acceleration internally and those dependent on favorable capital markets.
-
Anthropic told prospective investors that Q3 revenue reached $16.8 billion, up from $11.6 billion in Q2 and $4.7 billion in Q1, with an adjusted gross margin near 85%, operating margin above 11% and positive free cash flow. It is seeking a valuation exceeding $2 trillion against long-term infrastructure commitments of at least $518 billion over ten years. The evidence challenges the view that AI is predominantly an infrastructure-spending phenomenon — but revenue growth and financial productivity are not equivalent. signal: The next phase is won not by producing more intelligence, but by monetizing its economic consequences.
-
Credit protection costs jumped: Oracle's five-year CDS widened 26bps to 288bps, SoftBank reached 424bps, and CoreWeave exceeded 1,000bps. CoreWeave carries roughly $48 billion in bonds and loans; a 2032 bond traded near 77 cents on the dollar, yielding around 15%. The market is distinguishing self-financed platforms, contract-backed operators and capital-dependent developers — and the third faces a feedback loop where higher risk raises costs, lower returns constrain financing, and delays weaken growth. signal: The AI credit correction could accelerate industry concentration even as demand expands.
-
ByteDance is reportedly approaching a deal with MediaTek for data-center semiconductors, as Chinese firms diversify hardware. Meanwhile TSMC is expected to report roughly 65% year-on-year profit growth today, and ASML customers discuss orders toward 2029–2030. Two partially competing architectures are emerging — a U.S.-led system built on deep capital markets, and a China-centered one emphasizing substitution and domestic supply. signal: The Great Split is becoming a contest between competing systems of capital formation.
-
EU leaders convene in Brussels today on energy prices, Chinese competition and the 2028–2034 budget. The bloc's goods trade deficit with China reached roughly €360 billion in 2025. Europe combines advanced industry with fragmented capital markets and elevated energy costs — a structural constraint on investment velocity. The agenda must connect cheaper reliable electricity, long-term private capital and accelerated Industrial AI adoption. signal: Europe cannot finance a new industrial revolution through a fragmented capital system while competing against integrated ecosystems.
The Hidden Pattern
The AI Return Spread as the new divide
AI applications generate increasingly substantial revenues and advanced semiconductor suppliers retain exceptional demand — but the cost of financing infrastructure has become sharply differentiated.
This suggests a proposed measure, not a published index: AI Return Spread = incremental cash return on AI investment − risk-adjusted cost of capital. It is positive when returns exceed the capital hurdle, negative when financing and operating costs exceed the value produced. Crucially, it can differ dramatically within the same ecosystem: a semiconductor manufacturer, a hyperscaler, a leveraged developer and a capital-efficient adopter can all participate while their outcomes diverge.
The most coherent reading is capital concentration: demand remains strong, but credit is becoming selective — strong balance sheets gain, leveraged developers face pressure. The downside is a financial feedback shock in which inflation, spreads and project delays reinforce one another. The positive alternative requires proof that productivity benefits are spreading beyond infrastructure suppliers.
How to read the next regime
- The AI economy is separating into revenue leaders and financing-dependent operators.
- The cost of capital is becoming a decisive competitive variable; cash-rich platforms preserve flexibility.
- Technological acceleration can coexist with financial contraction — the correction does not prove AI demand is weakening.
- The next rotation may favor capital-efficient AI adoption as computing costs fall and enterprise monetization strengthens.
- Europe's competitiveness depends on uniting financial integration, energy productivity and Industrial AI into one agenda.
In the next economy, competitive advantage will belong to those who expand the return on intelligence faster than the cost of financing it.