“Intelligence is getting cheaper. Capital is getting more expensive. The winners will own the spread.”
Three numbers define today's regime. China's manufacturing engine is returning to expansion just as America's financial brake tightens — and both are strengthening at once. The result is a Global Velocity Index holding at 108.0 even as the terrain beneath it shifts.
The AI Supercycle is not slowing. Its cost of capital is rising faster than its cost of intelligence is falling. That reframes the entire market: the next phase will not be defined by AI versus non-AI, but by who sits on the right side of three divides.
The Evidence
three numbers, one regime
China's official manufacturing PMI returned to expansion at 50.1, with production at 51.7 and new orders at 50.5. High-tech manufacturing profits had already risen 54.7%, strengthening the evidence that China's recovery is being led disproportionately by the industries supplying the Next Economy.
In the United States, the opposite pressure is emerging through finance. The 10-year Treasury reached 5.293% and the 30-year 5.621%, while job openings declined to 7.08 million and consumer confidence fell to 81.9.
Brent fell 2.5% to $102.59, providing some physical-cost relief — but not enough to offset the repricing of long-duration capital. The GVI remains 108.0 because the productive engine and the financial brake are strengthening simultaneously.
The Second-Order Effects
how the regime propagates
-
At 5.29% on the US 10-year, the bond market increasingly determines which projects survive. The recursive mechanism: AI capex up → financing demand up → yields and spreads up → hurdle rates up → marginal projects fail → capital concentrates in stronger balance sheets.
The first-order effect is higher financing costs; the second-order effect is concentration. The largest cash-generative platforms gain an advantage not because their technology is better, but because their WACC is structurally lower than that of challengers. signal: Lower WACC, not better tech, becomes the decisive edge.
-
Every incremental accelerator now creates derivative demand: compute → memory → optics → networking → power → transformers → cooling → grid. Those layers cannot expand at software speed.
Europe's new EIB/BNP Paribas guarantee programme — potentially catalysing €2.8 billion of grid-equipment investment — illustrates where capital is migrating. The more abundant intelligence becomes, the more valuable its physical bottlenecks become. This is a rotation within AI, not away from it. signal: Physical bottlenecks inherit the rents that compute created.
-
At 5%+ long rates, unused compute, underutilised agents and speculative data-centre capacity become expensive balance-sheet assets. AI adoption is no longer enough. The relevant sequence becomes: AI deployment → utilisation → productivity → margin → cash flow. Companies that cannot complete that chain will find AI increases revenue while destroying economic value. signal: ROIC, not AI revenue growth, separates value from vanity.
-
China's return to expansion coincides with extraordinary strength in high-tech manufacturing. If confirmed, the transmission could run: China industrial acceleration → component demand → commodities → industrial automation → power equipment → global capex cycle — broadening the AI trade toward the global physical-industrial stack. signal: A durable China recovery broadens AI into a global industrial cycle.
-
High rates may improve the Supercycle's quality. Weak projects lose financing, speculative capacity is cancelled, scarcity assets acquire pricing power, and M&A accelerates. The result: higher hurdle rate → capital discipline → consolidation → higher productivity of capital. signal: Tighter money purifies the ecosystem rather than ending it.
The Evolutionary Market Rotation
four phases of the Supercycle
Phase I — AI Narrative
Model capability. GPU scarcity. Multiple expansion. Largely behind us.
Phase II — AI Cash Flow
Revenue conversion. Margins. Free cash flow. Balance-sheet strength. Happening now.
Phase III — AI Bottlenecks
Memory. Networking. Power. Transformers. Cooling. Grid infrastructure. Beginning.
Phase IV — AI Productivity
Industrials, banks, healthcare, transportation, logistics and consumer businesses using AI to lift ROIC, margins and asset productivity. The potentially largest phase still lies ahead.
What Changes for Capital Allocation
two hierarchies
For investors, the relevant hierarchy is becoming: cash flow → scarcity → productivity → optionality. Cash flow protects against higher rates, scarcity creates pricing power, productivity expands margins, optionality creates the next growth curve.
For boards, the hierarchy is equally clear: capacity → utilisation → productivity → cash flow → ROIC > WACC. Anything that stops before cash flow remains an AI investment thesis. Anything that reaches ROIC above WACC becomes competitive advantage.
the rotation in brief
- The AI trade is not ending — it is rotating, from models and compute toward cash flow, bottlenecks and productivity.
- 5%+ rates accelerate concentration, as cash-rich leaders gain structural advantage over externally financed challengers.
- Physical scarcity becomes the next source of excess returns — power, memory, networking, cooling and grids grow more valuable as intelligence scales.
- China may be the trigger for the next industrial leg, broadening AI capital formation into a global industrial and infrastructure cycle.
- ROIC becomes the ultimate AI benchmark — the winners will produce the greatest incremental cash return per dollar of intelligence deployed.
GVI 108.0. China accelerates. America absorbs the higher hurdle rate. The first phase rewarded intelligence; the second rewarded scale; the next will reward capital productivity. Intelligence is getting cheaper, capital is getting more expensive — the winners will own the spread.