The 5% AI Economy

“In the 5% AI economy, access is valuable. Contracted access is power.”

AI is accelerating the physical economy — and raising the price of the capital required to finance it. Europe approaches the open with one number dominating everything else: 5.25%.

The U.S. 10-year Treasury yield is back near 5.25% after touching 5.34% on Thursday, its highest level since 2002, capping the steepest quarterly rise in yields in 32 years. Nasdaq futures are nevertheless up roughly 0.4%, while European futures point marginally lower.

This is becoming the central paradox of the AI supercycle. The investment boom is strong enough to lift manufacturing across Europe and Asia, push memory capacity toward full utilization, lock technology companies into 20-year energy contracts and attract sovereign capital deeper into the AI stack. But that same investment intensity is helping sustain nominal growth, energy demand and capital demand at precisely the moment bond markets are demanding higher returns.

Today's U.S. payroll report becomes the next test: consensus expects +90,000 jobs, unemployment at 4.1%, with wage growth critical to the rate path. Markets now price only about a 25% probability of an October Fed hike, but still expect another increase before year-end. The issue is no longer whether AI growth is real. The fundamental question is whether AI-driven productivity and cash flow can rise faster than the cost of capital the AI supercycle is itself helping to push higher.

1. 01 Bonds — The Hurdle Rate Is Becoming Structural

   The U.S. 10-year touched 5.34% Thursday before buyers returned, trading around 5.25% in Asia this morning. Thursday's initial claims were only 197,000, ISM manufacturing remained expansionary at 54.5 with a sharp rise in input prices, and Fed officials John Williams and Philip Jefferson pushed back against an immediate October hike, cutting the market probability to roughly 25%.

   This is more consequential than the next 25 basis points. The bond market is increasingly distinguishing between the policy rate and the structural price of capital. The transmission chain is becoming: AI capex → stronger industrial demand → more energy, equipment and capital → higher nominal growth → higher term premium → higher hurdle rate. The Fed may pause in October; the cost of long-duration capital does not necessarily follow. AI may eventually be disinflationary — but its build-out is capital-intensive first. signal: The policy rate can pause while the structural price of capital keeps rising.

2. 02 Accelerated Computing — The Bottleneck Moves From Chips To The Industrial System

   September manufacturing data show AI demand spreading beyond technology equities into the physical economy. Eurozone manufacturing PMI reached 52.9, its highest since May 2022; Taiwan's PMI rose to 56.7; South Korean export demand grew at its fastest pace in 15½ years. Micron has disclosed $32 billion of customer financial commitments, around $150 billion of remaining performance obligations, and agreements covering most of its 2027 HBM output — calling memory the chief constraint in AI infrastructure.

   The bottleneck is migrating: GPU → HBM → network → power → capital, and every migration redistributes economic rents. The first AI phase rewarded ownership of compute; the next rewards companies that can pre-contract the next constraint before everyone recognizes it as scarce. Micron's customer deposits show customers now financing capacity themselves to secure access — a structural shift in bargaining power. The Great Split is becoming capacity with contractual scarcity versus capacity dependent on spot demand. signal: Capacity with contractual scarcity deserves a different multiple.

3. 03 Power + Sovereign Capital — Capital Is Moving Upstream

   Amazon and Constellation have signed a 20-year agreement covering 690 MW from Maryland's Calvert Cliffs nuclear plant, supporting more than $3 billion of investment including a 190 MW uprate expected between 2030 and 2032, and helping relicense the plant for another 20 years. Meanwhile, Singapore's roughly $400 billion Temasek remains strongly positive on AI despite the bond selloff, planning to raise AI-related exposure to as much as 15% of its portfolio over five years, from about 6% today. Hyperscaler AI capex is expected to exceed $1 trillion by 2027.

   Technology companies once bought electricity; they increasingly contract the assets that produce it. Sovereign funds once bought technology stocks; they increasingly fund the infrastructure beneath them. The emerging AI capital stack — sovereign capital, hyperscaler balance sheets, private capital, energy assets, contracted compute — is the next stage of capital concentration. Capital is moving upstream toward the physical inputs AI cannot function without, where duration, scarcity and contractual cash flow meet. signal: Capital is no longer chasing AI companies — it is moving upstream to the inputs.

4. 04 Geopolitics — Managed Interdependence Becomes Managed Chokepoints

   The Trump administration has slowed export licensing for aircraft parts bound for China and is considering mechanisms that could restrict components such as landing gear, seeking leverage as Washington pushes Beijing to loosen its grip on critical rare-earth supplies used in autos, semiconductors and aerospace. This comes only days after the two sides agreed to tariff reductions on roughly $60 billion of bilateral goods following the Trump–Xi summit.

   The relationship is moving neither toward decoupling nor détente, but toward selective connectivity plus selective coercion. Each side is identifying the other's hardest-to-replace dependencies — China: rare earths and critical materials; the United States: aviation, advanced technology and market access. The strategic asset is no longer domestic production but control of a chokepoint the other side cannot replace quickly. Managed interdependence does not eliminate geopolitical risk — it monetizes dependency. signal: Managed interdependence monetizes dependency; single-source exposure now carries a valuation penalty.

5. 05 Europe — The Bond Market Is Becoming The Competitiveness Market

   France unveiled a 2027 budget containing €54 billion of savings, including €43 billion of new measures. Yet its 10-year borrowing cost climbed to 4.96%, its highest since July 2002, and France–Germany spreads widened beyond 140 basis points. French debt has reached about 119% of GDP, and Paris expects to issue a record €340 billion of debt next year. Eurozone inflation is expected to rise to 3.6% today after September readings of 5.0% in Spain, 4.1% in Italy, 3.4% in France and 3.3% in Germany — with markets already pricing four additional ECB hikes over the coming year.

   Europe now faces a three-way squeeze: higher energy cost, higher capital cost, higher fiscal cost. It must finance defense, grids, energy resilience, industrial policy and accelerated computing at the same time its sovereign cost of capital is rising. The competition is no longer simply Europe versus the United States on technology — it is Europe versus its own financing structure. The winners will be those that convert sovereign borrowing into productive assets fast enough to generate growth before interest expense consumes the fiscal space. signal: In a 5% world, delay has a compounding cost.

Key Takeaways

The strategic agenda for today

• Yesterday's question was whether yields could stay above 5%; today's is what a 5% world does to the economics of a capital-hungry AI supercycle.
• It forces memory customers to prepay, hyperscalers to contract power decades ahead, and sovereign capital toward infrastructure.
• It turns trade dependencies into geopolitical leverage and forces governments to compete for capital alongside the companies they subsidize.
• This does not end the AI supercycle — it changes who captures it.
• The first AI age rewarded scale; the next will reward control of the contract.