The AI Trade Is Repricing the World’s Cost of Capital

“AI can make intelligence cheaper while making capital more expensive.”

The most important market signal this morning is not another record for the S&P 500. It is the real interest rate.

U.S. 30-year inflation-adjusted Treasury yields are approaching 3%, close to an 18-year high. German and British 10-year real yields are around their highest levels in more than a decade. And one of the forces increasingly pushing them higher is the extraordinary quantity of capital required to finance the AI infrastructure build-out. Alphabet, Amazon and Meta have issued almost $220 billion of bonds in 2026, already more than twice their combined issuance during the whole of 2025.

This changes the unit of analysis. The AI trade began by repricing semiconductors; then data centers, power, memory, grids and infrastructure; then corporate balance sheets and private credit. Now it is beginning to reprice something far larger: the opportunity cost of capital itself. The same AI companies generating extraordinary earnings growth are also absorbing an extraordinary share of the world's incremental savings. The result is a paradox — AI can make intelligence cheaper while making capital more expensive.

1. 01 The AI Supercycle Is Beginning to Move the Risk-Free Rate

   Inflation-adjusted borrowing costs have risen sharply across major developed economies. U.S. 30-year real yields are near 3%, while British and German real yields have reached roughly decade highs. Reuters identifies the combination of hyperscaler borrowing, large government deficits and reduced central-bank bond purchases as important contributors. Alphabet, Amazon and Meta alone have already sold almost $220 billion of bonds this year.

   Until now, markets treated AI capital formation largely as a sector phenomenon. That is no longer sufficient. The chain is becoming: accelerated computing demand up → AI infrastructure CapEx up → debt issuance up → demand for savings up → real yields up → WACC up → valuation dispersion up. AI is beginning to influence the denominator used to value the rest of capitalism. signal: AI is becoming large enough to change the price of the money that finances it.

2. 02 From Credit Risk to Relevance Risk

   BNP Paribas credit strategists argue that the extraordinary competition for high-quality capital created by technology-company issuance is beginning to affect credit outside technology. Bloomberg data show CDS spreads on LVMH, Sanofi and BAE Systems have risen more than 10% since year-end. BNP says the effect could ultimately reach sovereign debt.

   A financially sound company does not need to deteriorate operationally to face a higher cost of capital. It merely needs to compete for funding against an asset class offering higher growth, greater strategic relevance and stronger AI exposure. Capital markets can penalize a company relatively even when its absolute fundamentals remain intact. signal: Companies without a credible AI or productivity angle risk a structurally higher relative cost of capital.

3. 03 The Demand Side of the AI Trade Is Still Accelerating

   China's SMIC reported Q2 revenue above $3 billion for the first time, with attributable profit tripling to $479.2 million. Wafer shipments rose 14% sequentially, average selling prices increased 5.7%, and utilization reached 93.7%. South Korea's KOSPI is up about 1.8% this morning and nearly 11% for the week; Japan's Nikkei is around 1.5% higher.

   The capital-market warning is not being triggered by collapsing AI demand — quite the opposite. It is being triggered because demand is so strong that the infrastructure required to satisfy it is consuming extraordinary quantities of capital. The risk is not AI demand failure; it is AI capital-intensity success. signal: The next phase may be less about who builds the frontier model and more about who owns the scarce industrial inputs to deploy it.

4. 04 The Fed Gives Equities Relief While the Bond Market Sends a Different Message

   July U.S. producer prices were unchanged, below expectations for a 0.2% increase, with year-on-year PPI slowing to 4.7% from 5.5%. Combined with mild CPI and weak July payrolls, the report cut the market-implied probability of a September Fed increase to about 35%, from 55% a week ago. The S&P 500 nevertheless closed at a record 7,798.99, up 0.65%, and the Nasdaq added 0.81%.

   The market is pricing two apparently contradictory signals: short-end monetary pressure down, but long-duration real capital scarcity up. The Fed controls an overnight rate; the economy must finance AI factories, grids, defense, energy security and fiscal deficits over decades. Those are different capital markets. signal: The next Great Split may run directly through the yield curve.

5. 05 Europe Enters the Capital Competition at Exactly the Wrong Time

   Euro-area inflation stood at 2.9% in July, and 57 of 69 economists in an August Reuters poll expect the ECB to raise its deposit rate to 2.50% in September. The euro-area economy expanded 0.4% in Q2, with 2026 growth upgraded to 0.8%. Yet Europe must finance energy security, defense and infrastructure while competing with U.S. AI issuance for global savings. Q2 STOXX 600 profits are now expected to grow more than 22%, the strongest pace since 2022.

   Europe must simultaneously fund defense, energy independence, grids, transport, industrial restructuring and Industrial AI — at the precise moment the global cost of long-duration capital is rising. It can ration capital across fragmented national priorities, or raise the productivity of capital by integrating projects and financing pools at continental scale. signal: European capital-market integration is no longer a technical policy project; it is an industrial-security requirement.

Key takeaways

the third AI trade

• The first AI trade was about intelligence; the second about infrastructure; the third is becoming about capital.
• Real yields near multidecade highs are being driven partly by ~$220B of hyperscaler bond issuance competing for global savings.
• Credit is beginning to price a "Relevance Spread" — a financing premium on companies with weaker AI-productivity pathways.
• AI demand is not failing; it is succeeding so intensely that its capital intensity is raising the cost of money.
• The decisive question is which cash flows can outrun the real rate that AI itself is helping to create.