MACRO INSIGHT 01
The Era of 5% U.S. 10-Year Treasury Yields — Can the AI Boom Withstand High Rates?
How Treasury yields, credit terms and operating start dates are reshaping AI infrastructure investment
Key Conclusion
U.S. long-term yields above 5% are more likely to differentiate AI investment by funding capacity and contract structure than to stop it universally. Businesses with established operating cash, committed funding, creditworthy customers and credible operating schedules have greater relative endurance. Leverage-, renewal- and residual-value-dependent operators may need to redesign financing even while demand grows. The potential for AI growth must be distinguished from an individual project’s ability to recover invested capital.
Research Summary
On 29 September 2026, the U.S. Treasury’s 10-year nominal par yield was 5.26% and its real counterpart was 2.91%. Of the 51bp nominal increase since 31 August, 47bp corresponds arithmetically to the real-yield move; this is not a causal attribution. The report examines why high rates and AI investment can coexist when current demand for equipment, power and capital arrives before future productivity benefits. Matched-quarter cash flows from Microsoft and Meta, alongside CoreWeave’s earnings and distinct financing facilities, demonstrate why large revenue, adjusted EBITDA or long debt maturity alone does not guarantee capital recovery. Illustrative project models then show how higher loan rates raise required utilization and how delayed cash receipts reduce net present value. The base case favors selective investment continuity, differentiated by funding, customer contracts and execution, rather than a universal end to the AI boom. For Korean substrate, MLCC and cooling suppliers, customer funding, firm orders, acceptance and collection are more informative than aggregate U.S. AI CAPEX alone. The key monitoring variables are credit terms, paid use and renewals, power availability, operating starts and cash recovery—not Treasury yields in isolation.
Key Points
- High rates and AI investment can coexist: current capital, power and labor demand may precede the productivity payoff.
- Treasury yields are not actual borrowing rates. Benchmarks, spreads, maturity, hedges and customer contracts must be assessed separately.
- Microsoft and Meta’s matched-quarter cash flows illustrate different investment buffers, not standalone AI profitability.
- In the illustrative model, raising the loan rate from 8% to 10% lifts utilization needed for 1.30× DSCR from 87.1% to 92.1%.
- AI investment outcomes depend on the continuity of funding, power and operation, customer acceptance and cash recovery.