Inflation Exposes the Cost of the AI Supercycle

“The future will not belong to those who spend the most on intelligence, but to those who turn it into the highest sustainable return on capital.”

Global financial markets are confronting a decisive contradiction: demand for artificial intelligence infrastructure continues to accelerate, but the financial conditions required to sustain that expansion are deteriorating. Today's inflation surprise, sharp repricing of AI-related equities and extraordinary semiconductor order visibility reveal a two-speed investment regime — technological expansion without guaranteed financial returns.

The critical question is no longer whether the AI Supercycle will continue. It is which companies and economies can convert extraordinary investment into productive cash flow before financing costs erode their competitive advantage.

ASML's customers are discussing equipment requirements for 2029 and 2030, yet AI infrastructure operators are experiencing sharp equity declines as investors reassess financing, depreciation and future earnings. This is not the end of the AI investment cycle. It is the beginning of a more unforgiving phase of capital allocation.

1. 01 The Inflation Shock

   Headline CPI rose 0.6% month on month and 3.7% year on year; core inflation rose 0.3% monthly and accelerated to 2.6% annually. Energy climbed 5.2% and gasoline 9.1%. Following the release, the U.S. 10-year Treasury yield briefly reached roughly 5.32%, near levels last seen in 2002. Higher expected rates make long-duration financing more expensive and reduce the present value of future cash flows — acutely so for capital-heavy data centers. signal: The new capital hurdle is becoming structural, not temporary.

2. 02 Wall Street Splits Into Two Markets

   At about 12:30 p.m. ET the S&P 500 traded at 7,719.86 (−1.10%), the Nasdaq at 26,751.71 (−1.66%) and the Dow at 51,516.03 (−0.30%). Beneath the indexes, Oracle fell ~7.8%, CoreWeave and Nebius more than 11%, Dell ~11.5% and HPE 7.8%, while healthcare and staples advanced. Companies with established demand and margins are being separated from those whose valuations depend on future utilization and continued access to financing. signal: The market is repricing the financial structure through which AI is delivered.

3. 03 ASML — The Physical Cycle Stays Strong

   ASML reported €11.7B revenue, €4.96B net income and a 57.5% gross margin, raising full-year sales guidance to €44–45B and guiding Q4 to €14.2–15.2B. Customers are discussing requirements into 2029 and 2030, and planned 2027–2028 capacity may prove insufficient. Supplier profitability does not guarantee downstream profitability. signal: Industrial demand remains powerful; its financial valuation is becoming selective.

4. 04 Capital Moves Toward Distribution

   Bank of America reported ~$30B revenue but net income of ~$6.9B after a $1.2B penalty. Morgan Stanley posted $20.3B revenue, ~$4.9B net income, 22.6% return on tangible common equity and $232B of net new wealth assets. BlackRock drew $205B of net inflows, lifting AUM to ~$15.49T, and has floated treating computing capacity as a financial asset. signal: Securitization redistributes AI risk through the system; it does not eliminate it.

5. 05 Infrastructure Is Being Repriced

   Saudi Aramco's Dhahran headquarters sustained damage in an attack attributed to the Houthis; Brent traded near $106. Eutelsat surged ~23% intraday after Ukraine indicated SpaceX had restricted certain Starlink services — though OneWeb cannot immediately replace it. signal: Resilience is becoming a factor of production — and a component of valuation.

The first economy is driven by the extraordinary investment needs of accelerated computing — data centers, semiconductors, electricity and long-duration financing. The second is driven by deploying AI across existing businesses to improve productivity, integration and asset utilization. The first requires substantial new capital; the second can generate attractive returns from capital already deployed.

Morgan Stanley estimates Alphabet, Amazon, Microsoft and Meta could spend ~$700B in capital expenditure in 2026, with AI-related debt issuance approaching $570B — forecasts, not realized totals. On a hypothetical $10B floating-rate project, a sustained 100-basis-point rise adds roughly $100M in annual interest. The next infrastructure winners will combine contracted demand, reliable power, realistic depreciation and financing able to survive higher rates.

Evidence of productive AI is emerging: Citigroup reports developing some products up to 40% faster and cutting certain vendor costs ~30%, while JPMorgan cautions that heavy investment need not yield immediate quarterly benefits. The decisive metric becomes incremental return on invested capital.

For Europe, ASML and Eutelsat show world-leading and strategically indispensable capabilities alongside structural dependencies. Competitiveness requires affordable power, deep capital markets and banking integration — turning engineering, manufacturing, energy and mobility strengths into sustainable returns.

Key takeaways

investors, boards, policymakers

• Investors: favor demonstrable cash conversion, durable pricing power and limited dependence on repeated external financing; judge by free-cash-flow yield, incremental ROIC and rate sensitivity.
• Boards: subject AI capital expenditure to explicit productivity targets, and integrate energy resilience, cybersecurity and financing duration into allocation.
• Policymakers: prioritize affordable energy, efficient capital markets and recognition that technological sovereignty requires commercially sustainable capabilities, not only public commitments.

The next Great Split will not be between those who invest in AI and those who do not — it will be between those who can finance intelligence, those who can make it productive, and those whose capital structures cannot sustain either.