The Next Economy Research Centre

“The technology sets the opportunity. The capital structure sets the risk. Cash flow determines who wins.”

The AI Supercycle has entered a more consequential phase: technology demand continues to accelerate, but the financial system is becoming increasingly exposed to the cost, concentration and duration of the capital required to sustain it.

Three developments define today's assessment. Europe's semiconductor champion has delivered exceptional results. CoreWeave is pulling AI infrastructure deeper into institutional credit. And the IMF has elevated AI-related financial concentration into a global stability concern.

The decisive question is no longer whether AI will transform the economy. It is whether the capital structure financing that transformation can withstand higher interest rates, technological obsolescence and delays in the productivity dividend.

ASML's third-quarter revenue rose from €7.52 billion a year earlier to €11.66 billion, with installed-base management generating €3.29 billion. Management now guides to €44–45 billion for the year and warns that planned 2027–2028 capacity may be insufficient for demand. Some customers are already discussing orders for 2029 and 2030.

Semiconductor scarcity is moving upstream — from GPUs and memory into the capital equipment required to make future chips. A shortage of lithography tools can limit the industry's ability to resolve downstream chip scarcity, prolonging exceptional returns for suppliers. But strong demand does not eliminate cyclical risk: customers could revise orders if utilization or AI profitability disappoints. For Europe, ASML represents a strategic position in the global AI value chain that is difficult to replicate.

CoreWeave is pursuing greater participation from insurance investors as it finances its build-out. The Financial Times reports it has raised more than $35 billion this year against roughly $52 billion of obligations, with an $8.5 billion GPU-and-Meta-backed financing receiving investment-grade treatment. Its target of 8 GW by 2030 is an ambition, not operating capacity; free cash flow is expected to remain negative through 2028.

AI infrastructure is increasingly presented to lenders as a financeable asset class. But the analogy has limits: a power plant may run for decades, while an accelerator can become technologically uncompetitive within a few years. Risk arises when financing maturities, equipment economics and customer contracts do not align. Monitor debt-service coverage, residual-value guarantees, customer concentration and refinancing exposure — not merely backlog.

The United States is investing heavily in computing, but much of the equipment is imported. Reuters estimates ~70% of AI-related capital expenditure goes toward imported servers and semiconductors; the August trade deficit reached $105.6 billion. AI investment can lift equity valuations and corporate CapEx without an equivalent immediate increase in domestic value added.

This explains why semiconductor exporters can post extraordinary earnings before deploying economies realize broad productivity gains. The next competitive advantage will depend on how much of the AI cycle remains within a national economy — converting imported hardware into domestic software, services and productivity. No new independently verified economy-wide industrial-AI productivity inflection today.

The IMF's October assessment identifies two competing forces: a negative energy-supply shock from Middle East conflict, raising prices and weakening activity; and a positive AI investment-demand shock, supporting investment and trade while potentially adding to inflation. This reinforces the two-stage framework — Stage I infrastructure inflation (compute, energy and capital demand rising), then Stage II industrial-AI productivity (deployment and automation lowering unit costs).

The critical uncertainty is the transition between them: AI can raise inflationary pressure through infrastructure spending before generating enough productivity to offset it. Energy prices and long-term yields must now be built into AI infrastructure valuation models.

The IMF projects Eurozone growth of ~1.1% in 2026 against global growth of 3.1%. Europe cannot close that gap through public expenditure alone; it must improve the productivity of existing industrial assets while attracting private investment. The October 15–16 European Council will test its competitiveness and financing ambitions. No final new EU AI financing package verified today.

Three priorities: protect scarce industrial capabilities — equipment, photonics, industrial software, power electronics; accelerate conversion, judging AI Factories by productive workloads and domestic value added, not announced GPU counts; and strengthen capital-market integration for long-duration investment. The governing metric: domestic productive value per euro of AI capital.

Five key takeaways

What it means for capital allocators

• For investors — distinguish scarcity profits from capital-dependent growth; ASML's reported profitability and CoreWeave's financing ambitions are fundamentally different exposures.
• For boards — prioritize cash-flow timing; value depends on when a return becomes available to finance the next cycle.
• For central banks — monitor the capital-demand shock; AI can support activity while raising inflationary pressure and financing requirements.
• For Europe — build around existing strengths in semiconductor equipment, automation, power systems and engineering software.
• For policymakers — measure domestic economic conversion; investment announcements are not evidence of productivity or competitiveness.

The countries manufacturing the infrastructure earn profits today; the countries financing it assume risk today; the countries deploying intelligence productively may capture the largest long-term gains — but those outcomes will not necessarily accrue to the same companies, investors or countries. For markets, the question is no longer how much capital AI can attract, but which parts of the AI economy can compound that capital into sustainable returns.