The Next Economy Research Centre

“Computing defines the frontier. Electricity determines access. Capital determines who can sustain the journey.”

The AI Supercycle is entering a new phase of financial selection. Demand for computing infrastructure keeps expanding, but the cost of financing it is rising while reliable electricity becomes a decisive competitive advantage.

For much of the cycle the central scarcity was advanced semiconductors; it has since extended to grid connections, execution and long-duration financing. Those constraints are now beginning to interact — and the market is learning to distinguish the promise of intelligence from the financial capacity to deliver it.

Investor caution has extended beyond individual AI companies into broader corporate credit. The 129-basis-point figure is a broad index observation, not evidence of an AI-specific default cycle. The deeper risk is structural: AI infrastructure demands very large upfront capital, its most valuable equipment can depreciate faster than conventional infrastructure, and returns depend on utilization, demand and energy costs — a mismatch between financing structures and asset life.

The transmission is conditional: higher spreads → higher project financing costs → greater selectivity → potential deployment delays. Operators with strong cash flow and long-duration financing stay insulated; leveraged developers reliant on repeated refinancing are exposed. Expect greater concentration of ownership, not a generalized closure of AI credit markets.

Developers are turning to on-site generation because grid connection can require three to six years in parts of the US. Rystad sees fuel-cell manufacturing capacity rising from about 1.8 GW to roughly 4 GW by 2030, with around 9 GW of reported orders and frameworks and about 40% of US data-centre capacity potentially considering dedicated power.

Bloom Energy's Q3 confirms the demand: revenue $1.236B (+138.1% YoY) and GAAP operating income of $225.4M — but approximate free cash flow of −$180.6M, against 2026 guidance of $4.4–4.6B. The suppliers of alternatives to constrained grids are themselves becoming capital-intensive. On-site generation trades grid dependence for exposure to gas prices, fuel logistics and dedicated infrastructure — the grid arbitrage. Electricity readiness may increasingly command an economic premium.

Siemens and Nvidia report AI agents exploring up to 50× more chip-design alternatives, electrothermal workflows configured in under five minutes rather than days, and prefabricated power infrastructure with Delta cutting some deployment timelines by up to 50% — company-reported claims, not independently verified across the industry. AI is increasingly applied to building the infrastructure that supports AI, a recursive productivity loop.

American Express announced a restructuring of about 3,500 employees (5% of its workforce), citing efficiency — but workforce cuts alone do not prove higher total-factor productivity. The economy-wide industrial AI productivity inflection remains unconfirmed.

ByteDance, Alibaba and Tencent reportedly placed orders for more than $35B of Nvidia RTX PRO 5500 workstation GPUs; Nvidia warns it may not meet the volumes, and orders are not deliveries. Europe's strengths lie largely outside frontier foundries — lithography, electrical infrastructure, industrial engineering and software.

Moody's affirmed France's Aa3 on October 23, but a stable rating does not guarantee stable borrowing costs: the French-German spread reached about 178 bps on October 22 and the 10-year briefly exceeded 5.14%. The EU AI Gigafactories programme still targets up to seven facilities and more than €30B (up to €10B public, at least €20B private); the tender closes November 12.

Key takeaways

For investors, boards and policymakers

• The financial cost of AI infrastructure is becoming as important as its technological capability.
• Electricity availability is creating a new hierarchy of assets — operational capacity commands a premium over announced capacity.
• Alternative power systems create opportunities but new fuel-price, financing and operating exposures.
• Europe's industrial capabilities — semiconductor equipment, engineering software, power electronics and automation — offer a credible route to competitiveness.
• Financial credibility and capital productivity must become central to Europe 2030.

The next phase of the Great Split will distinguish economies that can finance intelligence from those that can power it — and ultimately from those capable of turning it into productive wealth.