AI Expansion Hinges on Access to Capital Markets, Says Fortune

The author argues that the next phase of AI competition between companies and countries will depend largely on access to cheap capital at scale. The author argues that the United States retains an advantage because American capital markets are deeper than those elsewhere. The author argues that China and Europe must close a financial gap to close the technological gap.

AI is turning technology into a capital-intensive industry after half a century in which software could reach billions of users at negligible marginal cost. The latest AI systems require factories, advanced chips, substations and enormous supplies of electricity. Even the world's richest companies are finding the expansion difficult to fund from income. The five largest American hyperscalers are on track to spend $730 billion on capital expenditure in 2026. A shrinking proportion of that spending is coming from the hyperscalers' own balance sheets. The hyperscalers are expected to finance up to a third of that spending with debt.

The author presents Nvidia's deal as one answer to financing AI infrastructure beyond the largest companies. Project finance has long funded infrastructure with high initial costs, repaid through steady returns over many years. The author describes the new effort as an attempt to make computing capacity an asset class that pension funds and insurers are comfortable owning. Nvidia's announcement set a financing target of half a trillion dollars.

Nvidia's announcement prompted concerns about circular financing, in which Nvidia's money returns to Nvidia through chip orders. However, the author argues that the arrangement operates in the opposite way. The author writes that Nvidia and Nvidia's six partners will, for the most part, not lend their own money. The capital comes from third parties, including pension funds, insurers, sovereign wealth funds and household savings. The author argues that the arrangement does not artificially boost demand but restructures supply costs for competitors beyond the hyperscalers. The author argues that the arrangement connects the AI expansion to the wider financial system.

The author warns that the comparison with conventional infrastructure is imperfect. Energy facilities generally last decades, respond to foreseeable demand and generate predictable returns. However, graphics processing units are short-lived assets, AI demand is unpredictable, and returns are too uncertain to project yet. The author argues that both the potential benefits and potential harms of AI remain speculative. The author argues that a successful financing model could reshape the AI market much as venture capital changed the risks of starting a business.

Eurozone households alone hold about 10 trillion euros, approximately $11 trillion, in bank deposits. The author writes that European institutional investors would rather buy U.S. technology bonds than back European startups. The author argues that Europe lacks the financial mechanisms to turn abundant savings into investment. The author argues that capital-market integration would improve European competitiveness more than another round of AI gigafactories.

The author argues that China's state financing is effective at building infrastructure but less effective at matching capacity to demand. Chinese media reported last year that as much as 80% of newly built computing capacity sat unused after a policy-driven construction spree. The Stanford HAI AI Index 2026 estimates that private AI investment in the United States last year exceeded investment in China by more than 23 times.

The author argues that no other country matches the United States in directing pensioners' and households' savings into risky, long-lived assets at scale. The author writes that American bond markets, private credit funds and asset managers are already seeking assets of this kind. The author argues that financial structures will help determine AI winners through the effective allocation of capital, not simply through the amount available. The author concludes that the next phase of AI depends on financing expansion beyond a small group of companies. The author argues that Nvidia's competitors will need equally inventive financing mechanisms.