Foreign capital backs North American AI expansion

Foreign investors are taking a growing role in financing North America’s artificial intelligence infrastructure build-out as technology companies tap global pools of capital to meet spending approaching $1 trillion, according to JPMorgan Asset Management.

Charles Wu, head of Asia-Pacific alternatives institutional client strategy at JPMorgan Asset Management, told the SuperReturn Asia conference in Singapore on Wednesday that banks and institutional investors from Asia, the Gulf and other regions were increasingly supplying capital for computing assets concentrated in North America.

“Compute training assets are still largely in North America,” Wu said. “On the banking side and also on the asset management side, the capital providing the funding for these assets is largely becoming more global.”

The widening investor base is helping finance an extraordinary expansion in data centres, computing capacity and associated power infrastructure. Wu said US hyperscalers continue to fund much of their artificial intelligence capital expenditure from operating cash flow, but are also drawing about $250 billion from bond markets and roughly the same amount from bank loans.

That shift is increasing scrutiny of how much debt is ultimately tied to a small group of technology companies. Wu said JPMorgan Asset Management was watching lease obligations, concentration risk and financing raised through structures that may sit outside corporate balance sheets.

“The reality is a lot of these funding securities, whether you’re talking about investment-grade bonds, 144a, private placements, even private credit structured deals — a lot of them seem to all triangulate to basically five key hyperscaler parties,” Wu said. Those companies can appear in transactions as guarantors, tenants, funding sources or customers.

JPMorgan Asset Management has separately estimated that the full data-centre build-out could cost about $5 trillion through 2030 over the coming years, with roughly $2 trillion potentially financed through investment-grade credit markets. Its strategists said hyperscaler bond issuance reached $194 billion in the first half of 2026 and could finish the year at $279 billion.

The scale of borrowing has intensified debate over whether conventional credit indices capture the full exposure. A growing share of data-centre financing uses private placements, project vehicles, bank construction loans and other structures outside standard investment-grade corporate bond benchmarks. That can make the aggregate leverage associated with artificial intelligence infrastructure harder for investors to measure.

The Bank of England added to those concerns on Wednesday, saying growing AI-related debt had broadened financial-market exposure to the sector. Its Financial Policy Committee estimated AI-related debt issuance at about $450 billion by early September, more than double the total for 2025, and warned that disappointment over future AI earnings could amplify a market correction.

The financing picture is not uniformly fragile. JPMorgan Asset Management says major hyperscalers still have comparatively strong balance sheets, with average net debt to earnings before interest, tax, depreciation and amortisation well below typical investment-grade issuers. Investors are also differentiating between borrowers rather than treating all AI-linked debt alike.

At SuperReturn, Jean-Christophe Aubert, senior director of infrastructure investments at PSP Investments, said his organisation was assessing concentration and the eventual ability to realise large infrastructure assets. The volume of capital required made those questions particularly important, he said.

The geographic pattern could also change as artificial intelligence moves from training models towards inference, or their everyday use. Panel members said computing for inference benefits from proximity to users, creating a case for more infrastructure investment across Asia even while North America remains the dominant centre for training capacity.

Mohsin Pirzada, head of funds at Qatar Investment Authority, said Asia was important on both the demand and supply sides. He pointed to decentralisation driven by end-user demand, data sovereignty and security requirements as factors likely to influence where new computing infrastructure is built.