(Sept 30): Foreign investors are increasingly funding the growth of artificial intelligence computing across North America, JPMorgan Asset Management’s Charles Wu told an audience at the SuperReturn conference in Singapore on Wednesday.

Foreign banks and institutional investors including pensions, sovereign wealth funds and insurance companies from regions including Asia and the Gulf are playing a major part in funding the expansion of a sector where capital expenditure stands at almost US$1 trillion (RM4.1 trillion), said Wu, the bank’s head of Asia-Pacific alternatives institutional client strategy.

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

How global investors approach the AI trade — including the mammoth funding needs for the world’s data centres — has been a key topic of conversation at the SuperReturn event this week. Most panelists have struck a positive tone, but there are also detectable notes of caution.

“We’re being very careful with the way we assess the risk,” said Jean-Christophe Aubert, a senior director of infrastructure investments at PSP Investments. “Concentration is a big point, but also for us, given the amount of capital that we would have to put to work, is trying to understand how can we ultimately realise those assets going forward.”

While US hyperscalers are funding much of the AI capex need through operating cash flows, they are also getting about US$250 billion from the bond market and roughly the same amount from bank loans, said Wu. He warned that his firm was keeping an eye on concerns including lease commitments and concentration risk, as well as the risk that some of the financing is being raised off balance sheet.

“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,” he said. “Either as a guarantor, as a tenant, as a funding source, or as a customer.”

While the US remains the dominant market, several of the panelists agreed that Asia is likely to be close behind. The nature of inference — the actual use of AI models — benefits from being close to end users, which will help drive huge demand for infrastructure in the region.

“Asia is equally important as it plays an interesting role in both the demand side of the equation and the supply side through the physical stuff,” said Mohsin Pirzada, head of funds at Qatar Investment Authority. “What we’re starting to see is a decentralisation, and that’s being driven by end consumer demand” as well as data sovereignty and security, he said.

The panelists also agreed that energy remains a major bottleneck for a variety of reasons. With gas being a go-to power source, a shortage of gas turbines is causing challenges for data centres trying to scale. I Squared fund partner Chenhua Shen said Asia’s power markets were less liberalised, making access to utilities harder.

“It becomes a major obstacle for Asia to leapfrog,” she said. “But we are still very optimistic in a way because Asia enjoys very abundant renewables.”

Uploaded by Magessan Varatharaja