Data as of: 2026-10-04 Issued: 2026-10-04 (America/Toronto) Evidence key: CONFIRMED (primary source), REPORTED (multiple outlets), RUMORED (unverified market reporting), ANALYTICAL (inference).
Power, not funding, is the binding constraint on the AI build-out — but funding terms are shifting anyway. At SuperReturn Asia in Singapore this week, infrastructure investors said they were wary of owning AI data centers as equity. Chenhua Shen, a fund partner at I Squared Capital, said uncertainty over future GPU generations and cooling loads made it more sensible to be a credit investor than an equity investor. Charles Wu, head of APAC alternatives at JP Morgan Asset Management, said he was “more excited about providing power to data centres” than holding equity in AI training facilities. Justin Chan-Sew, managing director of infrastructure debt at Ares Management, called power a “binary constraint,” noting US grid-connection times of four to five years and gas as the quickest route to scale.
Gulf states are turning that constraint into a pitch. Mohsin Pirzada, head of funds at the Qatar Investment Authority, told the same panel that Qatar’s cheap LNG-backed energy, land resources and coordinated compute procurement are designed to make domestic AI investment attractive. The pitch has precedent: Brookfield and QIA-owned Qai formed a $20 billion AI-infrastructure joint venture last December, QIA seeded a Blue Owl data-center platform with more than $3 billion in September 2025, and PIF-backed Humain is raising an initial $2.5 billion for Saudi data centers.
The signal is analytical but consistent: within days, three separate institutions articulated the same preference — lend to the power, avoid the equity. If that consensus hardens, the AI infrastructure boom will be funded increasingly by private credit and sovereign balance sheets rather than equity funds, repricing who bears technology-obsolescence risk in the data-center stack.