Overview
- Broadcom CEO Hock Tan and Intel CEO Lip‑Bu Tan said at a Singapore conference on Monday that access to capital and higher cost of borrowing may be the biggest constraint on expanding AI infrastructure.
- Bank research and recent company results show demand for AI compute remains strong with UBS forecasting average hyperscaler cloud revenue growth of about 58% year‑on‑year for the September quarter.
- Analysts are tracing a shift in bottlenecks from chips to power to capital, with Goldman Sachs saying grid capacity for 2026–27 is largely booked and hyperscalers are locking power contracts for 2028 and later.
- Markets and financiers are already reacting: rising bond yields and investor caution have driven semiconductor selloffs, at least one large data‑centre IPO to be shelved, and the creation of multi‑billion financing platforms and syndicated loan packages tied to chip and data‑centre leases.
- The new financing structures concentrate credit risk because a few hyperscalers drive most demand and some vendors now act as lenders, which raises scrutiny on customer profitability and could slow or reshape the pace of the multi‑trillion dollar AI buildout.