Goldman Sachs has updated its forecast for AI infrastructure spending by US hyperscalers, projecting capital expenditures could reach $1.4 trillion by 2027. This figure exceeds the bank's previous estimate of $1.1 trillion and the broader Wall Street consensus of $920 billion.
The bank previously projected spending at $405 billion in 2025 and $750 billion in 2026. Cumulative spending for the period from 2026 through 2031 is estimated at $7.6 trillion, including $5.1 trillion for compute, $2.1 trillion for data centers, and $358 billion for power infrastructure.
Key companies driving this investment include Microsoft, Amazon, Alphabet, Meta, Oracle, and OpenAI. These firms are moving from experimental AI deployment to full-scale commercial implementation, with median stock valuations for AI infrastructure currently at 26 times forward price-to-earnings ratios.
A significant shift in funding is underway, with over $400 billion of the 2027 expenditure expected to be covered by investment-grade bond issuance. This marks a departure from the traditional cash-flow-driven model used by major tech companies over the past two decades.
Analysts identify several constraints that could impact the buildout, including power availability, land accessibility, and the affordability of memory chips. The influx of new bond supply will also require absorption by fixed-income markets.
