Goldman Sees Hyperscaler AI Capex Rising 50% to $1.2 Trillion

Spending on AI infrastructure by the five largest US hyperscalers is set to increase by more than half next year to $1.2 trillion, according to strategists at Goldman Sachs Group Inc.

That figure is above the Wall Street consensus of $1.1 trillion in 2027 capex, the team led by Ryan Hammond said in a note. To break even on their outlays, the companies would need annual artificial intelligence revenues of about $300 billion in the coming years, they said.

“Based on consensus estimates, capex in 2027 is on track to reach a larger share of GDP than any technological investment cycle since the railroad build-out in the late 1800s,” Hammond and his colleagues said.

Amazon.com Inc., Alphabet Inc., Microsoft Corp., Oracle Corp. and Meta Platforms Inc. are on course to spend $800 billion this year developing their AI capacity. While next year’s figure represents a jump, the pace at which capex is rising is set to slow, the Goldman team said.

“The rate of both capex growth and upside surprises to consensus estimates will diminish relative to recent quarters,” the strategists wrote. Hyperscaler capex now exceeds cash flow from operations, increasing the need for debt and equity financing. Potential constraints include shortages of power, labor and memory as well as restrictions on data-center construction.

They expect hyperscaler capex growth to ease from nearly 100% this year to 54% in 2027 and to 12% in 2028, when it’s projected to hit $1.4 trillion.

AI users would need to spend roughly $1 trillion annually on applications for hyperscalers to generate solid returns and for application providers to earn strong margins on their computing costs, the strategists said. That compares with about $1.5 trillion of global software spending in 2026.

Read more: Meta’s Best Month Since 2013 Has It on Cusp of $2 Trillion Level

Current estimated AI revenue remains below the level required for hyperscalers to break even, though sales are rising quickly, the Goldman team said. Cloud revenue growth at Amazon, Alphabet, Microsoft and Oracle accelerated to 48% in the second quarter of this year from 25% in 2024. Amazon, Alphabet and Microsoft have also reported a combined $1.7 trillion of cloud revenue backlog.

Some investor skepticism about the durability of the AI boom is reflected in share prices. The median AI infrastructure stock now trades at 22 times forward earnings, down from 32 times in April, the strategists said. The hyperscalers have de-rated to their lowest multiples in more than a decade and trade at the smallest premium to the median S&P 500 Index company on record.