Hyperscalers' AI spending will rise by 50% in 2027 — Goldman Sachs

Goldman Sachs strategists predict that the largest U.S. hyperscalers will increase their spending on artificial intelligence infrastructure by more than half next year—to $1.2 trillion. To recoup these investments, companies will need to generate about $300 billion in annual revenue from AI, they estimate.

Amazon, Alphabet, Microsoft, Oracle, and Meta plan to spend about $800 billion on AI infrastructure in 2026. And in 2027, their combined capital expenditures will rise by 54% to $1.2 trillion, according to a forecast by Goldman Sachs analysts led by Ryan Hammond, Bloomberg reports. Goldman Sachs’ estimate is 9% higher than Wall Street’s consensus forecast.

"In 2027, capital expenditures could account for a larger share of GDP than during any technology investment cycle since the construction of railroads in the late 19th century," Hammond and his colleagues stated.

At the same time, the rate of spending growth will slow: Goldman Sachs estimates that it will fall from nearly 100% this year to 12% in 2028, when the aggregate spending of hyperscalers will reach $1.4 trillion. Hyperscalers’ capital expenditures already exceed their operating cash flow, which increases these companies’ need for debt and equity financing. Power, labor, and memory shortages, as well as restrictions on data center construction, could pose additional constraints on the expansion of AI infrastructure, the bank’s strategists note.

For hyperscalers to recoup their investments, AI users will need to spend about $1 trillion annually on applications. This will allow developers to achieve high profit margins, even after accounting for computing costs. By comparison, global software spending in 2026 is estimated at approximately $1.5 trillion, according to Bloomberg.

According to Goldman’s estimates, current AI revenue has not yet reached the level needed for hyperscalers to break even. At the same time, the cloud business is growing: the combined growth rate of cloud revenue for Amazon, Alphabet, Microsoft, and Oracle accelerated to 48% in the second quarter from 25% in 2024. Amazon, Alphabet, and Microsoft also reported a combined cloud order backlog of $1.7 trillion, according to Bloomberg.

At the same time, investors have become more cautious in their assessment of the prospects for AI infrastructure. The average price-to-earnings ratio for companies in this segment has fallen to 22 times forward earnings from 32 in April. The price-to-earnings ratios of the hyperscalers themselves have fallen to their lowest level in more than a decade, and their premium relative to the median S&P 500 company has reached an all-time low, Bloomberg concludes.

This article was AI-translated and verified by a human editor