Is big tech spending too much on AI? Goldman strategists think so

Investors are growing increasingly concerned that US technology megacaps are spending too much on artificial intelligence, according Goldman Sachs Group Inc. strategists.

Companies that the strategists refer to as “hyperscalers” — including Amazon.com Inc., Meta Platforms Inc., Microsoft Corp. and Alphabet Inc. — have utilized about $357 billion for capital expenditure as well as research and development in the past year, the team led by Ryan Hammond said.

A “significant portion” of this spending was on AI and represents nearly a quarter of the S&P 500’s total for capex and R&D, the strategists said.

“Today’s hyperscalers will eventually be required to prove that revenues and earnings will be generated from their investments,” Hammond wrote in a note. “Early signs that may not be generated, could lead to valuation de-rating.”

Amazon is expected to utilize $63 billion for capex this year, an increase from $53 billion in 2023, according to data compiled by Bloomberg. Meta and Google-owner Alphabet will spend record amounts in 2024.

The buzz around AI has powered US stocks to record highs this year, with Nvidia Corp. among the biggest beneficiaries of the trade. Investors broadly expect the frenzy to remain a key feature of a rally in the second half, although some are betting on sectors such infrastructure providers and utilities to lead gains for the remainder of 2024.

Even so, the strategists said that current AI spending “still pales in comparison” to capex levels seen during the dot-com crash.

At the height of the tech bubble at the turn of the millennium, the group of tech, media and telecom stocks was spending more than 100% of cash flows from operations on capex and R&D, Hammond said. Today, the total bill stands at 72%.

Still, expenses associated with depreciation could represent a risk to profitability, the strategist said.

“As demonstrated in the tech bubble, sales revisions will be a key indicator for investors to assess the durability of the AI trade,” Hammond said.