AI capital expenditures are eating into the pools of money once devoted to stock buybacks.
Vast data-center ambitions and a memory-chip crunch are driving Big Tech's artificial-intelligence spending higher - and it's likely coming at the expense of stock buybacks and dividends.
The biggest AI hyperscalers - that is, large cloud-services and technology providers like Amazon.com (AMZN), Alphabet $(GOOGL)$ $(GOOG)$, Meta Platforms (META), Microsoft $(MSFT)$ and Oracle $(ORCL)$ - will spend $755 billion on capital expenditures this year, an 83% jump, Goldman Sachs analysts said in a note on Thursday. To help facilitate that, those companies cut buybacks by nearly two-thirds in the first quarter, according to the analysts.
They added that the hyperscalers are now putting 20% of their total spending toward buybacks and dividends, compared with 34% on average from 2017 to 2022.
Google parent Alphabet didn't buy back any stock in the latest quarter, after repurchasing around $15.1 billion in the same period a year before. Amazon hasn't bought back stock in nearly four years, according to FactSet data, and Meta declined to repurchase shares in both of the last two quarters.
Of the hyperscalers, only Microsoft repurchased stock in the March quarter, and at a roughly equivalent rate to what it had done a year before.
Apple $(AAPL)$ is not spending nearly as much on AI as its Big Tech peers, but the company roughly halved buyback spending to $12.3 billion in its most recent period.
Much of the tech industry's spending increase has come as the costs for memory capacity rise, with AI demand chewing up available supply and threatening to push prices higher for electronic devices.
Both Microsoft and Meta, during their earnings calls last month, highlighted the effects of "higher component pricing" on their spending plans. Meanwhile, videogame giant Nintendo called out those higher costs in its earnings materials on Friday.
Investors have grown more worried over Big Tech's astronomical spending levels on AI. The Goldman analysts said the hyperscalers' capital-expenditure growth this year is on pace to reach 100% of their cash flows from operations.
The analysts noted that for capex growth next year to match levels this year, the hyperscalers would need to spend around $1.4 trillion. The end result, they said, is that Big Tech could have to borrow hundreds of billions of dollars more.
The consensus view "estimates that cash flows from operations for these companies will equal $980 billion," the Goldman analysts wrote. "Without sharp upward revisions to hyperscaler cash-flow growth, these companies would need to increase net debt by $400 billion."
For the S&P 500 SPX at large, the analysts forecast aggregate capital spending among the index's member companies will increase 33% this year, largely driven by Big Tech. But amid worries about the economy and the Iran war, companies in the index are expected to increase gross buybacks by only 3%.
"Uncertainty related to tariffs has largely been replaced with uncertainty related to commodities and supply chains," the Goldman analysts said, adding that companies were likelier to rethink their spending on share repurchases during times of elevated uncertainty.
The analysts added that they expect buyback growth to be subdued in 2027 - but said Big Tech's spending spree could translate into buybacks elsewhere.
"Some of the buyback headwind from the hyperscalers will likely be offset by increased buyback activity among the beneficiaries of that capex, such as semiconductor firms," the Goldman analysts wrote.
Comments