By Robbie Whelan
Big tech companies are starting to strike gold with artificial intelligence, but investors are still wringing their hands over the multibillion-dollar bets on data centers and chips.
Four of the biggest names in technology -- Microsoft, Alphabet, Meta Platforms and Amazon.com -- on Wednesday reported earnings showing that sales are growing thanks to the proliferation of AI tools. That progress, however, comes at a steep cost. Capital expenditures on the infrastructure needed to satisfy demand are climbing steadily higher.
Microsoft, Alphabet, Meta and Amazon last year combined for $410 billion in capital expenditures and are expected to spend more than $670 billion more on capex in 2026, according to a Wall Street Journal tally. Morgan Stanley estimates that tech companies will spend $2.9 trillion on chips, servers and other pieces of data-center infrastructure between 2025 and 2028.
The costs of those inputs, which keep the AI engine humming, keep rising. Prices for memory chips, which are essential for zipping data back and forth between the computer processors and databases that allow AI models to run, have skyrocketed as surging demand for AI tools has caused a shortage. There are demand-driven capacity crunches in everything from fiber-optic cabling to electric power to water for cooling chip factories to undeveloped land for data centers.
"We're seeing constraints across the board. The hyperscalers who are trying to get into the gold mine -- they're having to wait, or spend more to get in," said Brent Thill, a tech analyst at Jefferies. "It's good for the picks and shovels, but it's not good for the people who are assembling all the pieces."
Wednesday's results highlighted a bifurcation in Big Tech between the fortunes of companies that have invested heavily at several levels of the AI stack, including designing cutting-edge microchips, cloud services and models, and those that are more reliant on partners for hardware or services that they or their customers need.
Alphabet posted the strongest quarterly results of the four, reporting an 81% increase in net income compared with a year eralier. Its Google Cloud unit, which rents out access to its in-house AI chips, known as Tensor Processing Units, posted $20 billion in first-quarter revenue, a 63% year-over-year increase. Chief Executive Sundar Pichai said that AI is "lighting up every part of the business." The company's remaining performance obligations, or backlog of customer commitments, grew 398% year-over-year to $460 billion, largely on the strength of cloud sales.
In a major shift, Pichai announced that Google would begin direct sales of TPUs to a select group of external customers, the company's first significant step to commercialize its chips outside its own cloud business. Alphabet raised its capex projection for 2026 by $5 billion to a new range of $180 billion to $190 billion, a result of its recent acquisition of the data center energy provider Intersect.
Alphabet shares rose 7% in aftermarket trading.
Amazon, which has been rapidly expanding its in-house custom silicon business, saw income at its Amazon Web Services data-center segment surge 28% year-over-year, to $37.6 billion. Amazon produces AI chips including its Trainium accelerators and Graviton CPUs.
The company's shares rose 2.7% in afterhours trading, despite reporting that it had spent $59.3 billion more on property and equipment, mostly related to its AI investments, than in the year-earlier quarter. That left the $2.8 trillion company with just $1.2 billion in annual free cash flow, much lower than was typical in the years leading up to the AI boom.
Meta shares fell about 7% after the market close despite big gains in ad revenue it attributed to AI-driven improvements in ad targeting and tracking. The company far exceeded analysts' expectations for both quarterly sales and net income.
The social-media company told investors that it expected capital expenditures of between $125 billion and $145 billion in 2026, up from a previous estimate of $115 billion to $135 billion. The increased spending projection is due to "expectations for higher component pricing" and "additional data center costs," the company said.
Meanwhile, Microsoft beat Wall Street's expectations for sales, earnings per share and operating margins, but shares still declined after the bell and were later little changed. The company's Intelligent Cloud unit, which includes its AI infrastructure and Azure cloud computing businesses, reported $13.2 billion in operating income, lower than the $14.1 billion predicted by analysts surveyed by FactSet.
On an earnings call with investors, Microsoft Chief Financial Officer Amy Hood said the company expects to increase capital spending to $190 billion for all of 2026, with $25 billion of that reflecting higher-component pricing. Hood said despite rising costs, she was optimistic these investments would pay off once the company gets more infrastructure in place.
"I feel quite good about our ability to work through the physical sort of limitations," Hood said.
Microsoft shares have fallen about 20% over the past six months as investors remain concerned about the company's dependence on -- and at times tense relations with -- OpenAI. The company reached a truce with its longtime partner earlier this week.
"The moral of the story is, cloud businesses are accelerating, and you're seeing particular strength behind vertical integration," said John Belton, portfolio manager of the $1.4 billion Gabelli Growth Fund, which has about 30% of its holdings in Alphabet, Amazon, Meta and Microsoft.
"That means, if you're a cloud services company, and you have a full suite of computing services, from the chip down to the model, down to the application, you're doing much better, versus if you're just building data centers and running more third-party models."
Write to Robbie Whelan at robbie.whelan@wsj.com
(END) Dow Jones Newswires
April 29, 2026 21:21 ET (01:21 GMT)
Copyright (c) 2026 Dow Jones & Company, Inc.
Comments