SEATTLE — Mark Zuckerberg, Meta’s CEO, started 2023 by declaring it the “year of efficiency.” Like several of its big tech peers, Meta cut jobs and mothballed expansion plans.
Then came AI.
Zuckerberg started this year saying his company would spend more than $30 billion on new tech infrastructure in 2024. In April, he raised that to $35 billion. On Wednesday, he increased it to at least $37 billion. And he said Meta would spend even more next year.
Zuckerberg said he’d rather build too fast “rather than too late,” and allow his competitors to get a big lead in the AI race.
The tech industry’s biggest companies have made it clear over the past week that they have no intention of throttling their stunning levels of spending on artificial intelligence, even though investors are getting worried that a big payoff is further down the line than once thought.
In the past quarter alone, Apple, Amazon, Meta, Microsoft and Google’s parent company Alphabet spent a combined $59 billion on capital expenses, 63% more than a year earlier and 161% more than four years ago. A large part of that was funneled into building data centers and packing them with new computer systems to build artificial intelligence. Only Apple has not dramatically increased spending because it does not build the most advanced AI systems itself.
If investors are getting anxious, they’re going to have to learn to cope with their nerves. Last week, Alphabet’s share price dropped more than 5% after it reported a 91% increase in capital expenses. But Sundar Pichai, Alphabet’s CEO, made the case for patience.
“These things take time,” he said, and “the risk of underinvesting is dramatically greater than the risk of over-investing.”
The leaders of the biggest tech companies see a once-in-a-generation opportunity in the generative AI technology behind popular chatbots like ChatGPT. They believe it can revolutionize everything from the software that runs the complex operations of global companies to research on new drugs.
When ChatGPT debuted in late 2022, tech giants were beginning to dial back a burst of spending from the pandemic. But the industry’s brief embrace of austerity went out the window when they saw the potential of artificial intelligence.
This new wave of AI is wildly expensive. The systems work with vast amounts of data and require sophisticated computer chips and new data centers to develop the technology and serve it to customers. The companies are seeing some sales from their AI work, but it is barely moving the needle financially.
In recent months, several high-profile tech industry watchers, including Goldman Sachs’ head of equity research and a partner at venture firm Sequoia Capital, have questioned when or whether AI will ever produce enough benefit to bring in the sales needed to cover its staggering costs. It is not clear that AI will come close to having the same impact as the internet or mobile phones, Goldman’s Jim Covello wrote in a June report.
“What $1 trillion problem will AI solve?” he wrote. “Replacing low-wage jobs with tremendously costly technology is basically the polar opposite of the prior technology transitions I’ve witnessed in my 30 years of closely following the tech industry.”
Google’s parent company was the first big tech outfit to report its earnings for April through June, and it added weight to those overspending concerns. Though Alphabet had a 29% jump in profit, sales for ads on YouTube, which Google owns, were lower than expected and the massive jump in infrastructure spending — Google spent an average of $145 million each day in the quarter — rattled investors.
Microsoft was up next. As OpenAI’s largest investor, it had a jump on its peers and had been raising its capital spending every quarter since the start of last year. On Tuesday, Microsoft had a bit of unwelcome news: Its cloud computing business, where most of that AI work was being done, didn’t grow as fast as expected.
But instead of serving as a moment of caution, the miss (about 1 percentage point below what was expected) added fuel to the building frenzy. Executives said that Microsoft had more demand for AI than it could serve from its data centers, a problem that they expect will persist through the end of the year. That helped explain why they are building so furiously.
Satya Nadella, Microsoft’s CEO, said much of the capital spending went to acquire physical land and buildings, which they need to lock up in advance. But the remaining 60% was for what he called “the kit,” the expensive chips and other components of a computer network.
Microsoft’s executives also asked for patience, saying the spending would bring in revenue over a long time — “over 15 years and beyond,” said Amy Hood, the company’s financial chief. Digging into Microsoft’s numbers indicates that the company is on track for more than $5 billion in sales for generative AI products this year. But for a tech giant that just reported $245 billion in annual revenue, that’s still tiny.
The next day, Meta increased its predictions for how much it would spend. Zuckerberg said he was planning for the next generation of AI systems, and the next major update to the company’s main AI model will demand 10 times more computing power.
Meta gives away the advanced AI systems it develops, but Zuckerberg still said it was worth it. “Part of what’s important about AI is that it can be used to improve all of our products in almost every way,” he said.
Amazon told investors it had spent more than $30 billion on capital expenses in the first half of the year, and that it too would spend more in the rest of the year. Executives said they need to balance building enough to meet demand without getting ahead of what they really need.
“The reality right now is that while we’re investing a significant amount in the AI space and in infrastructure, we would like to have more capacity than we already have today,” said Andy Jassy, Amazon’s CEO. “I mean, we have a lot of demand right now.”
That means buying land, building data centers and all the computers, chips and gear that go into them. Amazon executives put a positive spin on all that spending. “We use that to drive revenue and free cash flow for the next decade and beyond,” said Brian Olsavsky, the company’s finance chief.
There are plenty of signs the boom will persist. In mid-July, Taiwan Semiconductor Manufacturing Co., which makes most of the in-demand chips designed by Nvidia that are used in AI systems, said those chips would be in scarce supply until the end of 2025.
Zuckerberg said AI’s potential is “super exciting.”
“It’s why there are all the jokes about how all the tech CEOs get on these earnings calls,” he said, “and just talk about AI the whole time.”
This article originally appeared in The New York Times.
Get more business news by signing up for our Economy Now newsletter.