To stay competitive in the artificial intelligence race, Meta increased its spending on data centers last year and began shifting away from virtual reality.
On Wednesday, the company said it would bet even bigger this year.
Meta, which owns Facebook, Instagram and WhatsApp, set its capital expenditure forecast for 2026 at $115 billion to $135 billion, nearly twice the $72 billion the company spent last year. Much of that will go to building computing facilities that power AI, alongside spending to hire researchers to develop a “superintelligent” AI model with godlike abilities.
The projected jump was enormous. Meta, which had spent $28 billion in 2023 and $39 billion in 2024, now plans to shell out sums that may be even bigger than larger AI rivals like Google, which spent $93 billion last year.
Meta’s core business of online advertising continued to grow, providing cash for its AI spending. For the fourth quarter, revenue was $59.89 billion, up 24% from a year earlier. Profit was $22.76 billion, up 9.2%. The company attributed the growth to its AI investments, which it said helped improve ad targeting and recommendations of videos and other posts to users.
“Our world class recommendation systems are already driving meaningful growth across our ads systems, but we think our current systems are primitive compared to what’s to come,” Mark Zuckerberg, Meta’s chief executive, said on a conference call with investors.
Meta’s stock rose 10% in after-hours trading.
“I don’t think investors care as much about capital expenditures with Meta now given the consistent traction with their AI optimization tools,” said Eric Seufert, an independent mobile analyst who closely watches Meta. “The investments are driving real growth across a number of dimensions.”
After falling behind in the AI race last spring, Meta overhauled its AI division and poured tens of billions into new talent and leadership. In its largest investment, in June, it put $14.3 billion into Scale AI, an AI data labeling startup, and appointed that company’s chief executive, Alexandr Wang, as its new chief AI officer.
Meta promptly reorganized its AI division and transferred leadership of its AI models and products from longtime executives to new hires. To develop its next AI model, the company created a new group, TBD Lab, which is being led by Wang and Nat Friedman, a former chief executive of GitHub.
TBD Lab’s model, code-named Avocado, is expected to be released in the first half of this year. How it performs against models from Google and OpenAI will be the first major test of Zuckerberg’s recent AI investments. Yann LeCun, who was Meta’s chief AI scientist before leaving last year, recently said that U.S. companies had a “herd” mentality and were too focused on large language models, the AI technology at the heart of popular products like ChatGPT.
Wednesday, Zuckerberg appeared to dampen some AI expectations. “I expect our first models will be good, but more importantly will show the rapid trajectory we’re on,” he said.
Meta’s AI spending spree continued last month when it bought Manus, an AI startup based in Singapore, for $2 billion. The deal came after a federal judge ruled in November that Meta did not violate U.S. antitrust law by buying WhatsApp and Instagram a decade ago, two acquisitions that experts said had opened the door for more dealmaking by Silicon Valley giants.
The bulk of Meta’s new AI spending will be on infrastructure like data centers, under an initiative it announced this month called Meta Compute. Meta hired Dina Powell McCormick, a former adviser to President Donald Trump and banker at Goldman Sachs, to be its new president and vice chair focused on data center banking deals.
Zuckerberg said Meta Compute would lead to data centers powered by “tens of gigawatts this decade, and hundreds of gigawatts or more over time.”
This month, Meta also laid off 10% of employees in its Reality Labs division, which oversees products like virtual reality headsets. Meta said that it had not given up on the metaverse, an immersive world that it had bet big on, and that it was still investing in products like augmented reality glasses.
Some investors are worried that Meta has fallen behind in AI even as its spending has ramped up, said Uday Cheruvu, a portfolio manager at the investment firm Harding Loevner. “The benefits of AI have to keep up with the rate of the spending,” he added.
Meta’s family of apps now has 3.58 billion users, up 7% from a year earlier. The company’s workforce grew 6% to 78,865.
Reality Labs posted revenue of $955 million in the quarter, down from $1.1 billion a year earlier. Meta said its AI smart glasses, which are developed by Reality Labs, were a hit for the company, selling more units than expected. The division lost $6 billion, more than a year earlier.
Zuckerberg said Wednesday that he expected this year to be the “peak” of losses for Reality Labs.
On Wednesday morning, Zuckerberg’s personal philanthropy, the Chan Zuckerberg Initiative, said it was laying off about 8% of employees, or around 70 people. Those reductions follow a restructuring of the organization around AI and science, changes that Zuckerberg said last year were important to ensure the organization kept a “focus on something that wasn’t going to be undone every few years.”
A spokesperson said the organization was still focused on scientific research, and was making “thoughtful adjustments to our team to ensure we are best positioned to drive this effort forward.”
Curated by Aisha Patel
