Meta has begun laying off employees to focus its resources on building data centers, training its own large-scale language models, and hiring AI talent.
A person familiar with the injury said. register The number will be about 700 people.
Job losses will be most severe at Meta’s Reality Labs, its social media division, and its recruiting division, according to The Information.
“After six years at Meta, my role has been impacted by today’s recent layoffs,” the woman, who until this morning worked as a senior recruiter at Meta, wrote in a LinkedIn post. “This time has been especially tough. I am grateful to have been offered full-time again last year after returning as a short-term employee in 2024, and I am extremely proud of what I have been able to accomplish in that time. My gratitude far outweighs my disappointment.”
in a statement to registerMeta said the job cuts were aimed at streamlining its business to work more effectively with AI, as Meta CEO Mark Zuckerberg said in a January earnings call.
“Teams across Meta regularly undergo restructuring or changes to ensure they are best positioned to achieve their goals. Where possible, we find alternative opportunities for employees whose positions may be affected,” the spokesperson wrote.
In a post-earnings note on Jan. 28, Zuckerberg said this is the year Meta begins to “flatten the team.”
“We are elevating individual contributors and flattening teams. Projects that once required large teams are now being accomplished by a single, highly talented person,” he writes. “I want as many of these incredibly talented people as possible to choose Meta as the place where they can have the greatest impact, delivering personalized products to billions of people around the world. And if we can do this, we can accomplish more and have more fun.”
Reuters recently reported that Meta plans to lay off 20% of its workforce, or about 15,000 people, but the layoffs, which reportedly began this week, have so far been small.
Meta said the company had 78,800 employees at the end of January, a number that had grown in recent years as it sought to build a bench of AI talent capable of building a platform that could compete with frontier model providers like Anthropic and OpenAI.
If Meta cuts its workforce by 20%, it would eliminate about 15,000 jobs, making Meta’s lowest headcount since 2021, when it had about 58,600 full-time employees.
Spending on AI
Meta is focused on building its own AI infrastructure and data center assets to rival competitors Anthropic, Google, and OpenAI, and has significantly increased spending in recent years to stay in the AI arms race. Expenses will rise 24% to $118 billion during 2025, and the company said it plans to spend between $162 billion and $167 billion this year (though it expects operating income to increase, meaning revenue will grow faster than expenses). Of this, capital expenditures, including data center expansion to strengthen AI efforts, will be between $115 billion and $135 billion.
The company is also designing its own custom chips for GenAI workloads that it plans to build over the next two years. The first in-house MTIA chip was released in 2023.
“MTIA 300 is used for ranking and recommendation training and is already in production. MTIA 400, 450, and 500 can handle all workloads, but in the near future and into 2027, we will primarily use these chips to support GenAI inference operations,” the company said.
According to the New York Times, the release of Meta’s next inference model (codenamed “Avocado”) has been postponed due to unsatisfactory results from internal testing. The news comes despite Meta offering dramatic nine-figure pay packages to lure AI researchers from competitors last year, and a reported hefty $100 million sign-on bonus for defectors from OpenAI.
Zuckerberg also invested $14 billion in Scale AI and named its co-founder Alexander Wang to lead Meta’s AI efforts. Mr. Wang reportedly clashed with Meta’s former chief AI scientist Yan Rukun, who called him too young and inexperienced after he left Mr. Wang. LeCun is also known as the godfather of AI. LeCun accused Zuckerberg of sidelining his former AI team after the release of the company’s disappointing Llama 4 model.
Recently, Facebook CFO Susan Lee spoke with analysts at the Morgan Stanley Technology, Media and Telecom Conference about the uncertain path of Meta’s AI investment returns.
“It’s not like this is the ROI in 2026 and this is the ROI in 2027. Frankly, this is painful to me,” Lee said. “I really wish that was the world we live in, but it’s not. And we have to be willing to make temporary bets, and that’s a big part of what we have to do in an intelligent and thoughtful way.”
He said Meta can accurately measure the cost of people and infrastructure to run the platform’s existing apps and experiences. You can also calculate how much it costs to build new AI capabilities, including employee and computing costs.
But there is a blind spot when it comes to estimating how much inference power the company will need if the AI products it makes need to scale quickly to meet user demand.
“The teams that are fundamentally working on AI training today have the most immediate and well-defined roadmap of how much capacity they have, for example, they think they need to train models over the next 12 to 24 months,” Lee said. “It’s like a demand roadmap from teams that have more certainty. For us, I think the hardest part of gaining certainty is the inference need, because we need to meaningfully predict the future to ensure lead times and production capacity.” ®
