Meta, the parent company of Facebook and Instagram, is preparing to lay off approximately 8,000 employees in May, or roughly 10% of its global workforce. And that is just the beginning. Another round of cuts will come in the second half of the year. How far will it go?
According to three Reuters sources familiar with management’s plans, Meta has chosen May 20 as the date for the first wave of layoffs. The company employs nearly 79,000 people, and the first round will affect approximately 8,000 of them. The details of the second wave, including its exact date and scope, have not yet been finalized. Management reportedly wants to retain the flexibility to respond to developments in artificial intelligence. Meta declined to comment on the plans. A spokesperson neither confirmed nor denied the scale of the planned cuts.
This is the largest round of layoffs since 2022, when Meta cut more than 21,000 jobs as part of its so-called “year of efficiency.” At the time, the company was facing a collapse in its share price and a crisis following pandemic-era overheating. The situation is different now. Last year, Meta generated more than $200 billion in revenue and posted a profit of around $60 billion. This time, the layoffs are not an emergency brake but purely a strategic choice.
Savings from layoffs will fund AI infrastructure
Meta plans to spend $115 billion to $135 billion on artificial intelligence infrastructure in 2026. This includes data centers, graphics processing units, networks, and proprietary chips. Compared with approximately $72 billion last year, that is nearly double. And the money has to come from somewhere.
Zuckerberg has said it openly: 2026 is the year when artificial intelligence will dramatically change the way we work. Projects that once required dozens of people can now be handled by one talented person with the help of AI tools. Smaller teams and more autonomous agents instead of employees. Meta is also building a new division called Applied AI, to which it is transferring engineers from across the company. Its goal is to develop autonomous AI agents capable of writing code and independently completing complex tasks. The company has also created a Meta Small Business unit, where it is transferring some employees from positions being eliminated.
In its longer-term outlook, the company is openly discussing investments of up to $600 billion in US AI infrastructure by 2028. Such a move logically requires savings to be made somewhere.
As recently as March, the company was advertising approximately 800 vacancies. There are currently seven listed on its website. A spokesperson declined to comment on either the disappearance of the job postings or the layoff plans.
Employee monitoring: every click, every keystroke
Alongside the layoffs came another announcement that caught employees by surprise. Meta informed them that it was launching a new tool called the Model Capability Initiative (MCI). It will run on company computers and in internal applications, recording mouse movements, keystrokes, and generally how employees work. The purpose? To train artificial intelligence models. Meta wants to build agents that will help people handle everyday computer tasks. To do that, it needs real examples of how people use computers.
A Meta spokesperson assured the BBC that the data will be used exclusively for model training and nothing else. The system reportedly includes safeguards to protect sensitive content. Access to employees’ activity on company devices was technically possible before, but specifically monitoring it for the purpose of AI training is new.
Employees see it differently. One of them, who did not wish to be named, described it as “very dystopian.” A company that is simultaneously preparing further layoffs will now record every move made by the workers who remain. According to another employee who recently left Meta, it is just “another way for the company to force AI down your throat.”
Shares surged
Markets reacted surprisingly positively to the news of the layoffs. Meta shares jumped 5.7% after the plans were announced. They are up approximately 3.7% since the beginning of the year, although they remain below last year’s record high.
Analysts at Simply Wall St estimate that Meta could reach revenue of $275.9 billion and profit of $92.1 billion by 2028. They estimate the shares’ fair value at $835, which would represent an increase of approximately 25% from the current price.
Investors are encouraged by precisely this combination: massive investment in AI infrastructure funded through layoffs and a shift to smaller, more efficient teams. The risk is that spending could outpace revenue. Analysts’ pessimistic scenarios project revenue of only $321.6 billion by 2029, while costs could continue to put pressure on margins.
Meta’s layoffs are not an isolated phenomenon. In recent months, Amazon has cut more than 30,000 corporate employees, or nearly 10% of its white-collar workforce. Block, a digital payments provider, laid off nearly half of its team in February. In both cases, management linked the cuts to efficiencies achieved through artificial intelligence. In total, more than 73,000 people worldwide have lost their jobs in the technology sector so far this year. For all of 2024, the figure was approximately 153,000.
Llama, Muse Spark, and the battle with OpenAI
Zuckerberg does not just want to cut costs. He wants to win the AI race. Meta is continuing to develop the Llama 4 models and is preparing Llama 4.5. This year, it launched Muse Spark, the first frontier model from Meta Superintelligence Labs. It is a multimodal model with reasoning, tool use, and multi-agent coordination capabilities. It is Meta’s first model not to be released as open source. Muse Spark is locked into the company’s products and a private API. It currently trails Google, Anthropic, and OpenAI in benchmarks, but experts say its architecture is on par with them.
Meta’s strategy is clear: open-source models (Llama) for the community and developers, and closed models for monetization. Meta is planning a paid API for premium models, while open models maintain developer interest and provide the underlying ecosystem.
