Meta’s $145B AI Push Is Reshaping Its Workforce and Its Risks

Meta has marked artificial intelligence as its North Star, and it is spending accordingly: between $125 billion and $145 billion in capital expenditure this year alone, much of it on data centers for AI workloads. Inside the company, the bill is coming due. Roughly 8,000 jobs are being eliminated, an internal employee-tracking program that logged keystrokes and screenshots has been paused after a staff revolt, and one of the field’s most decorated scientists has walked out the door. Two jury verdicts, delivered within 24 hours of each other in March, have added a fresh layer of legal exposure, even as the advertising machine that funds all of it keeps printing money.

The company behind Facebook, Instagram, and WhatsApp reported $26.77 billion in net income for the first three months of 2026, up 61% from a year earlier, on revenue of $56.31 billion, according to Meta’s first quarter 2026 financial results. Advertising made up nearly all of it. The question inside Menlo Park is no longer whether Meta can afford its AI bet. It is what the bet is doing to the people who work there, the products that already ship, and the courts that are now starting to look more carefully.

The $145 Billion Bet Reshapes Meta’s Ranks

Meta announced the latest round of cuts in late April, telling staff it would eliminate roughly 10% of its workforce, or about 8,000 jobs, in a push the company called an efficiency drive tied to its AI spending. Bloomberg reported the plan as a way to “boost efficiency and offset its heavy spending” on infrastructure. The same round also closed about 6,000 open roles, according to NPR and ABC News, bringing total positions affected to roughly 14,000. Cuts began arriving via early-morning emails, with workers given short windows to clear their desks.

The math behind the cuts is brutal. Meta’s capital expenditure range for 2026 now sits at $125 billion to $145 billion, up from an earlier forecast of $115 billion to $135 billion, with much of the incremental spending earmarked for data centers and AI accelerators. That is roughly twice the $72.2 billion the company spent in 2025, according to investor data compiled by 247 Wall St. The same earnings release warned that Meta could face a “material loss” tied to ongoing youth-related trials in the United States.

On the other side of the ledger, Meta’s advertising machine is funding all of it. The company pulled in $56.31 billion in revenue for the quarter, up 33% year-over-year, with advertising contributing nearly 98% of the total. Family daily active people hit 3.56 billion in March 2026. Headcount actually ticked up 1% year-over-year to 77,986, because Meta is still hiring aggressively into AI roles while cutting elsewhere, the release noted.

• $26.77B in Q1 2026 net income, up 61% year-over-year
• $56.31B in Q1 2026 revenue, up 33% year-over-year
• 77,986 total headcount as of March 31, 2026
• 10% workforce reduction announced in late April 2026

Inside the Model Capability Initiative

In April 2026, Meta rolled out an internal program it called the Model Capability Initiative, or MCI. The tool captured mouse clicks, keystrokes, and screenshots from employees’ work laptops to train AI agents that could one day perform those same tasks, according to reporting by the BBC and The Verge. A Meta spokesperson told the BBC the data was “not used for any other purpose” and that the tool had “safeguards in place to protect sensitive content.” The same spokesperson added: “If we’re building agents to help people complete everyday tasks using computers, our models need real examples of how people actually use them.”

Employees were not buying it. The Guardian reported that more than 1,600 Meta workers signed an internal petition demanding the company stop harvesting “employee ‘computer use’ data.” One employee, speaking anonymously to the BBC, called the program “very dystopian,” adding that workers expected a wave of additional job cuts. Another, who had recently left, told the BBC the tracking was “just the latest way they’re shoving AI down everyone’s throat.” An internal memo from Stephane Kasriel, a vice president in Meta’s Superintelligence Labs unit, later offered staff the ability to pause data collection for “up to 30 minutes at a time.” The pause came after Wired reported that data tables containing “full prompts and transcriptions, private conversations, people and performance data” had been made accessible to anyone inside the company, as The Guardian’s report on the Meta employee-tracking program paused after a privacy backlash detailed.

The 1,600-Signature Revolt

Mark Zuckerberg, Meta’s founder and chief executive, has made the company’s AI drive his personal priority. He defended the program in an internal meeting by arguing that the average intelligence of Meta’s workforce sits well above the average intelligence available elsewhere, and that capturing their work would make the models smarter. He told employees that AI models “learn by watching really smart people do things,” according to The Guardian’s account of the meeting. He added that the coding skills of Meta engineers would “dramatically improve a model’s coding abilities.”

AI models learn by watching really smart people do things.

The framing landed badly with the rank and file. The petition warned that “collecting and repurposing this kind of data raises serious concerns around privacy, consent, and trust in the workplace.” Some signatories compared Meta to an “employee data extraction factory,” The Guardian and other outlets reported. The mood made its way onto Meta’s internal forum, where one worker posted a meme from “The Office” reading: “0 days since our last nonsense.” Another internal wrinkle: the program captured activity from both U.S. and European employees, which drew the attention of EU privacy regulators.

Meta suspended MCI on June 22, 2026. “While we have no indication at this time that any data was improperly accessed by Meta employees, we’re pausing it while we investigate,” a Meta spokesperson told The Guardian and AFP. Wired’s reporting also pointed to a related round of firings tied to leaks about the program. Business Insider described the episode as sharpening the broader privacy debate inside the company. The same AI push that motivated MCI has also driven Meta’s push to automate content moderation, an effort now answering a large slice of the same Meta’s push to automate content moderation work.

A Brain Drain at the AI Helm

The departures are not limited to the rank and file. Yann LeCun, Meta’s chief AI scientist and a Turing Award winner considered one of the “godfathers” of modern AI, told colleagues in November 2025 that he would leave at the end of the year to launch his own startup. LeCun had led Meta’s AI research since 2013, and his exit marked the end of a 12-year run at the company. He announced the move on LinkedIn, writing that he had spent “5 years as founding director” of Facebook AI Research before taking the chief scientist role.

LeCun had found himself reporting to Alexandr Wang, the 28-year-old founder and then-CEO of Scale AI, whom Zuckerberg had hired in mid-2025 to lead a new Meta Superintelligence Labs. Meta invested $14.3 billion in Scale AI as part of the deal, according to The New York Times, the largest such AI infrastructure investment the company had made at the time. The Financial Times reported that LeCun, more than 35 years Wang’s senior, was uncomfortable with the arrangement and believed Wang had “no experience with research” and was on “a dead end” quest. Other senior researchers and engineers have also walked out the door, according to reporting by CNBC and The Information, though Meta has framed Wang’s first year as a learning curve rather than a failure.

Two Verdicts in Two Days

Outside the company walls, the legal calendar is closing in. On March 25, 2026, a Los Angeles jury found Meta and Google’s YouTube liable for the mental health harms suffered by a 20-year-old woman identified in court as Kaley, marking the first time a U.S. jury had held a social media platform liable as a defective product for features like infinite scroll and constant notifications. A day earlier, a jury in New Mexico ordered Meta to pay for its failure to protect young users on Instagram and Facebook. The two rulings landed within 24 hours of each other.

Both verdicts came against the backdrop of Meta’s own warning in its Q1 2026 earnings release that additional youth-related trials could “ultimately result in a material loss.” Joseph VanZandt, co-lead lawyer for families suing the social media companies, said the LA ruling showed accountability had arrived for the industry. New Mexico Attorney General Raúl Torrez said juries in his state and California had “recognized that Meta’s public deception and design features are putting children in harm’s way.” Meta and Google have said they will appeal both rulings.

A second phase of the New Mexico trial is scheduled to begin on May 4, 2026. The state is asking a judge to declare Meta a public nuisance and force the company to overhaul safety features across its apps, with The Guardian reporting New Mexico is also seeking $3.7 billion in additional penalties. The cases sit alongside roughly 2,000 other pending lawsuits brought by parents and school districts, according to the Los Angeles social media addiction jury verdict coverage.

The trial turned in part on internal Meta documents shown to jurors, including one saying: “If we wanna win big with teens, we must bring them in as tweens.” Zuckerberg took the stand on February 18, 2026, telling jurors that keeping young users safe “has always been a company priority.” He added: “If people feel like they’re not having a good experience, why would they keep using the product?” NPR’s reporting on the New Mexico jury verdict against Meta over child safety details the parallel case.

Beyond the Chatbot: Glasses and Prediction Markets

Even as Meta pours money into data centers, the company is also pushing the bet out of the data center and onto consumers’ faces. Meta’s smart-glasses partnership with EssilorLuxottica sold more than 7 million AI-equipped Ray-Ban units in 2025, according to CNBC, and the eyewear maker’s latest launch extends the line into more frames and price points. The glasses ship with a camera, open-ear audio, and a Meta AI assistant; the company has said true augmented reality will follow in 2027. The hardware push is one of the few places Meta has actually shipped consumer AI in volume.

Meta is also testing a new product line far from its core social media apps. The New York Times, via The Guardian, reported Zuckerberg has directed a small team to build a smartphone app tentatively called Arena, a prediction-market product that would compete with Polymarket and Kalshi. About $24 billion in wagers are placed each month on Kalshi and Polymarket combined, according to Pew Research Centre figures cited by The Guardian. The proto-app remains in development and may not ship, a design choice covered separately in Meta’s prediction-market app targeting Polymarket and Kalshi.

Mike Proulx, a research director at the analyst firm Forrester, told The Guardian that moving into prediction markets was “not a great look” for a company under legal pressure over its social media products. The criticism echoed a broader question facing Zuckerberg’s strategy: how to grow beyond an ad business that is now nearly 98% of revenue, at a time when regulators are scrutinizing the existing ads machine more closely. Zuckerberg’s bet, in other words, is no longer just about catching up on frontier models. It is about finding new places to put the company’s reach before the courtroom catches up with the old ones.

Frequently Asked Questions

What was Meta’s Model Capability Initiative?

Meta’s Model Capability Initiative was an internal program launched in April 2026 that logged mouse clicks, keystrokes, and screenshots from employees’ work laptops to train AI agents. The company said the data was “not used for any other purpose” and had “safeguards in place to protect sensitive content,” but more than 1,600 workers signed an internal petition calling for it to end. Meta suspended MCI on June 22, 2026.

Why is Meta laying off employees while reporting record profit?

Meta said in late April 2026 that it would cut roughly 10% of its workforce, or about 8,000 jobs, to “boost efficiency and offset its heavy spending” on AI infrastructure, per Bloomberg. The cuts come as Meta has raised its 2026 capital expenditure forecast to between $125 billion and $145 billion, much of it earmarked for AI data centers. The company still reported $26.77 billion in net income for the first quarter of 2026, up 61% from a year earlier.

What did Mark Zuckerberg say about training AI on employees?

At an internal meeting, Zuckerberg told staff that AI models “learn by watching really smart people do things,” according to The Guardian’s account of the meeting. He added that the coding skills of Meta engineers would “dramatically improve a model’s coding abilities.” The remark surfaced as the Model Capability Initiative was already drawing a petition signed by more than 1,600 workers.

When did Yann LeCun leave Meta?

Yann LeCun, Meta’s chief AI scientist and a Turing Award winner, told colleagues in November 2025 that he would leave the company at the end of the year to launch his own AI startup focused on “world models.” He had led Meta’s AI research since 2013. The Financial Times reported that LeCun had grown uncomfortable reporting to Alexandr Wang, the 28-year-old founder of Scale AI whom Zuckerberg had hired to lead Meta Superintelligence Labs.

What jury verdicts has Meta faced in 2026?

A Los Angeles jury on March 25, 2026, found Meta and Google’s YouTube liable for the mental health harms suffered by a 20-year-old plaintiff, awarding $6 million in total damages, with Meta on the hook for 70%. A day earlier, on March 24, a New Mexico jury ordered Meta to pay $375 million for failing to protect young users from child predators on Instagram and Facebook. Both companies have said they will appeal.