The Brief

Meta is internally discussing workforce reductions of up to 20 percent — roughly 16,000 of its 79,000 employees — to offset AI capital expenditures projected to reach $135 billion in 2026, nearly double last year’s $72 billion. No date or final headcount has been set, and a Meta spokesperson described the reporting as “speculative.” The cuts would represent the company’s largest restructuring since the 2022–2023 layoffs that eliminated 21,000 roles.

The Report

Meta Platforms is preparing for what could be its most significant workforce reduction since it shed a quarter of its employees during the “year of efficiency” in 2022 and 2023, according to three people familiar with the matter who spoke to Reuters. Top executives have begun instructing senior leaders to plan for cuts affecting 20 percent or more of the company’s nearly 79,000 employees — approximately 16,000 people.

The reductions are driven by two concurrent pressures. Meta’s AI infrastructure spending has escalated sharply, with capital expenditures rising from $39 billion in 2024 to $72 billion in 2025 and now projected at $115 billion to $135 billion for 2026 — a near-tripling in two years. Simultaneously, CEO Mark Zuckerberg has signalled that AI tools are compressing the number of employees needed for existing work. “Projects that used to require big teams now be accomplished by a single very talented person,” he said in January.

The company has been spending aggressively to recruit AI talent for its Superintelligence Labs division, offering compensation packages worth hundreds of millions of dollars to individual researchers. It acquired Chinese AI agent startup Manus for at least $2 billion in December and social-networking platform Moltbook earlier this month. Meta invested $14.3 billion in Scale AI last year, bringing its founder Alexandr Wang in as chief AI officer.

The proposed cuts come against a backdrop of technical setbacks. Meta’s flagship Behemoth model — the largest in the Llama 4 family — was cancelled after missing performance targets. Its successor, a superintelligence model codenamed Avocado, was delayed from March to May after internal benchmarks showed it trailing Google, OpenAI, and Anthropic in reasoning and coding. Leaders within the AI division have discussed temporarily licensing Google’s Gemini to power some Meta products while Avocado catches up.

Meta spokesperson Andy Stone characterised the layoff reports as “speculative reporting about theoretical approaches.” Meta’s stock fell approximately 3.8 percent on the news, to around $614.

The company posted record quarterly net income of $22.8 billion in Q4 2025, on revenue of $59.9 billion. Full-year revenue exceeded $200 billion for the first time. CFO Susan Li told analysts the company remains “capacity constrained” and expects to deliver higher operating income in 2026 despite the infrastructure ramp.

Meta is not alone in coupling AI investment with workforce reduction. Amazon eliminated 16,000 corporate roles in January. Block, under Jack Dorsey, has cut roughly half its workforce. More than 45,000 tech jobs have been eliminated globally in early 2026, with over half linked to AI and automation.

The last time Meta restructured at this scale, it cut 11,000 jobs in November 2022 and another 10,000 four months later — moves that preceded a period of record profitability. Reality Labs, the division responsible for the company’s metaverse ambitions, has accumulated more than $73 billion in losses and already shed over 1,000 employees in January.

The Angle

The arithmetic here is unusually legible. Meta posted $200 billion in revenue, $22.8 billion in quarterly profit, and a 41 percent operating margin — and its response is to consider eliminating one in five employees. The move is not driven by financial distress. It is driven by a resource allocation decision: that the capital required to compete in AI is large enough to make the existing workforce a variable to be optimised rather than a constituency to be maintained. The company is not cutting because it cannot afford people. It is cutting because it has decided to afford something else.

What makes this structurally interesting is the gap between the confidence of the spending and the performance of the product. Meta has committed to $135 billion in capital expenditure for 2026 and up to $600 billion in data centre investment by 2028. It is offering individual researchers compensation packages measured in hundreds of millions. And the models those researchers are building — Behemoth cancelled, Avocado delayed and underperforming, Gemini licensing floated as a stopgap — have not yet justified the bet. The company is restructuring its entire cost base around a capability it does not yet possess. That is not unusual in technology. What is unusual is doing it at this scale while posting record earnings — which means the constraint is not money. It is conviction. Zuckerberg has decided this is the correct trade, and the company is being reshaped to execute it before the results are in.

The broader pattern across Big Tech — Amazon, Block, Meta, all cutting headcount while increasing AI spend — is clarifying something the industry has been careful not to state directly. The efficiency gains from AI that these companies cite as justification for the cuts are not a side effect of AI adoption. They are the product. The same tools these companies are building to sell are being used internally to reduce the number of people required to build them. Zuckerberg’s line about single talented individuals replacing whole teams is not a future projection. It is a present-tense description of what is already being used to determine who stays.

The 16,000 people whose roles are under discussion are not being replaced by better employees. They are being replaced by the inference output of systems that, by Meta’s own internal benchmarks, cannot yet match the competition. The company is cutting humans to fund machines that do not yet work as advertised — on the structural bet that they will, soon enough for the arithmetic to close.

A company posting record profits while preparing its largest-ever workforce reduction is not a contradiction. It is a statement about what the next decade values and what it does not.