Meta Just Axed Hundreds of Jobs to Fund a $135B AI Gamble — Here's What They're Actually Building

Meta cut several hundred jobs on Tuesday — across Reality Labs, Facebook, global operations, recruiting, and sales. This is the second round of layoffs in 2026, following January cuts. And it's just the beginning.

Reuters reported that Meta is considering layoffs that could affect up to 20% of its workforce — over 15,800 roles from its roughly 79,000 employees. Meta called the report "speculative" but didn't deny the workforce planning discussions. The company's actions tell a clearer story than its press statements.

The $135 Billion Question Nobody's Asking

Meta plans to spend between $115 billion and $135 billion on capital expenditure in 2026 alone. That's nearly double what it spent in 2025. Total spending guidance sits at $162-169 billion for the year.

Where's this money going? Data centers, Nvidia H100 and H200 GPUs, custom MTIA chips, and a $600 billion data center construction program running through 2028. Meta is building what it calls "Hyperion" — a data center complex nearly the size of Manhattan.

The company paused approximately 12 data centers in late 2022 to redesign them for GPU-dense, liquid-cooled AI workloads before resuming construction. Every new facility is AI-optimized from the ground up, achieving a Power Usage Effectiveness (PUE) of 1.08 — significantly better than industry average.

Reality Labs Takes the Heaviest Hit

Reality Labs has absorbed the deepest cuts across Meta's 2026 restructuring. The division — which oversees VR hardware and early-stage AR development — already lost 1,500 positions earlier this year. Resources are being reallocated from metaverse projects to AI research and development.

This is Zuckerberg's clearest signal yet: the metaverse bet is being downsized to fund the AI bet. After years of multi-billion dollar Reality Labs losses (over $50 billion since 2019), the company is shifting its infrastructure muscle toward Llama models, Superintelligence Labs, and Meta Compute.

Who's Actually Getting Cut

Data from the 2026 layoff rounds shows a pattern: mid-level management, quality assurance teams, customer support staff, and internal IT departments are the primary targets. These are roles that Meta's leadership apparently believes can be augmented or replaced by AI systems.

The tech industry has shed 59,000 jobs in 2026 so far, with Amazon, Meta, and Block among the largest contributors. Meta's cuts aren't unique — they reflect a sector-wide pattern of replacing human headcount with AI infrastructure spending.

What Meta Is Building With $135 Billion

The spending breaks down into several major initiatives:

Llama Model Ecosystem: Meta's open-source LLM family continues to grow. Llama models power everything from Instagram recommendations to WhatsApp business tools. The investment funds training runs for next-generation models that Meta gives away for free — a strategy designed to commoditize the AI model layer and keep Meta's proprietary data advantage intact.

Superintelligence Labs: Meta's research division focused on frontier AI capabilities. This team is racing to build AGI-capable systems, competing directly with OpenAI, Anthropic, and Google DeepMind.

Meta Compute: Announced in January 2026, Meta Compute formalizes the company's position as an AI infrastructure giant. It's not just building AI for Meta products — it's building compute capacity that could eventually be offered as a service.

Custom Silicon: The MTIA (Meta Training and Inference Accelerator) chip program aims to reduce dependence on Nvidia. While Meta remains Nvidia's largest customer, custom chips give Meta negotiating leverage and architectural control over its AI workloads.

The Math Doesn't Lie — But It's Uncomfortable

Meta generated $164 billion in revenue in 2025. Spending $135 billion on capex in 2026 means the company is betting nearly its entire annual revenue on AI infrastructure. The only way this works is if AI dramatically increases Meta's advertising revenue — which currently accounts for 97% of income.

Meta's argument: AI-powered ad targeting, content recommendations, and Llama-based business tools will generate returns that dwarf the infrastructure costs. The company points to AI-driven improvements in ad relevance that added $10+ billion in incremental revenue in 2025.

The counterargument: no company in history has successfully spent this much on infrastructure without either creating entirely new revenue streams or facing a painful correction. Meta is betting that AI is the new internet — and that being the infrastructure provider, not just a user, is where the value accrues.

What This Means for Everyone Else

If Meta follows through on 20% workforce cuts while doubling AI spending, it sends an unmistakable signal to the rest of the tech industry: human labor is being replaced by GPU racks at an unprecedented scale.

For developers: Meta's open-source Llama ecosystem means more powerful free models, but the company's massive compute advantage makes it harder for startups to compete on model quality.

For workers: The jobs being cut — middle management, QA, support, IT — are exactly the roles AI companies claim their products can automate. Meta is eating its own dog food.

For investors: Meta stock rose on the layoff news. Wall Street has decided that headcount is a cost to cut, not a capability to grow. This incentive structure will spread to every public tech company.

The Bottom Line

Meta is making the largest corporate bet on AI infrastructure in history. $135 billion in a single year. A potential 20% workforce reduction. Reality Labs gutted to fund GPU clusters. A data center nearly the size of Manhattan.

Either Mark Zuckerberg sees something the rest of us don't, or this will be studied in business schools as the most expensive pivot since IBM's mainframe-to-services transformation. The next 18 months will tell us which.