Meta Is Firing 8,000 People to Pay for AI. Zuckerberg Is Not Apologizing.

Correction (October 9, 2026): An earlier version of this article attributed the layoff memo to Mark Zuckerberg; it was sent by chief people officer Janelle Gale, and Zuckerberg's remarks came at an internal meeting. It also said Meta raised its capex guidance on the day of the layoffs (it did so on April 29), described the jump from 2025 capex as 73 percent (the top of the range is roughly double), gave Q1 2026 revenue growth as 16 percent (it was 33 percent), and included unsourced claims about Llama 5, a specific August layoff round and Meta's Workplace product. Those have been corrected or removed.

Meta started laying off approximately 8,000 employees today — roughly 10 percent of its global workforce — weeks after raising its 2026 capital expenditure guidance to between $125 and $145 billion, up from $115 to $135 billion. The company is also dropping plans to fill 6,000 open roles. Chief people officer Janelle Gale's memo described the cuts as an effort to run the company more efficiently and "offset the other investments we're making." At an internal meeting reported by The Wall Street Journal, Zuckerberg named "compute and infrastructure" and "people oriented things" as Meta's main cost drivers. The connection is not hard to read: the AI infrastructure bill is enormous, and the math demands something be cut. That something is people.

This is not Meta's first round of 2026 cuts. Layoffs already hit Reality Labs in January and several hundred roles in March, and leaders have told staff more cuts are possible later in the year. What is different today is the scale and the candor. Zuckerberg told employees the company needs to get "somewhat" smaller — reading less like a pivot in strategy than a deliberate reallocation of capital from salaries to servers. Meta's 2025 capital expenditures were already $72.2 billion. The 2026 guidance tops out at $145 billion. Roughly doubling compute spending needs to be funded somewhere.

What $145 Billion Buys

The capex surge is not going toward one thing. Meta said the latest increase reflects mostly higher component pricing, along with additional data center costs, but the overall budget is going toward everything at once: NVIDIA H100 and Blackwell GPU clusters, Meta's in-house AI accelerator chips (MTIA), data center construction across the United States and Europe, and the power infrastructure those facilities require. Meta is also training its own models while running inference for Meta AI across WhatsApp, Instagram, Messenger, and Facebook. At that scale, compute is not a line item. It is the business.

The math explains the squeeze. Every dollar spent on salaries is a dollar not spent on a GPU cluster that might determine whether Meta's AI assistant is competitive with ChatGPT in two years. Zuckerberg has watched Google commit to an equivalent capex pace, has watched Amazon build out AWS AI services at scale, and has watched Microsoft deploy Copilot into its enterprise installed base. The bet Meta is making is that raw infrastructure investment, combined with open-weight Llama models and product distribution across four apps with billions of users, is the correct response. That bet requires an enormous amount of capital. It also requires making hard choices about headcount.

The Uncomfortable Arithmetic

There is a version of this story where the layoffs are a failure — a signal that AI is not delivering returns fast enough to cover its costs. That is not what the evidence shows. Meta's revenue grew 33 percent year-over-year in Q1 2026, to $56.3 billion, on the back of its ad business. The money machine is working. The issue is that it is working well enough to justify spending even more, which in turn requires finding the capital from somewhere within the organization.

Eight thousand people losing their jobs so that Meta can buy more Blackwell chips is a brutal headline, but it reflects a logic that is now standard across big tech. Microsoft did versions of this. Google did versions of this. The difference with Meta is that Zuckerberg said the quiet part out loud. He told his employees that compute and people are the company's biggest costs, and that the company needs to shrink. Whether that honesty is admirable or alarming depends on which side of the severance package you are on.

What This Signals for the Industry

Meta's announcement today is a data point in a broader trend that is accelerating: the large-scale displacement of knowledge workers at AI companies, not by AI directly, but by the investment decisions AI forces. When a company decides to double its compute budget, it is also implicitly deciding that the value created by additional headcount is lower than the value created by additional compute. That calculus is being made right now at every major technology company, and Meta is just the most explicit about it.

The companies that will come out of this period strongest are not necessarily the ones with the most employees or even the most revenue. They are the ones with the most efficient inference infrastructure and the distribution to monetize it at scale. Meta has both. The 8,000 people starting their job searches today are the cost of that position.

Sources: Meta Q1 2026 results · Meta Q4 2025 results (SEC) · Forbes Australia · Yahoo Finance