Meta Lays Off 8,000 Employees, Bets on AI Superintelligence
Meta launched a new round of global layoffs on May 20, 2026, cutting approximately 8,000 positions, which accounts for about 10% of its total workforce. This is the largest structural adjustment since 2026, with layoffs starting in three waves from 4 AM local time.
This adjustment also cancels about 6,000 planned recruitment positions, while another 7,000 employees are reassigned to AI-related work lines. The affected areas include the Integrity team, cybersecurity team, and content design department. U.S. employees will receive a severance package of 16 weeks' pay plus an additional 2 weeks for each year of service. Reports indicate that the median annual salary for employees has decreased by about $30,000 due to tightened raises.
Financially, Meta reported a revenue of $56.3 billion in Q1 2026, a 33% year-on-year increase, marking the largest single-quarter growth in nearly five years, with a net profit of $26.7 billion. Meanwhile, the company has raised its capital expenditure budget for 2026 to a range of $125 billion to $145 billion, more than doubling from $72 billion in 2025, with funds clearly directed towards the development of "personal superintelligence" technology and data center construction.
Prior to this round of layoffs affecting 8,000 people, Meta had already cut hundreds of positions in departments like Reality Labs and Facebook from January to March this year. Including this adjustment, the total number of layoffs since 2022 has exceeded 30,000. Internally, over 1,500 employees had previously signed a petition opposing the company's use of employee data to train AI models that could replace their jobs, and employee morale has reportedly continued to decline.
Zuckerberg stated in an internal memo that "AI is the most decisive technology of this era, and success is not guaranteed," positioning this adjustment as a transformation of the company rather than merely a cost-cutting measure. According to Goldman Sachs, layoffs across the industry due to AI-related restructuring have reached over 16,000 per month in 2026, with Cisco also announcing a cut of about 4,000 employees, indicating that such adjustments are not isolated within the tech industry.
From a financial perspective, the market interprets this round of layoffs as a dual signal of "cost-cutting" and "betting on AI"—the positions eliminated are non-core AI roles in auditing, security, and content design, with the freed-up salary budget being directly redirected towards superintelligence research and computational infrastructure represented by the Hyperion data center co-built with Blue Owl Capital. On the day of the announcement, Meta's stock price only rose by 0.1%, indicating that investors had already priced in expectations for AI transformation, and there was no panic selling due to the layoffs themselves, but rather a focus on whether the expansion in capital expenditure can deliver long-term AI competitiveness.
Supplementary data: The Hyperion data center project aims to reach an initial capacity of 2 gigawatts by early 2030, making it one of Meta's largest publicly disclosed data center investment projects.
Source: Public Information
ABAB AI Insight
Zuckerberg's layoff strategy has repeated the same set of actions at least three times: a large-scale layoff of 11,000 people in November 2022, another cut of 10,000 in 2023 as part of the "efficiency year", and in January 2025, a precise reduction of about 3,600 people (about 5%) under the guise of "performance"—the "Met Some Expectations" rating was directly linked to the layoff decisions. Before each round of layoffs, the company adjusted the rating distribution—by the end of 2022, the proportion of the lowest rating increased from 7%-12% to 16.5%, and the mid-2025 review target was further raised to 15%-20%. The mechanism of "first adjusting ratings, then laying off based on that" has been repeatedly used as a fixed mechanism rather than a one-time decision.
The teams affected in this round are concentrated in auditing, security, and content design—these have been cost centers that expanded continuously over the past three years due to regulatory pressures and content governance needs. The saved labor budget was announced almost simultaneously with the increase in capital expenditure for 2026 (from $72 billion to $125 billion to $145 billion), with funds clearly directed towards the Hyperion data center and the development of "personal superintelligence". This is a typical capital reallocation action of "cutting cost centers while expanding infrastructure", motivated by the need to seize the computational power window for large models and avoid falling behind competitors like OpenAI, Google, and Anthropic.
This path is highly similar to Amazon's layoffs of about 18,000 corporate roles during its "efficiency year" from 2022 to 2023, redirecting budgets towards AWS and warehouse automation; it is also akin to IBM's repeated structural layoffs over the past decade using "performance re-ranking" rather than directly announcing strategic contractions. Meta is currently in a phase closer to "high-intensity capital expansion" rather than mere contraction—while the total number of employees has decreased to about 71,000, capital expenditure has reached a historical high, representing an expansion-type restructuring of "reducing headcount while increasing computational power".
Essentially, this reflects a combination of "capital concentration" and "transfer of pricing power": the cost structure previously dominated by labor-intensive functions like auditing, trust, and security is being replaced by a capital-intensive structure dominated by computational power and model capabilities. Mechanistically, as AI models can take on some auditing, content generation, and security monitoring tasks, the marginal output of these roles is replaced by computational power, leading companies to naturally shift budgets from "labor unit costs" to "computational power unit costs"; once this shift crosses a critical scale (like Meta's hundreds of billions in capital expenditure), it will inversely suppress the bargaining power of labor in the same industry, creating a chain reaction—Cisco's concurrent layoff of about 4,000 people serves as evidence.
ABAB News · Cognitive Laws
Labor is a replaceable cost; computational power is the asset being bet on.
First rewrite the assessment criteria to justify layoffs.
Employees are shrinking, while capital expenditure is expanding—this is true expansion.