Wired published an investigative piece Thursday โ based on conversations with 16 current and former Meta employees โ confirming that Meta will cut about 10% of its workforce, roughly 8,000 people, on Wednesday May 20. The cuts add to the ~25,000 Meta has announced over the past four years and arrive against Q1 2026 profits of nearly $27 billion. The framing in the piece is the part worth pinning: record-high profits, record-low morale. Median total compensation at Meta fell to $388,200 last year from $417,400 in 2024, with the company cutting the equity portion of annual raises by 5% (2026) on top of last year's 10% reduction. Meta shares are off about 5% year-to-date. The perverse incentive shape Wired flags is concrete: layoff packages include a minimum of 16 weeks severance plus 18 months of paid health care, which has turned the upcoming cut into something people are openly hoping to be selected for. One Instagram employee quoted: "Everyone is just like, do it now, jesus fucking christ."
The novel mechanic in this story โ the part that sets it apart from the broader Big Tech layoffs narrative โ is the AI-training feedback loop. Wired confirms Meta installed corporate software to track employees' activity "solely in the name of training AI," with Meta's spokesperson responding that "there are safeguards in place to protect sensitive content, and the data is not used for any other purpose." One Meta policy staffer's quote captures the mood: "American employees being used to train the AI models that will replace them." This is structurally different from the Mercor data-labor story covered last week โ there, displaced workers were being recruited to train AI as gig labor; here, the same workforce is being instrumented in-situ as training data while being cut. The data-labor pipeline has moved inside the company doing the layoffs, which is a tighter loop with worse incentive geometry than the gig-marketplace version.
The institutional response is the part to track most carefully. UK Meta workers are now registering signatures to form a labor union with United Tech & Allied Workers, the UK-based tech labor group whose parent Communication Workers Union just unionized UK Google DeepMind employees earlier this month over concerns about military AI sales. The organizers' pitch to colleagues, per Wired: "Our leadership are escalating their cruel and short-sighted behaviours. We need to create an incentive for them to treat us with basic humanity." Tech-worker unionization in the UK is now hitting two of the largest AI-employing labs in two consecutive months โ Google DeepMind in early May, Meta in mid-May. If a third major lab follows by July, the pattern crosses from "isolated incidents" to "institutional response in formation," and the regulatory window opens up across the EU and UK. US workers have weaker labor-law leverage but the same incentive stack, and the equity-comp reductions at Meta are leading-indicator data points for how the wage compression will look elsewhere.
For builders deploying AI in regulated contexts, three implications worth tracking. First, if your AI product replaces internal job functions, the workforce-tracking-to-train-AI pattern is now a documented anti-pattern with reputational risk โ employees notice and organize. Second, the UK unionization trajectory means EU-style AI labor protections are arriving institutionally rather than legislatively first, which often produces faster contract changes than legislation. Third, the "record profits, record layoffs" combination is going to become the standard frame for AI-driven restructuring narratives โ Anthropic enterprise market-share gains (covered yesterday), Cisco's 20% earnings pop, Meta's $27B Q1 alongside its 10% workforce cut โ these data points compound into a coherent story that regulators and journalists are now positioned to use against the AI vendor stack. The pattern matters more than any individual data point; the Meta story is one anchor, and the next two months will determine whether it's an inflection or just another step on the curve.