On an internal company forum, a Meta employee posted a meme from the television series “The Office” with a caption that read: “0 days since our last nonsense.” The image, a wry tally of dysfunction, surfaced in June after the company suspended a programme called the Model Capability Initiative. That initiative, rolled out in April, had captured the clicks, keystrokes and browsing activity of US-based staff to train AI agents. More than 1,600 employees signed a petition demanding its end, some describing the parent of Facebook, Instagram and WhatsApp as a “data extraction factory.” The programme was paused when private conversations and performance metrics inadvertently became visible to all staff, a breach that also risked scrutiny from European regulators because it swept up exchanges between colleagues on both sides of the Atlantic.

The surveillance experiment unfolded against a backdrop of relentless cost-cutting. This spring, Meta eliminated roughly 8,000 roles—about 10 percent of its workforce—and reassigned another 7,000 employees to AI projects. Founder and chief executive Mark Zuckerberg, who holds near-absolute voting control, has framed the cuts as a necessary trade-off to fund a capital expenditure bill of up to $145 billion this year, nearly double the previous year’s figure. Days after the layoff notices went out in waves across Asia, Europe and the Americas, Zuckerberg argued in a public interview that fears of AI-driven job losses were overstated, suggesting that tools which empower individuals could create more employment than automation destroys. Inside the company, however, the mood was captured by the anonymous complaints of “mind-numbing” tasks and a pervasive fear of the next restructuring.

Meta’s AI push is, by its own admission, a game of catch-up. Viewed from Silicon Valley, the company has lagged behind Google, OpenAI and Anthropic in producing cutting-edge models. To close the gap, Zuckerberg invested over $14 billion last year in the San Francisco startup Scale AI and installed its 28-year-old chief executive, Alexandr Wang, to run a new “superintelligence” lab. The move triggered an exodus of research talent, most notably Yann LeCun, a Turing Award winner often called a godfather of modern AI, who had led Meta’s AI research since 2013. LeCun, suddenly reporting to a manager more than three decades his junior, left at the end of 2025 and later told the Financial Times that Wang had “no experience with research” and was pursuing a “dead end.” The stakes extend beyond social networks: Meta is doubling down on smart glasses and exploring a prediction-market app, even as its legal challenges multiply.

This human toll is not confined to Menlo Park. A review of regulatory filings by Business Insider shows that some of the world’s richest technology companies are spending billions simply to part ways with employees. Amazon recorded an estimated $2.7 billion in severance costs through 2025, while Oracle and Intel each disclosed $1.8 billion in similar charges. Intel’s layoffs, part of a turnaround effort, cut over 25,000 jobs at an average cost of more than $70,000 per person. Oracle’s headcount fell by 21,000 in a single year as it redirected resources toward cloud and AI infrastructure. Meta itself did not break out severance expenses, but the pattern is unmistakable: the industry’s AI pivot is being financed in part by the departure of tens of thousands of workers.

For now, the Model Capability Initiative remains suspended while Meta investigates. A spokesperson said the programme was designed with privacy safeguards and that there was no indication data had been improperly accessed. Yet the pause leaves an unresolved question hanging over the company’s open-plan campuses, where the meme’s deadpan clock has, for a moment, stopped ticking.