Silicon Valley tech firms developing advanced artificial intelligence are promoting shorter workweeks while subjecting their engineers to grueling 90-hour schedules. According to former and current employees at firms like OpenAI and Meta, surging workloads, mandatory internal reassignments, and relentless output demands are expanding traditional office hours rather than contracting them.
The Executive Promise Versus The 90-Hour Reality on the Ground
For years, executives driving multi-billion-dollar investments into artificial intelligence have anchored their public relations campaigns around a compelling narrative: advanced software will eventually shoulder the cognitive load. Four years ago, a Google engineering director predicted a four-day workweek by 2025. Earlier this year, OpenAI formally challenged corporate peers to trial a four-day workweek without salary cuts, arguing that accelerated productivity makes workforce restructuring inevitable.
But the balance sheet tells a different story. A former technical employee at OpenAI revealed that the firm never instituted its own recommended four-day work schedule. Instead, internal culture leaned heavily on aggressive crisis meetings, mandatory weekend shifts, and ruthless performance reviews that triggered sudden terminations. According to reports gathered by the BBC, engineers operating within OpenAI and Anthropic frequently endure development cycles pushing past 90 hours across a standard seven-day period.
Here is the financial friction point: efficiency gains do not translate to resting capital. When Anthropic boasts that its chatbot Claude operates autonomously for seven-hour stretches, or when Meta CEO Mark Zuckerberg champions lean teams accomplishing more than they ever could have, the corporate response is rarely to scale back human hours. It is to scale up expectations.
The Bottom Line
- Productivity Paradox: Research from the University of California, Berkeley demonstrates that software engineers using artificial intelligence accelerate their output only to absorb a broader array of validation and maintenance tasks.
- Labor Arbitrage: While Meta reduced headcount by cutting 1 in every 10 employees, remaining personnel report being drafted into high-urgency AI projects with zero right of refusal.
- Operational Burnout: Ex-employees at major technology providers note that infrastructure shifts toward AI leave technical debt and backend errors that demand late-night manual intervention.
Forced Reassignments and the Collapse of Work-Life Boundaries
The internal pressure cooker extends far beyond routine product launches. At Meta, employees describe being abruptly drafted into urgent artificial intelligence units through what insiders term “forced recruitment.” Staff members are stripped of their prior project allocations and reassigned to build software engineering assistants and model-testing architectures. Refusing the transition forces an immediate resignation.
These specialized units operate under continuous strain. Staffers log deep into the night and across weekends, maintaining an unyielding state of digital readiness. Unlike European counterparts restricted by statutory 48-hour weekly limits under regional labor rules, American technology hubs operate in an unregulated legal framework regarding maximum work hours for those over 16.
The administrative bleed affects non-AI teams as well. Former Google engineer Amin Shali departed the firm after corporate resource reallocation—directing critical memory storage and processing units away from legacy products and toward artificial intelligence infrastructure—triggered constant engineering failures that required overnight crisis management.
Empirical Evidence Confirms the Expanding Workday
Economic theory suggests that technological automation should compress human labor requirements. Empirical field research disrupts that assumption. An eight-month observational study conducted by the University of California, Berkeley tracking technology workers deploying AI tools found that participants accelerated their operational tempo, assumed wider task scopes, and systematically lengthened their daily hours.
| Technology Firm | Reported Executive Stance | Observed Employee Work Impact |
|---|---|---|
| OpenAI | Advocated for external 4-day workweek trials. | Reported 70 to 90-hour workweeks during intensive product sprints. |
| Meta | Stated smaller teams achieve more output via AI. | Enforced mandatory reassignments to AI units with “on-call” weekend expectations and 10% workforce reductions. |
| Anthropic | Highlighted autonomous 7-hour continuous AI task execution. | Maintained intense development cycles exceeding standard white-collar hour baselines. |
| Pushed heavy infrastructure investments into machine learning models. | Engineers reported increased night shifts handling legacy system failures due to diverted computing resources. |
Workers take on supplementary validation tasks either out of professional ambition or survival anxiety in volatile labor markets.
