February 2026, two researchers focused on artificial intelligence and labor relations told Business Insider that the "996" work culture is seeping from China's tech sector into Silicon Valley. Originally seen as a high‑intensity rhythm unique to Chinese internet companies—working 9 a.m. to 9 p.m., six days a week—it is now emerging inside U.S. AI firms in various altered forms: from implicit online on‑call expectations to output‑cycle‑driven performance assessments.
The scenes the researchers describe are concrete rather than abstract: AI teams, eager to gain an edge in model performance competitions, require engineers to extend training, tuning and testing windows, making nights and weekends become "normal" working time. Job postings that mention a "high‑intensity pace" or "rapid iteration" are often read by candidates as code for expected overtime. Online presence indicators on Slack and project management tools gradually become implicit metrics of loyalty and engagement.
They note that this culture can indeed increase short‑term output frequency: models are deployed faster, research papers are published more quickly, and product iterations accelerate. But the costs are both quantifiable and not. Direct consequences cited by interviewees include engineers' emotional exhaustion, sleep deprivation and professional burnout; indirect effects involve reduced creativity, impaired team collaboration, and weakened attention to safety and ethical considerations in decision‑making.
The article also outlines several concrete risks: first, increased staff turnover, as prolonged high pressure drives top talent to leave early or shift to more stable career paths; second, the sacrifice of internal learning and post‑mortem reviews, where short sprints replace systematic research, ultimately creating technical debt and reducing model transparency; third, under rapid release pressure, review gaps are more likely to be overlooked, raising the chance that products will introduce societal risks.
The researchers do not categorically oppose intense effort. They acknowledge that short‑term overtime can be reasonable at particular stages or to meet urgent needs. The key, they say, is boundaries and institutional safeguards: set clear limits on working hours, ensure paid time off and mental health support, include sustainability metrics in performance reviews, and have management lead by example in modeling a reasonable work pace. These measures aim to make "acceleration" a controlled strategy rather than a sacrifice of individual well‑being for fleeting speed.
For junior engineers, the article offers practical reminders: watch for implicit expectations behind job descriptions during interviews and onboarding; ask about typical team hours and vacation usage rates; practice boundary management at work by tracking hours and agreeing with supervisors on a sustainable pace. The researchers emphasize that investors and founders also bear responsibility—incorporating long‑term R&D capacity and talent retention into company valuation metrics.
Business Insider's report highlights a paradox: the AI field fundamentally requires deep thinking and long‑term accumulation, while a "996"‑style continuous high‑pressure approach suits repetitive output. The researchers' conclusion is clear: if Silicon Valley in 2026 continues to prioritize speed over humane management, short‑term gains may come at the cost of industry talent depletion and elevated ethical risk.