A work culture once synonymous with the grueling tech industry grind of China is now finding fertile ground in the heart of Silicon Valley. The so-called “996” work schedule — working from 9am to 9pm, six days a week — is reportedly gaining traction among AI researchers and engineers at some of the most prominent technology companies in the United States, raising serious questions about sustainability, mental health, and what the race to build artificial general intelligence is actually costing the people driving it.
What Is ‘996’ and Where Did It Come From?
The term “996” originated in China, where it became notorious as a de facto standard at major tech firms like Alibaba, ByteDance, and Huawei. The practice drew international criticism and even sparked a grassroots protest movement among Chinese developers, who created an open-source repository called “996.ICU” — a dark reference to the intensive care unit, implying that working such hours would land you in one. Chinese courts and labor regulators eventually moved to crack down on the practice, declaring it illegal under existing labor laws.
Now, according to a report from Business Insider, echoes of that same culture are emerging in Silicon Valley, particularly within AI-focused teams where the competitive pressure to ship models, secure funding, and outpace rivals has created an environment where extreme hours are increasingly normalized — and in some cases, implicitly expected.
The AI Arms Race Is Driving the Burnout
The timing is hardly a coincidence. The AI industry is currently in a period of intense, almost unprecedented competitive pressure. Billions of dollars are flowing into foundation model development, inference infrastructure, and AI applications. Every week brings a new model release, a new benchmark record, or a new funding round that reshapes the competitive landscape. In that environment, the pressure to work longer and harder is structural, not just cultural.
This connects directly to a broader ideological push within certain corners of the AI industry. AI scaling has been described by some researchers as something that “must be pushed to the maximum” — a mindset that, when applied to human workers as well as compute resources, begins to look less like ambition and more like exploitation. When the dominant philosophy frames progress as something that must happen as fast as physically possible, the people doing that work inevitably bear the cost.
Researchers and engineers working on large language models, reinforcement learning systems, and multimodal AI are reportedly logging consistent 70 to 80 hour weeks, with some describing weekends as functionally non-existent during critical development cycles. The culture is particularly acute at well-funded startups where equity incentives and mission-driven rhetoric are used to justify the demands placed on staff.
The Human Cost of Moving Fast
Burnout as a Feature, Not a Bug
What makes the 996 dynamic particularly insidious in the AI context is that it is often dressed up in the language of purpose. Employees are told they are working on technology that will change the world — cure diseases, solve climate change, transform education. That framing makes it psychologically difficult to push back on unreasonable expectations, because doing so can feel like opting out of something important. The mission becomes a mechanism of control.
Mental health professionals and labor researchers have long warned about the dangers of this dynamic. Chronic overwork is associated with significantly elevated risks of cardiovascular disease, depression, anxiety, and cognitive decline — the very cognitive capacities that AI researchers need most. There is also a well-documented relationship between exhaustion and error rates, which in the context of AI safety research carries implications that extend well beyond individual wellbeing.
Who Bears the Burden?
It is also worth asking who, specifically, is most exposed to these pressures. Junior engineers, researchers without tenure or established reputations, and employees on visa sponsorships — groups with less structural power to refuse unreasonable demands — are often the ones absorbing the worst of these cultures. This has implications not just for individual welfare but for the diversity and long-term health of the AI talent pipeline. As major tech companies simultaneously embrace AI and reduce their broader workforces, the pressure on remaining AI-focused employees only intensifies.
What This Means
For workers currently in or considering roles at AI companies, the normalization of 996-style expectations represents a genuine and pressing concern. Unlike previous tech boom cycles, the AI race shows little sign of decelerating, meaning that the conditions producing these work culture patterns are unlikely to self-correct without deliberate intervention — either from company leadership, organized labor, or regulators.
Policymakers are beginning to pay closer attention to the AI industry more broadly. US senators have been taking steps to educate themselves about AI ahead of potential regulation, though labor conditions within AI firms have not yet emerged as a primary legislative focus. That may change as burnout, attrition, and the mental health consequences of extreme work cultures become harder to ignore.
For the companies themselves, the calculus is more complicated than it might appear. Burning out top talent is expensive — recruiting, onboarding, and bringing new researchers up to speed on complex systems takes months. If the 996 culture drives away experienced engineers or degrades the quality of their output, the short-term velocity gains may come at a significant long-term cost to both product quality and, critically, AI safety.
And for the broader public, it raises a quiet but important question: if the people building the most powerful AI systems in history are exhausted, overworked, and operating under extreme pressure, what does that mean for the quality and safety of what they’re building? Experts have consistently emphasized that understanding the conditions under which AI is developed matters just as much as understanding the technology itself.
Key Takeaways
- The ‘996’ work culture — 9am to 9pm, six days a week — that became infamous in China’s tech industry is now reportedly taking hold among AI researchers and engineers in Silicon Valley.
- Competitive pressure in the AI industry is a primary driver, with teams under intense pressure to ship models and outpace rivals, creating environments where extreme hours are normalized and often implicitly expected.
- The human cost is real and compounding — chronic overwork carries documented risks to physical and mental health, and the cognitive degradation associated with burnout has direct implications for AI safety and output quality.
- Regulatory and structural intervention may be necessary to address these conditions, as market incentives alone are unlikely to reverse a culture that has become embedded in the competitive logic of the AI arms race.
