Illustration by Charter · Photo by Roman_Gorielov, Getty

San Francisco’s tech scene is having a dangerous déjà vu moment.

The city has become “the ultimate 996 city” again: employees working 9am to 9pm, six days a week, chasing AI-driven riches. This isn’t just scrappy startups putting cots in offices. Google’s Sergey Brin recently sent a memo telling employees they need to “work at least 60 hours a week” in the office to accelerate Google’s AI efforts.

The pressure is very real: I’ve had conversations with many long-term tech employees in the Bay Area who are burning out. Those looking to move aren’t finding many appealing options: big tech is “depressing as hell” but as one friend noted “I guess I’ll stick with it versus hopping into the 996 and only having time with my kids on Sunday.”

Sign up for Charter's newsletter to get the handbook for the future of work delivered to your inbox.

We’ve seen this movie before. The late ’90s dotcom boom. Elon Musk’s “hardcore” Twitter takeover. In China, which Eric Schmidt says tech workers need to emulate, it’s been illegal since 2021. Such 80-hour weeks were standard during my own early consulting days in the “work hard, play hard” 1990s—when the partnership had 18 men and one woman.

The pattern always ends the same way: burnout, turnover, and diminishing returns.

When more becomes less

Here’s what Brin and other executives pushing extreme hours miss: presence isn’t productivity. More importantly, research consistently shows diminishing returns: working more than 40-50 hours a week might boost short-term productivity but on a prolonged basis leads to burnout, fatigue, decreased motivation, and impaired cognitive function. It’s not just bad for employee wellbeing, but for work quality. Beyond 50 hours, there’s a negative correlation with productivity.

Former Google Android engineering leader Chris DiBona pushed back on Brin’s memo, writing that “the best developers know that once they’ve been working singularly on a problem for four, five, six, eight hours in a day, their accuracy and productivity drop into the toilet and they’ll spend more time fixing their stress-induced screwups than creating good code.”

This isn’t theoretical. Gaming and software development have long used “crunch culture”—70 to 100 hour weeks before launches—but successful companies pair these sprints with recovery periods. The current AI gold rush is different: it’s demanding sustained extreme hours without the necessary downtime.

The real cost of the 996

While some companies are literally supporting this pace with “peptide Fridays,” the human cost is mounting. Burnout rates continue climbing, and we’re seeing the same talent exodus patterns that followed previous boom-and-bust cycles.

The executives driving these demands—many of whom genuinely “live to work”—are making a classic mistake, assuming what works for them works for everyone. They’re also forgetting a fundamental truth any CEO-athlete would know: you don’t win marathons by sprinting every mile.

Because that’s what AI development is: a marathon, not a sprint.

A better path forward

The companies that will win the AI race aren’t necessarily those working the most hours—they’re the ones working most effectively. This means:

  • Protect peak performance hours rather than demanding maximum hours. Blocks of focus time—no internal meetings allowed—boost productivity. DiBona insisted on a four-hour “no meetings” block for his teams every afternoon. Bonus points: turn off phone and computer notifications.
  • Build sustainable rhythms that include both intense focus periods and recovery time. After a demanding month before a launch, a week of downtime helps with recovery and retention of employees.
  • Focus on personal growth, not just rest. Research by Nick Petrie, a leadership development coach, found that shifting from consistent “high performance” mode to other challenges and developing new skills reduced burnout. Leaders need to rotate staff to enable growth —sometimes at the expense of short-term optimization.

The irony of the current moment is that the very technology teams practicing 996 hours are building—AI that can automate routine tasks and enhance human creativity—should reduce the need for extreme hours, not increase it.

San Francisco may be reliving its 996 moment, but the smartest companies will learn from history instead of repeating it.

Read more from Charter

The handbook for the future of work, delivered to your inbox.

Subscribe
EDIT POST