I keep thinking about the moment I read that Silicon Valley startups were posting “996 required” in job listings. It was August 2025. A 36kr article tracked how Chinese overtime culture had jumped the Pacific. Rilla, an AI startup, advertised seventy-hour weeks with free Saturday lunches. Another founder said the first two years of a startup left “almost no option besides 996.” Reddit threads filled with American engineers calling it “modern slavery.” One commenter noted, with bitter accuracy, that China’s own Supreme Court had already ruled 996 illegal.
The irony is almost too neat. The work rhythm that defined China’s tech boom, the one that left young people in Shenzhen and Hangzhou with cervical spondylosis at twenty-six, is now being repackaged as hustle gospel for San Francisco. Ramp, a fintech company, used corporate card data to prove it: Saturday spending in San Francisco tech offices spiked in 2025, a pattern that did not exist in 2024 and did not appear in New York. The data showed lunch orders, team dinners, office supplies. The kind of small transactions that add up to a culture shift.
I noticed something else in the numbers. While Silicon Valley investors like Mike Moritz of Sequoia praised Chinese workers for “six to seven days a week, fourteen hours a day,” the Chinese internet industry itself was quietly shrinking. A 36kr analysis of ten major Chinese tech companies found that Alibaba cut 66,000 jobs in recent years. Baidu and NetEase kept trimming. The headline said the money was flowing “from people to machines.” Tencent still paid an average of 1.13 million yuan per employee, but AI hiring was the only category growing. Everything else contracted.
The delivery rider story keeps coming back to me too. In February 2025, JD.com announced it would finally pay social security for its full-time delivery workers. Meituan and Ele.me followed within days. For an industry born in 2009, this was sixteen years of waiting. The articles treated it as breakthrough news, which it was. But the fine print was more honest. Most riders are classified as independent contractors, not employees. Over 80 percent sit in a legal gray zone created by platform algorithms. A single platform would need 17.3 billion yuan annually to cover all active riders, more than its entire yearly profit. Riders themselves often prefer the rural health cooperative, which costs 400 yuan per year, over urban social security that deducts 700 yuan monthly from paychecks they may not keep for fifteen years. The average rider stays on the job for 9.7 months. Turnover exceeds 30 percent annually.
What strikes me is how these stories layer on top of each other. The same month that JD.com promised riders social security, it also launched its own外卖 platform and declared war on Meituan. By July, the three-way battle between JD, Meituan, and Alibaba had burned through an estimated 80 billion yuan in subsidies. Meituan’s Wang Puzhong told LatePost that most of it was foam. He described visiting a restaurant in Nanjing and finding twelve of the day’s first orders were for twelve-packs of bottled water, bought at prices that made no economic sense. In Suqian, a city in Jiangsu,外卖 volume quadrupled. But nobody was eating out anymore. A bowl of noodles cost 18 yuan in the dining room and 6 yuan through the app. The price distortion was, in his word, unsustainable.
Meanwhile Pinduoduo turned ten in December 2025 and barely made a sound. No founder letter. No AI announcements. Employees received a hoodie, a commemorative coin, and blind-box chicken mascots. The company holds 423.8 billion yuan in cash, more than almost any other Chinese internet firm, and chooses to wait. It did not join the外卖 war. It did not announce large AI investments. Its cross-border platform Temu pushed toward 90 billion dollars in annual sales across nearly one hundred countries, but the headquarters mood was described as quieter than before. Some employees no longer worked the legendary intensity. Pay raises slowed. The social media ban loosened. The article compared founder Colin Huang to Warren Buffett, gathering cash in a downturn.
I do not know what to make of the fact that China’s most famous work culture is now being admired abroad at the exact moment Chinese companies are trying to move past it. Maybe that is how cultural exports work. The thing you are trying to leave behind becomes the thing others want to import. Or maybe it is simpler: when growth slows, the mythology of hustle becomes more valuable than the reality. You sell the story of overtime to investors while your own algorithms replace the overtime workers.
The Ramp data showed something small but real. Saturday office spending in San Francisco. Free lunches on the sixth day. A few thousand people, maybe tens of thousands, borrowing a rhythm that millions in China have already burned through. I wonder if they know what comes next. The体检 reports that grow longer each year. The cervical vertebrae and thyroid nodules. The father in the hospital you cannot visit because you are debugging a model. The 35-year-old who discovers that skills accumulated over a decade depreciate faster than the hardware they run on.
One of the researchers interviewed thirty Chinese大厂 youth for a study published in China Youth Research. She called their labor market a “Mobius strip,” a surface with no clear inside or outside, no boundary between core and edge, no finish line to the performance competition. The workers themselves used two different phrases for the same job: “致富梦,” the dream of getting rich, and “青春饭,” a meal you eat only while you are young. Both described the same work. Both were true.
I keep thinking about that. How the same job can be a dream and a countdown at the same time. How the same culture can be exported as inspiration while it is being automated away at home. How a delivery rider’s social security can be both a historic breakthrough and a math problem nobody wants to solve.
The补贴 wars will end. The泡沫 will deflate. Someone will calculate whether the Saturday lunches were worth it. I am not sure the answer will change anything.