China’s AI boom is creating a different kind of entrepreneur
Rest of World Lin Zhang
China's gig workers are turning into one-person AI startups instead of employees. It's survival, not ambition — and it's quietly shaping how Chinese AI models get built.
Based on reporting by Rest of World, Lin Zhang — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
There's a word Chinese internet users started using around 2020 to describe a feeling everyone already had but couldn't name: neijuan, or involution. It's the sense of running harder and harder on a treadmill that never moves — competing so fiercely that all the extra effort just raises the bar for everybody else, with nobody actually getting ahead. Borrowed from an old anthropology term about colonial Java, it now describes something very modern: graduates with degrees and no jobs, a housing market that stopped being a ladder, and a generation that grew up believing hard work equals luck, only to find the equation broken.
Out of that anxiety a new kind of worker has emerged, one who isn't quite an employee and isn't quite a founder either. Using generative AI, a single person can now write copy, design graphics, edit video, run customer service, and manage an online shop — essentially becoming a company of one. Unlike the entrepreneurs of the mid-2010s who dreamed of becoming the next Jack Ma, this crowd has more modest aims: cover the rent, pay social insurance, keep some breathing room. They know platforms can cut off traffic overnight and that algorithms, not effort, often decide who eats. AI, in their hands, is less a rocket ship and more a survival kit.
That same logic of scarcity happens to be shaping China's AI industry itself. With U.S. chip export controls limiting access to top-tier hardware and capital markets more cautious than in Silicon Valley's free-spending years, companies like DeepSeek can't just throw more GPUs at every problem. Instead they've leaned into model compression, architectural tweaks, and open-source releases — squeezing more performance out of less. It's not simply cheaper engineering; it's a different philosophy of innovation, one built under pressure rather than abundance, closer to the resourceful improvisation known elsewhere as jugaad than to Silicon Valley's burn-money-first playbook.
The state has noticed the overlap. Local governments are now running pilot programs offering computing vouchers, discounted office space, and repurposed industrial parks to encourage exactly this kind of solo, AI-powered entrepreneurship — partly to soak up workers laid off by big tech firms, partly to build a domestic AI ecosystem from the ground up. It looks like empowerment, but it comes with strings: support that rises and falls with policy cycles, and risk that often lands not on venture capital but on family savings and parents' pensions when a shrinking real estate market has already hollowed out the middle class's safety net.
What's forming isn't a copy of Silicon Valley and it isn't the old catch-up model either. It's messier, still finding its shape, full of contradictions between freedom and precarity. But it suggests something worth sitting with: the next wave of useful AI innovation might not come from wherever capital is thickest, but from wherever the squeeze is tightest and people have no choice but to make less do more.
My take — AI-written commentary, not fact-checked reporting
I've said for a while that the 'more compute solves everything' story was always going to hit a ceiling somewhere, and China's chip constraints just forced that reckoning early. There's something almost bracing about watching efficiency become a survival trait instead of a marketing slogan — it's the opposite of the Western VC habit of mistaking spend for progress. I don't think this makes China's AI scene more humane, the family-as-safety-net part is genuinely brutal, but it does mean the next real breakthroughs in efficient, deployable AI are more likely to come from people who never had the luxury of wasting compute in the first place.
Read more about this at: Rest of World