
As of June 2025, the number of one-person companies in China had passed 16 million, 27.4% of all enterprises. New registrations were growing at 47%; in other words, one in every three newly registered companies was a one-person company.
16 million is roughly the population of all of Australia running their own one-person companies in China.
That's no accident; it's a shift in the startup paradigm that's happening right now.
OPC stands for One Person Company (一人公司 in Chinese). For overseas readers: in China, OPC has become both a legal form (a limited liability company with a single shareholder) and a policy buzzword, with local governments rolling out programs to support AI-powered solo founders.
But if you still think of it as "one person doing everything and working themselves to death", that's out of date.
The 2026 OPC is essentially a "1 founder + N AI digital employees" model. You handle strategic decisions, creative judgment and core value creation; AI handles standardized, repetitive work like code generation, content creation, data operations and customer service.
Lian Yunbo, head of the AI Center of Excellence at the Cloud Computing and Big Data Research Institute of the China Academy of Information and Communications Technology (CAICT), gives a precise definition:
"The rise of OPCs in the AI era isn't fundamentally about shrinking headcount; it's about restructuring the smallest production unit of a business. It's a lightweight organizational model of 'one core decision-maker + multiple digital employees + platform resources'."
So the key isn't "one person"; it's "one person + AI computing power".
Many people see an OPC as an "upgraded sole proprietorship" (个体户, an individual business in China), but the two are fundamentally different:
| Dimension | Traditional sole proprietorship | OPC (one-person company) |
|---|---|---|
| Core dependency | Human labor, with a ceiling on productivity | AI agent systems that can scale without limit |
| Legal status | Not a separate legal entity, unlimited liability | Separate legal entity, limited liability protection |
| Operating model | One person does everything | One person directs an AI army |
| Growth logic | Linear: another hour, another unit of income | Leveraged: content/code/AI systems enable nonlinear growth |
| Core moat | Diligence and stamina | Depth of insight + ability to work with AI |
| Financial risk | Personal assets not separated from business risk | Separate assets, isolated risk (if compliant) |
Simply put, a sole proprietor is "one person fighting alone"; an OPC is "one person commanding an army".
In one phrase: the barriers came down.
Five years ago, if you wanted to build a SaaS, you had to code or have money to hire people who could; to do content, you had to write, edit and run it all yourself; to sell cross-border, you needed English, overseas channels and supply-chain resources.
Now? AI can write the code, AI can generate the content, translation across languages happens in real time, and AI agents can do a first pass on customer inquiries.
The 2026 One-Person Company Insight Report published by the "Honghu Hui" (鸿鹄汇) community in Hangzhou proposed a key metric, HACR (human-AI cost ratio): every ¥1 spent on AI can effectively replace about ¥72 of development labor costs.
¥1 doing the work of ¥72 isn't cost-saving anymore; it's structural change.
Entrepreneurship has shifted from "capital-intensive" to "insight-intensive": you don't need to raise millions and hire dozens of people; you just need deep enough insight into a field, and the ability to use AI to turn that insight into products, services and revenue.
Case 1: An AI grading product, 15,000 users in 4 months
Li Yunfan, born in 1998, joined the "Honghu Hui" community when it was founded. He built "Zuowen Shuo" (作文说), an AI essay-grading product. AI did 80% of the programming, while he was strategist, product manager and chief operating officer rolled into one. It passed 15,000 users within 4 months of launch.
Case 2: An architecture firm as a one-person company, with margins up 20%–30%
Wang Kaili of Ningbo Wuzhongshengyou Architectural Design Co., Ltd. turned his architecture firm into a one-person company, juggling the roles of designer, business development, operations and social media. Compared with when he led a team, he cut management and staffing costs, and his profit margin actually rose by 20% to 30%.
Case 3: A Tsinghua PhD's AI-assisted startup
Xu Zhanwei, a Tsinghua PhD born after 1995, registered Shenzhen Shouyi Computer Co., Ltd. and uses AI for expense reimbursement, logo design and more, focusing money and people on what matters most: product R&D.
These cases share one trait: not going it alone, but one person orchestrating all the resources.
OPCs have their advantages, but there's one red line you must never cross: commingling assets.
Article 23, Paragraph 3 of China's revised Company Law states: if the shareholder of a one-person company cannot prove that the company's assets are independent of their personal assets, they bear joint and several liability for the company's debts.
This is called "reversal of the burden of proof", which means:
This is no joke. In judicial practice, most OPC shareholders who end up with joint liability do so because of commingled assets.
How to avoid it? Three things are essential:
Remember: an OPC's limited liability protection isn't "set and forget"; it rests on compliant operation. If you can't keep assets independent, the limited liability is an illusion.
Hangzhou currently has China's most mature OPC ecosystem:
Other places are following: Guangdong, Shanghai, Chengdu and others have all issued support policies for AI-driven OPCs.
If OPCs appeal to you, here's an actionable path:
Step 1: Find your "insight leverage"
The core of an OPC isn't technology; it's your deep insight into a field. Cross-border e-commerce, content operations, finance and tax, design... wherever your understanding beats more than 80% of people, that's your leverage.
Step 2: Build an AI system, don't just "use AI"
The difference: using AI means occasionally asking ChatGPT; building an AI system means automating content production, customer service, data analysis and operations with AI agents. You don't manage "the work"; you manage "the system".
Step 3: Validate with an MVP: sell first, build later
Don't develop in isolation for six months. Share your idea on social media first, find a minimal audience and validate willingness to pay. Once people are willing to pay, invest in building the product.
Step 4: Compliance first
Register a one-person company, open a separate corporate bank account and get an audit every year. It's a slower path, but a steadier one.
An OPC isn't a cure-all; it has a ceiling: limited scale, narrow fundraising options and one person carrying the pressure of every decision. But if what you're after isn't valuation and scale but freedom and profit, an OPC may be the best option this era offers ordinary people.
Future competition won't be about how many employees you have, but how many AI agents you can direct.
One person can be an army.
Sources: People's Daily, 36Kr, Sina Finance, Alibaba Cloud Developer Community, Gaopeng Law Firm (all in Chinese)