A one-person company looks like being your own boss. On the books it looks more like a subcontractor

For the past two years I've been building myself a system for writing a public account. Topic selection, sourcing, drafting, images, pushing the draft — most of the steps run on AI, and I keep the judgement and the final check for myself. It really can do the work of a small editorial team.

One night last month, staring at the dashboard, I had the thought again: if one person can run this thing, maybe I could take it one step further and treat it as a real business.

Behind that thought is a phrase that's been everywhere lately: the one-person company.

A one-person company — One Person Company — means one person using AI tools to run an entire business chain themselves, from product to build to sales, hiring nobody, or almost nobody.

The reason it keeps coming up is that the cost structure genuinely changed. One estimate puts the annual cost of solo founding at somewhere around $18,000, down from over $200,000, with a tool stack running a couple of hundred a month — about one-twentieth of hiring a person. OpenAI's Sam Altman has been predicting since 2024 that the first billion-dollar company with a single founder is close; he later revised that to "dangerously close."

Domestic numbers point the same way. One statistic has the share of companies started by a solo founder rising from 23.7% in 2019 to 36.3% in 2025. Places like Lin-gang in Shanghai and Zhongguancun in Beijing have rolled out dedicated support policies for one-person companies.

So this isn't just a concept. Tools, costs and policy are all paving the road for "one person can run a company too."

Except what's been paved is the road, not the traffic on it.

Here's what I eventually understood: whether a one-person company can keep running doesn't really hinge on whether one person can carry every step. What actually gets you stuck is the other end — whether upstream is still sending work out.

What big companies have been doing these past few years is taking whole functions they used to keep in house and slicing them into outsourced projects. One engineering role can become three chunks of project-based work. That sliced-off, leaked-out work is what a one-person company eats.

Economists call this trickle-down: opportunity seeps downward layer by layer like water. What a one-person company catches is the water that reaches the very bottom layer. When the layer above stops leaking, the bottom is the first to run dry.

Two people in one report lay the difference out clearly.

One is a developer making small bullet-hell games. He shipped six in a year, used AI to replace 70% of the art and to write code, held the cost per game to 1,000–1,500 yuan, cleared close to 20 million yuan in total revenue for the year, and took home 800,000 to 1,000,000.

The other left a big company to do cross-border e-commerce. He'd been on around 600,000 a year. Two months in, sales were a little over 10,000, profit a little over 1,000; with some agency-operations work on the side, his income shrank by 90%.

The AI tools in their hands weren't much different. The difference is that the first one caught an opening that was still ramping up, and the second walked into a market where trust needs a personal introduction to build — and he didn't have one.

There's a blunter number too. That game developer's revenue runs through platforms like Douyin. The platform takes half, the streamer's cut is a bit over 40%, and what finally lands in the developer's hands is roughly 8% of the revenue.

You think you're going it alone. You're actually standing at the far end of a chain where someone else already divided the take.

Put all of this together and the shape of the one-person-company business gets fairly clear.

It doesn't quite count as a business you hold in your own hands, come rain or shine. It's more like subcontracting work that rises and falls with the wider economy and the platform cycle. When upstream is ramping and the platform is handing out traffic, one person can eat well. When upstream tightens and the platform changes its rules, that's also who goes hungry first.

It isn't the same thing as building a company in the real sense either. A company building its own product is slowly accumulating its own demand and its own channels; a one-person company, most of the time, is taking pre-defined work inside someone else's channel. The first is building a house. The second is renting a desk.

None of this means you can't do it. The game developer made real money. But whether it's worth pulling out your day-job leg too depends on whether you can see clearly which opening your work comes from, and whether that opening is currently open or closing.

A few signals worth watching: in your industry, are big companies pushing work out, or pulling it back in; the platform your income depends on — over the last six months, have its cut and its traffic rules loosened or tightened; the work in your hands — is it something only you can do, or something that can be handed to a cheaper person at any time.

If those three drift the wrong way, the window for a one-person company narrows, no matter how fluent you are with AI.

That thought last month — I didn't act on it in the end. Not because of the hard work, but because I realised the "work" my system catches is really just time squeezed out around my day job plus a little platform tailwind, and neither is steady enough to hold up pulling out the other leg.

The phrase "one-person company" turns "one person can run a company" into "one person can make a stable living off this." The stretch in between is whether upstream is willing to keep leaking work down to you.