Every venture capitalist in Silicon Valley begged David Holz to take their money. He said no. Not once, not in some dramatic press conference moment, but repeatedly, quietly, over years. While his competitors raised billions, burned through it building large organisations, and diluted their founders down to single-digit stakes, Holz kept the whole company and kept building.
That part of the story is solid. The numbers usually attached to it are not, and this page used to attach them too. So before the story, the ledger.
What is actually known
Verified: Midjourney has taken no outside funding. David Holz owns the company. It charges four flat subscription tiers, from $10 to $120 a month, with no free tier, no enterprise contracts and no sales team. Forbes reports the company as profitable.
Estimated: Revenue somewhere between $300M and $500M a year. Forbes put it at around $300M for 2024; the ~$500M figure has been widely quoted since. Nobody outside the company can reconcile the two, because Midjourney publishes nothing.
Unknown: Headcount, actual profit, margin, subscriber count. Public headcount estimates range from roughly 40 to 163 people — a spread wide enough that any "revenue per employee" figure built on it is arithmetic theatre. Profit is reported, never disclosed. We do not print a per-employee number for Midjourney, and you should distrust anyone who does.
That distinction — verified, estimated, unknown — is the whole point of this page. For how Midjourney sits against companies that do disclose, see our AI revenue leaderboard, where every row carries its source and confidence level.
Why the efficiency story survives the missing numbers
Strip out the fake precision and the interesting part is still standing. A company with no outside capital, whatever its exact size, is funding its own research from subscription revenue in the most capital-intensive corner of technology. It competes with organisations that have raised tens of billions. It has never needed a round.
You do not need a per-employee ratio to see why that matters. You need only the structure of the business.
The structure: radical simplicity
Midjourney's product runs on a text prompt and returns images. The pricing is four tiers and nothing else — Basic at $10 a month, Standard at $30, Pro at $60, Mega at $120. No free tier. No freemium bait-and-switch. No ad-supported model. No enterprise sales team negotiating custom contracts through months-long procurement.
Every customer pays the same way, which means the company needs no salespeople, no account managers, no solutions engineers, no customer success team running quarterly business reviews. The entire revenue engine is self-service. That, not a headcount trick, is what lets a small team serve a very large number of subscribers.
Why Holz said no to venture capital
This is the part that cuts against everything the startup ecosystem tells founders. The conventional playbook is clear: raise as much as possible, as fast as possible, from the most prestigious investors possible. Growth at all costs, profitability later.
Holz looked at that and saw dilution, loss of control, and pressure to grow headcount before growing revenue. He saw board meetings and investor updates pulling attention away from the product. He decided none of it was necessary.
The result is that he owns the entire company. At a conservative revenue multiple, Midjourney would be worth several billion — and Holz gave up not a single percentage point. In an industry where founders routinely raise billions and end up owning five to ten percent of what they built, that is a rare outcome.
Ownership means control. Holz can decide the product, the strategy and the pace without approval from a board or investors who need a specific return by a specific date. That freedom shows up in everything about how the company operates.
The product is the marketing
Midjourney spends approximately nothing on marketing. No paid ads, no content team, no social media managers, no influencer partnerships, no growth hackers running conversion experiments.
So how do people find it? Because every image it produces is an advertisement. When someone shares a striking generated image — in a presentation, a pitch deck, a feed — that image sells the product to everyone who sees it. The aesthetic became recognisable enough that the output does the selling.
Most AI companies misunderstand this. They assume they need a sales team, a marketing team, a content team and a growth team. Midjourney is the standing argument that when the product is good enough and its output is inherently shareable, the traditional marketing machine is optional. Product quality is the cheapest customer acquisition channel ever invented, and most companies buy marketing because their product cannot sell itself.
What it says about AI business models
The dominant narrative says AI requires enormous capital. Foundation models cost hundreds of millions to train, compute is astronomical, you need billions just to stay competitive. That is true at the frontier of model development. It is false for companies that apply AI to a specific problem and charge for the value delivered.
Midjourney is not trying to build general intelligence or compete on reasoning benchmarks. It generates images for people who want images, at a price they will pay. The ambition is calibrated to what a small team can deliver profitably.
The narrative also says consumers will not pay for AI, that enterprise contracts are where the money is. Midjourney is the counterexample: individuals paying $10 to $120 a month, at scale, with no salesperson involved. The consumer AI market keeps being underestimated by investors fixated on enterprise SaaS metrics.
The profitability gap
Here is the comparison that actually holds up. Most of the largest AI companies do not disclose a profit, and the ones whose economics are visible are spending far more than they earn — deliberately, funded by investors who treat the losses as an investment in future dominance.
Midjourney is reported as profitable and has been self-funding since launch. No audited figure exists, so treat the margin estimates you see with the same suspicion as the headcount ones. But the structural claim is sound: a company that takes no outside money and keeps operating has been covering its own costs. If the venture market froze tomorrow, that would not change.
We keep the profitability picture for the whole sector in one place rather than in every article — the revenue and profitability ranking lists which AI companies disclose a profit, which merely project one, and which say nothing at all. The short answer is that almost none disclose.
The lessons that survive
The wrong takeaway is that every AI startup should bootstrap. Some genuinely need hundreds of millions to build their technology; you cannot self-fund frontier training from subscription revenue.
The real lessons are narrower. Venture capital is a tool, not a requirement — the existence of a profitable, unfunded company in the most capital-hungry sector in tech should end the assumption that raising is mandatory. Simplicity scales: four pricing tiers and one product surface mean fewer people are needed to run the business, and people are the most expensive thing any company buys. Quality is a growth engine: if every use of your product is an advertisement, the marketing budget becomes optional.
And one more, learned the hard way on this page: be suspicious of the numbers that make a story too neat. Midjourney's achievement is real. The tidy arithmetic that used to sit on top of it was not.
Change log
- 7 August 2026 — Substantially corrected. The previous version of this article was titled "Midjourney Makes $500M a Year With 40 People and $0 Raised" and built its central argument on $12.5M of revenue per employee. Midjourney has never published a headcount; current public estimates range from about 40 to 163 people, which makes any per-employee figure unusable. We removed the calculation, the headcount claim in the title and headline, and the derived profit figure of "$350–400M". We also removed a subscriber claim — "16 million paying users at roughly $31 per month" — that implied nearly $6B in annual revenue and therefore contradicted this article's own revenue figure by an order of magnitude. Revenue is now given as the $300M–$500M range, profitability as reported rather than calculated. Sector-wide comparisons moved to the AI revenue leaderboard so the figures live in one place instead of ageing separately in several.
---
Keep Reading
- AI Revenue Leaderboard 2026 — every major AI company by run rate, valuation and disclosed profitability, with sources and confidence levels.
- OpenAI's revenue machine — the opposite end of the spectrum: the most-funded AI company and how its money works.
- Cursor's valuation run — another AI startup with unusual capital efficiency and explosive growth.
- How to Make Money with Claude AI — practical ways to build income with AI tools.
