On 23 July 2025, the Stockholm company Lovable announced it had passed $100 million in annual recurring revenue, eight months after crossing its first $1 million. TechCrunch, reporting the same day, framed it as eight months from launch and added the numbers underneath: 2.3 million active users, 180,000 paying subscribers, 10 million projects created, and 45 full-time employees.
Four months later, TechCrunch reported the figure had doubled to $200 million ARR. The company had by then raised more than $225 million, including a $200 million Series A at a $1.8 billion valuation.
Numbers that steep invite one of two lazy readings: that it was pure timing, or that the product was simply so good it sold itself. Lovable's own account of its first months points somewhere more useful.
It did not start at zero
The most important fact about Lovable is buried in its own January 2025 post on reaching $10 million ARR in two months: the product grew out of GPT Engineer, an open-source project by Anton Osika that had already become well known.
That changes the arithmetic of the launch completely. A commercial product built on top of a widely used open-source project starts with a population who have already installed the thing, already trust the author, and already tell other people about it. The two months from zero to $10 million were not two months of work. They were two months of conversion, following a much longer period of building an audience for free.
This is the pattern we described in distribution is the moat, in its most extreme form. The open-source project was the distribution. The commercial product was the monetisation of an audience that already existed.
The channels were named, not stumbled into
Lovable's own list of what drove the first phase is unglamorous and specific: a Product Hunt launch; presence on X, TikTok, YouTube and LinkedIn; co-marketing partnerships with Supabase, Replicate and Resend; user-generated content; and hackathons.
Two of those deserve attention because they are available to companies with no budget.
Co-marketing with adjacent tools is the cheapest distribution in software. Supabase sells a database. Replicate sells model hosting. Resend sells email. None competes with Lovable, and every one of them has customers who need what Lovable does. Joint content puts a product in front of a pre-qualified audience at the cost of writing something together. If you sell developer tools, the list of companies whose customers need you but who do not compete with you is usually longer than you think.
User-generated content is the other. A tool that produces a visible artefact, in this case a working app, has a structural advantage: every satisfied user creates a demo. Hackathons formalise that, turning it into a scheduled supply of public examples. If your product's output is invisible, this channel is closed to you, and it is worth asking whether you can make the output shareable by design.
The pricing decision that cost $1.5 million in one day
The detail from this story most likely to be useful to a smaller company has nothing to do with AI.
Osika told TechCrunch that Lovable moved Team tier users onto the cheaper Pro tier and, as a result, "lost $1.5 million ARR in a single day."
Deliberately deleting $1.5 million of committed revenue is not a decision most founders can bring themselves to make. It is the correct one when a tier is misallocating customers: if people are paying for a plan whose value they are not receiving, the revenue is borrowed against future churn and future complaints. Cutting it converts a slow bleed into a single visible cost.
The general form: a price that is wrong keeps being wrong, and the longer you leave it, the more revenue you are protecting and the harder it becomes to fix. We covered the same dynamic from the other direction in your first price is too low and, for usage-based AI products specifically, in pricing AI features when every click has a cost.
What 45 people and 180,000 subscribers imply
Run the numbers TechCrunch reported in July 2025 side by side. About $100 million ARR across 180,000 paying subscribers is an average of roughly $555 per subscriber per year, or a little over $46 a month. That is self-serve pricing, not enterprise contracts.
And $100 million ARR across 45 employees is on the order of $2 million of recurring revenue per person. Both figures describe the same company: one that sells to individuals and small teams through a product-led funnel, with no meaningful sales organisation.
That model has a specific vulnerability. Self-serve subscribers at $46 a month leave without a phone call, and a competitor's better release is one click away. High revenue per employee is a strength while growth is fast and a fragility if churn rises, because there is no contracted revenue underneath.
Osika also credits an unusual structural choice. He told TechCrunch he resisted pressure to move the company to Silicon Valley: "It was tempting, but I really resisted that. I [can] sit here now and say, 'Look, guys, you can build a global AI company from this country.'"
Reading this honestly
What a smaller team can actually take from Lovable: build the free artefact first and let it accumulate an audience before charging; name your distribution channels explicitly and work them rather than hoping for word of mouth; partner with the companies whose customers you need and who do not compete with you; make your product's output shareable; and fix broken pricing immediately, even when the correction costs real money.
What is not transferable: the moment. Lovable launched into the steepest adoption curve any software category has had, with buyers actively searching for AI tools and investors funding them at unusual speed. A comparable product with comparable execution in 2019 would not have produced these numbers, and saying otherwise would be dishonest.
It is also worth being clear about what these figures are and are not. ARR is a run rate, not audited revenue. Fast-growing self-serve companies can post large ARR while churn is high, and Lovable has not published churn, profitability, or gross margin. The company is roughly two years old. The interesting question is not how it reached $200 million ARR but whether the subscribers who produced it are still there in 2028.
Where the company chose to be
Osika has been explicit that staying in Sweden was a decision rather than an accident. He told TechCrunch that the pressure to relocate was real: "It was tempting, but I really resisted that." His stated reason is a hiring argument, not a patriotic one: "There is more available talent if you have a strong mission, and you have a lot of urgency coming together as a group and working."
The claim underneath is that in a market where every well-funded competitor bids for the same engineers in the same square mile, being somewhere else is a recruiting advantage rather than a handicap. Whether that generalises is unproven, but it is testable, and it costs nothing to consider before signing a lease in an expensive city.
The funding kept pace with the revenue. Lovable's own blog records the $200 million round in July 2025 and, in December 2025, a post titled "Lovable raises $330M to power the age of the builder." A company adding capital that fast is buying time to hold a position, not runway to find one.
Figures are from Lovable's own posts of January and July 2025 and TechCrunch's reporting of 23 July and 19 November 2025, checked July 25, 2026.
Discussion
Sign in with Google or just a name. No email link, no password to remember.