Lovable has crossed $600 million in annual run-rate revenue, co-founder and chief technology officer Fabian Hedin said on stage at the HumanX summit in Amsterdam. The interesting part of that number is its shape rather than its size: the Stockholm company was near $200 million a year ago by the reckoning of Forbes senior editor Iain Martin, who conducted the interview, and about $500 million as recently as June. Tripling in twelve months, then adding a fifth of that again in a quarter, is a curve that has flattened while the valuation attached to it has not.
Key takeaways
- Annual run-rate revenue has reached $600 million, against roughly $500 million in June and about $200 million a year earlier.
- Hedin said employees at nearly two-thirds of Fortune 500 companies use the product, and that Deutsche Telekom alone runs more than 2,000 applications built on it.
- Applications made on the platform pull close to a billion visits a month, which Hedin said is an order of magnitude beyond the traffic Lovable's own site sees.
Adoption that skipped procurement entirely
In his conversation with Forbes at the summit, Hedin described a pattern that inverts enterprise software's usual direction of travel. Employees find the tool themselves, solve a problem nobody had budgeted for, and the employer surfaces later holding thousands of applications and asking what to do with them. Lovable's enterprise work starts at that point: mapping which app touches which data, who can edit it, and what each of those editors indirectly gains access to.
The company subsequently clarified that Hedin's Fortune 500 claim describes individual employees, not corporate contracts. Microsoft and Nvidia sit on the named-customer list beside Deutsche Telekom, and the builders are spread across finance, HR, product and design rather than concentrated in engineering. Hedin's illustration was a sales leader at Uber Eats who understood the restaurant pitch process end to end and shipped a tool generating a tailored pitch for each one β hours for the first version, roughly ten minutes for later changes, against a queue that used to consume months of engineering backlog.
Why Lovable argues it is not in the code business
Pressed on why anyone would use Lovable rather than going straight to Codex or Claude Code, Hedin drew the distinction at the output. Those produce code; Lovable, in his telling, produces a product and increasingly a business, because hosting, deployment and scaling arrive with it. Lovable Cloud has since added payments, email, SEO and ads. He put the share of incoming ideas that are really business ideas at 80 percent, and offered a Brazilian founder whose AI education platform β built on Lovable, with 50 staff running internal processes on it β is tracking toward roughly $20 million in annual recurring revenue this year.
There has never been a model picker in the product. A single task can route across more than a hundred models weighed on cost, speed and quality, with the savings passed through and the cheaper model often the faster one as well. Hedin said customers began asking for exactly that once the industry's appetite for maximising token spend showed up on invoices, and that Lovable is now tuning its own models to push cost down on particular jobs. For a company reselling inference at scale, routing is margin, and margin is what a 22-times revenue multiple eventually has to be justified with.
A million projects a week, and nobody reading them
Volume is the uncomfortable half of the story. More than a million projects are created weekly, past any possibility of human review, so the platform scans all of them automatically, continues scanning after a user walks away, repairs critical dependency flaws outright and proposes fixes for the remainder. Hedin said the scanner is free to every user and the first of its kind in the category.
He did not wave the risk away. Security teams worry about the volume of new code, rightfully so in his words, and Lovable works with CIOs on who builds what and which data each application may reach. His sharper point concerned AI agents: large enterprises have long relied on permission systems that were imperfect but impenetrable, a form of security through obscurity. An agent can navigate them. That observation applies well beyond anything built on Lovable.
What the money says about the market
Investors have front-run the revenue. Lovable has raised more than $700 million in two rounds eight months apart β $300 million from Menlo Ventures and CapitalG last December at a $6.6 billion valuation, then $400 million in August from Menlo Ventures and the Scaleup Europe Fund at $13.3 billion. The vocabulary is moving too. Chief executive Anton Osika said last week the company is well past vibes, and Hedin said he favours agentic coding over the label Andrej Karpathy coined two years ago, on the grounds that calling any of this a vibe no longer fits. Both founders looked at San Francisco early and stayed in Sweden, where Hedin said access to talent turned out to be a competitive advantage.
FAQ
Is Lovable's $600 million figure audited revenue?
No. It is a self-reported annual run-rate quoted by a co-founder on a conference stage, and run-rate annualises a recent period rather than reporting what the company actually collected over the past year. Lovable is private and publishes no audited accounts.
What does two-thirds of the Fortune 500 mean here?
It counts individual employees rather than corporate agreements. Lovable clarified that Hedin meant people at nearly two-thirds of Fortune 500 companies are using the platform, frequently having adopted it on their own before any procurement conversation happened.
Can users choose which AI model builds their app?
No. Lovable has never offered a model picker and instead routes each task across more than a hundred models based on cost, speed and quality, the same approach many AI coding assistants expose as a setting. The company is also fine-tuning its own models for some tasks.
