Vibe Coding Boom: Lovable Tops $600M Run Rate as Agents Hit APIs

The headline number

Vibe coding now has a revenue milestone to point to. Lovable co-founder and CTO Fabian Hedin told the HumanX summit in Amsterdam that the Stockholm company's annualized run-rate revenue has passed $600 million. In June the company had put that figure at roughly $500 million.2122 Several outlets lay out the same climb: about $200 million in November 2025, then $300 million and $400 million in quick succession in early 2026, then $500 million by June.2329

Hedin added two more numbers. He said people at two-thirds of Fortune 500 companies use Lovable, naming Microsoft, Nvidia and Deutsche Telekom as customers. He also said apps built on the platform draw close to a billion visits a month, which he described as about ten times the traffic to Lovable's own site.2124 The money behind the company tells a similar story. Lovable raised $300 million at a $6.6 billion valuation last December, then $400 million at $13.3 billion in August, co-led by Menlo Ventures and the Scaleup Europe Fund.2128

The news is about more than one startup's revenue. Read together, the reports show vibe coding becoming three things at once: a consumer business with unusually fast growth, a bottom-up channel into large enterprises, and a stack built on APIs that more and more targets AI agents as well as people.

Where the reporting agrees, and where it softens

The outlets agree on the basic facts: $600 million, up from about $500 million in June, announced on stage at HumanX.2124 They differ on how much weight to give it.

Most early coverage repeated the Fortune 500 claim without qualification. One summary went further and said two-thirds of Fortune 500 companies use Lovable "for their apps."27 TechCrunch, however, added a clarification from Lovable: Hedin meant that people at two-thirds of those companies use the product.21 That is a different claim from enterprise-wide deployment. The Next Web's account backs the narrower reading. Hedin described staff finding the tool on their own, after which their employers show up holding thousands of apps and asking what to do with them.22 One analysis says the share has risen from about half the Fortune 500 six months earlier, and calls the pattern a shadow-IT wave, not signed contracts.23

The more skeptical coverage also points out that $600 million is a run rate, meaning a snapshot of current sales projected over a year, not revenue earned in a completed year. It adds that the traffic number isn't new: Lovable had already reported more than 900 million monthly visits in August.28 Another breakdown calls all of the numbers company-reported and unaudited.23

The caveats matter, but the growth is real. Even the cautious outlets accept the direction of travel. One analysis notes that the valuation doubled while the implied revenue multiple fell from about 33x to roughly 22x, because revenue grew faster than the price investors paid.23 When revenue outpaces a sharply rising valuation, investors are being asked to believe less, not more. Our reading: the Fortune 500 line is overstated as written, but the core business numbers are holding up.

User growth: big numbers that don't line up

Outside Lovable, the user-growth picture is huge and inconsistent. GitHub Copilot is cited at more than 20 million all-time users, up 75% year over year, with 4.7 million paid subscribers.12 Another 2026 roundup puts it at "over 15 million."17 OpenAI Codex reportedly passed 4 million weekly active developers in April 2026.12 Vercel's v0 reports more than 4 million users.14

Lovable's project counts don't agree with each other either. One report gives 50 million projects, growing by about a million a week, up from 1.2 million in February 2025.14 Another says more than 60 million projects have been created since the public launch in November 2024.23 The more careful accounts warn that a project is not an app. The totals include duplicates, dashboards and experiments abandoned within the hour, so "50 million vibe-coded apps" overstates what the data shows.14

The surveys of professional developers disagree even more. A JetBrains survey found 90% of developers regularly using at least one AI tool at work as of January 2026, up from 85% in mid-2025.12 Sonar's survey of more than 1,100 professionals found 72% of AI-tool users using them daily, with an estimated 42% of committed code AI-generated or assisted.1319 On vibe coding specifically, though, results nearly flip depending on who asks. One dataset has 77% of developers saying vibe coding is not part of their professional workflow, while Adaptavist found 83.9% of 240 US and UK engineers vibe-coding in some form.14 Some widely shared market figures add to the confusion. One vendor report still lists Cursor at $100 million ARR and a $9 billion valuation, while other 2026 coverage cites a $29.3 billion valuation and ARR above $2 billion.111218

The most consistent finding is about who the new users are. Lovable says about four in five of its users are non-technical, with 45.7% identifying as founders and only 5.8% as engineers.14 Hedin said 80% of the ideas people bring to Lovable are about starting a business.22 Vercel data cited in one compilation puts non-developers at 63% of vibe-coding users and Asia-Pacific at 40.7% of global use.12 Professional developers use AI assistants heavily inside existing workflows. Most of the growth in "vibe coding" as a product category comes from people who weren't writing software before.

The API angle: from apps for people to endpoints for agents

The least-covered part of this story is the infrastructure underneath. Hedin's pitch is that Lovable doesn't output code but a product, and more and more a business, with hosting, deployment and scaling built in.24 Lovable Cloud now adds payments, email, SEO and ads.22 One report says the platform automatically routes prompts across more than 100 underlying AI models, weighing cost against quality, instead of letting users choose.23 Another growth history links an earlier jump, from under $500,000 to roughly $20 million ARR, to Lovable's Supabase integration, which gave users databases, authentication and storage.29 Lovable's moat, in other words, is mostly the set of APIs it wires together.

Cloud providers are now selling that template directly. Cloudflare has published a reference architecture for an "AI vibe coding platform." It routes prompts through an AI Gateway that spans OpenAI, Anthropic and Google, can serve cached answers to common requests such as a todo app to save on inference, runs untrusted generated code in isolated sandboxes, and hosts potentially millions of resulting apps, each in its own Worker.1 Cloudflare also offers prompts and a documentation MCP server built to make AI models write better code against its own APIs.1 In short, the company is courting the models as well as the developers.

The same shift is reaching companies outside developer tools. Vibe, a connected-TV ad platform, used one quarter to ship a public API, an AI-assistant connection via MCP, and a command-line tool aimed partly at coding agents, all writing to the same account data as its web interface.49 Its head of growth used the API to set up 250 account-based campaigns, each with its own audience and creative.4 The tools are free to all advertisers.4 Exposing a business through an API and an MCP endpoint is becoming the expected response to customers who now build software by prompting.

One vendor's summary of Andrej Karpathy's view goes further still: today's vibe coding is a transitional stage, and the end state is agents acting on APIs directly for people, which could make standalone apps unnecessary.18 That is a forecast, not a finding. Still, the Cloudflare and Vibe examples suggest companies are already building toward it.

The cost: secrets, security and trust

Putting APIs in the hands of non-developers brings predictable risks. Beginner guides now dwell on API keys. They tell readers never to commit keys to GitHub, never to paste them into an AI chat, and to keep calls on the server side.5 Others tell new builders to check whether keys are visible in browser code and to test with a second account to confirm that users can't see each other's data.7 One compilation reports that AI-generated code produces about 1.7 times as many issues as human-written code, and that 45% of samples fail OWASP Top-10 security benchmarks.12 Retool found that 93% of senior technology and security leaders worry about vibe-coded tools running in production.13

Our view: Lovable's $600 million is a real sign that prompt-built software has found paying customers at scale. The next phase depends on governance. Enterprises that discover thousands of employee-built apps will want audit trails, permissions and control over secrets. Platforms that can provide those through their APIs, and not only through their chat boxes, are best placed to keep growing.