AI App Builders: Lovable Hits $600M as Bolt and Bubble Diverge

The market splits in two

People still compare Lovable, Bolt and Bubble as if they were three versions of the same product. Through 2026, the market has made clear that they are not. Lovable and Bolt write real code, mostly React, from a chat prompt. Bubble keeps the app inside its own visual engine and does not let you export the code3335. At the same time, the AI-native editors that professional developers use, Cursor and the editor once called Windsurf, have moved past autocomplete. They now act as control rooms for teams of autonomous agents2612.

This piece argues that "vibe coding" platforms and AI coding IDEs are growing toward each other. Both are betting that software will be planned and built by agents, with people reviewing the results. The two groups still serve different users, though, and the differences in lock-in, pricing and production readiness matter more than any benchmark.

Lovable's revenue explains the hype

Lovable's growth is the main reason prompt-to-app tools get serious attention. Co-founder Fabian Hedin said at the HumanX summit in Amsterdam in September that the Stockholm company had passed $600 million in annual run-rate revenue. In June that figure was about $500 million6. In August the company raised a $400 million Series C at a $13.3 billion valuation, roughly double the $6.6 billion it was valued at in December 202562. Menlo Ventures and the EU-backed Scaleup Europe Fund, managed by EQT, co-led the round. Tencent, Salesforce Ventures and returning backers such as Accel and CapitalG also took part7.

The coverage mostly agrees on the growth path: $100 million ARR in July 2025, $200 million by November, $400 million by February 202658. One trackers' estimate puts the $13 billion valuation at about 26.6 times revenue7. Older data aggregators still list the $6.6 billion figure, which shows how quickly the numbers have changed13. These run rates come from the company and are not audited8. Even so, the September disclosure that two-thirds of the Fortune 500 use Lovable suggests demand reaches well beyond hobbyists8.

Lovable vs. Bolt vs. Bubble: where reviewers agree

Comparisons published this year broadly agree on how the three tools differ. Bolt is the fastest way to get a working prototype in the browser. It supports React, Vue and Svelte, and it shows the code at every step3140. Lovable is usually called the more complete full-stack option, producing React and TypeScript with authentication and a database attached33. Bubble, around since 2012, is described as the deepest no-code logic engine, with a large plugin ecosystem. The trade-off is a steep learning curve3736.

Entry pricing is also similar across the board. Lovable and Bolt both start around $25 a month. Lovable's tier includes about 100 credits, and Bolt's includes roughly 10 million tokens3339. One harsh review says those allowances run out quickly on real projects, with Lovable's credits covering maybe 20 to 30 serious prompts36. Bubble's quoted entry price ranges from $29 to $59 a month depending on the outlet and whether mobile is included. That spread shows Bubble's plans have been changing, not that the sources contradict each other343338.

Mobile is a clear point of difference. Bubble publishes natively to app stores on mobile-enabled plans. Bolt has a real but manual Expo pipeline. Lovable remains web-only and has no documented path to the app stores39.

Where the coverage diverges: lock-in and audience

The sharpest disagreement is about Lovable's portability. Some reviewers call code ownership Lovable's main advantage: if the company disappeared, your app would keep running3239. Another review strongly disagrees. It argues that Lovable's GitHub export relies on a forked Supabase setup that doesn't realistically run elsewhere, and calls this "lock-in dressed as portability"36. Coverage also differs on the backend. Some describe Lovable as depending on an external Supabase instance that users must configure themselves31. Others treat Lovable Cloud as a full-stack platform in its own right32.

Reviewers also disagree about who Lovable is for. One guide recommends it to non-technical founders33. A competing vendor's guide says it is not a good fit for that group31.

Who publishes these comparisons matters. Many are written by companies with competing products, including Bubble, Mocha, Replit and Superapp, and some say so openly343139. Bubble's own blog concludes that Bubble is the only option for production apps built without code34. Readers should treat these comparisons as positioning, not neutral testing.

Our reading: Lovable is the strongest web-first choice for getting from idea to deployed app, but "you own the code" means less than it sounds when the backend is tied to the platform. Bolt suits developers who want fast prototypes and real portability. Bubble is still the safer choice for complex logic, as long as you accept that you can never leave without rebuilding.

Cursor turns the editor into an agent control room

For professional developers, the important changes are happening in the IDE. Cursor 3.0, released April 2, 2026, added an Agents Window for running many agents in parallel and moving work between local, cloud and SSH environments. A tiled layout followed in version 3.129. Cursor's in-house Composer 2 model arrived in March, and Composer 2.5 followed in May as the default for new CLI sessions26. Version 2.4 in January introduced Skills, task-specific workflows the agent loads only when needed, separate from the always-on Rules26.

Cloud agents now run in isolated virtual machines and can be started from the web, Slack, GitHub or Linear2129. Self-hosted cloud agents became available to Enterprise customers in March29. One tutorial reports that in mid-October Cursor added interactive Canvases for cloud agents and an Environments API, and announced an Agent SDK. Those details come from a single outlet and should be treated as provisional21. One comparison also claims Cursor's parent company Anysphere is reportedly being acquired for about $60 billion. No other coverage here confirms that, so it is best treated as unverified14.

Windsurf is now Devin Desktop

The editor once known as Windsurf has gone through the most upheaval. In July 2025, OpenAI's planned $3 billion acquisition collapsed. Google then paid a reported $2.4 billion to license the technology and hire CEO Varun Mohan and key staff, and Cognition bought the remaining company days later1516. Some reviews date the Cognition deal to December 2025 at around $250 million1112, but the July 2025 sequence is the better-documented account.

Under Cognition, Windsurf 2.0 arrived in April 2026. It added the in-house SWE-1.5 model, Codemaps, the Fast Context retrieval system, and Devin built in as a cloud agent13. On June 2, an over-the-air update renamed the product Devin Desktop and kept users' settings and extensions20. Several outlets report that the Cascade agent reached end of life on July 1 and was replaced by Devin Local1220. One comparison, however, still describes Cascade as a live feature14. In March, pricing moved from credits to quotas, and Pro rose from $15 to $20, the same as Cursor1114.

The strategy looks deliberate. Devin Desktop supports the Agent Client Protocol, so other vendors' agents can run inside it19. Cognition released its SWE-2 model on September 10 and raised more than $2 billion at a $48 billion valuation19.

Why it matters

The two categories are drifting toward each other. Lovable's investors are funding the move from prototype to production, with enterprise integrations, governance, databases and payments7. Cursor and Cognition, meanwhile, are building agent dashboards that a non-specialist could plausibly use. Bolt's Plan Mode and Cursor's planning-first agents follow the same idea: plan first, generate second3227.

The near-term risk is cost, not capability. Credits, tokens, workload units and quotas all hide the real price of shipping software, and several reviewers flag that unpredictability as the biggest catch3639. Whichever tool wins over the next year will probably be the one that makes the bill as predictable as the output.