Lovable vs Cursor: $13.3B Valuation Reframes AI Coding Tool Race

Two separate markets that keep getting compared

The usual way to compare AI coding tools is a three-column table with Lovable, Bolt and Cursor side by side. That table hides the main point. These products sell to different buyers, and in 2026 each one has moved further toward its own buyer. Most of the year's comparison coverage agrees on a basic spectrum. Cursor is an AI layer inside a professional editor. Bolt is a browser sandbox built for fast experiments. Lovable is a chat-first route to a deployed full-stack app.110 The market news shows how far apart those lanes have drifted. Lovable doubled its valuation in about eight months. Cursor's parent, Anysphere, is reportedly in acquisition talks. Windsurf, the editor that once looked like Cursor's closest rival, was renamed into another company's agent brand.112623

My view is that "Lovable vs. Cursor" has stopped being a useful consumer question. The more useful questions are who owns the code, who owns the agent, and who pays for usage.

Lovable's growth does not settle the quality question

Lovable's financial story is the most dramatic in the category. The Stockholm company raised $400 million at a $13.3 billion valuation in August. Menlo Ventures and the EQT-managed Scaleup Europe Fund led the round, and the valuation was more than double the $6.6 billion mark from December.1417 The company reported $200 million in annual recurring revenue at the end of 2025, $300 million soon after, and $400 million by March.14 It said it had passed $500 million by early June.20 TechCrunch later reported that annualized revenue had crossed $600 million.15 One analysis prices the round at roughly 22 times forward revenue. It also cautions that the $600 million path was company-reported, not audited.11

The reported figures don't fully line up. One deal tracker dates a $330 million Series B to June 2026 at a valuation "north of $6 billion" and pegs the multiple at about 12 times ARR.13 Most other accounts place that $330 million round in December 2025 at $6.6 billion.1819 Headcount numbers range from 146 full-time staff in early 2026 to more than 1,000.15 Readers should treat secondary aggregators with care. The core facts hold up across several sources: $13.3 billion, $400 million raised, and revenue that roughly tripled since the end of 2025.1419

The reviews are much less consistent than the revenue figures. Lovable's own comparison guide presents it as a full-stack builder with real backends, one-click deployment, an Agent Mode for multi-file changes and Visual Edits for non-technical teammates.1 Independent reviewers mostly accept the speed claim and doubt what happens after the first build. One assessment says Lovable and v0 leave the most cleanup work for later. It rates Lovable's security as weak and its CI/CD support as limited.4 Another warns that app-generation platforms often ship with broken auth, missing rate limits and hardcoded secrets.4 A business-apps comparison goes further and calls Lovable best for frontend-heavy prototypes that struggle once backend complexity grows. That reviewer sells a done-for-you development service, which is worth keeping in mind.6 Developer-focused coverage points to "bug loops" on complex logic and lock-in to a fixed React-and-Vite stack.2

The reviews disagree on how capable the backend is, but they agree on what Lovable is for. It is widely named the easiest starting point for non-technical founders and the right tool when validating an idea matters more than production hardening.479 Investors appear to be paying for that validation loop, not for production-grade output. The credit-based model, which combines a subscription with metered top-ups, turns that loop into revenue very quickly.11

Bolt: catching up on features, far behind on revenue

Bolt keeps showing up in these comparisons, but the money gap is large. One market breakdown puts StackBlitz's Bolt at about $40 million in ARR, against Lovable's roughly $600 million.11 On features, Bolt has narrowed the gap. Lovable's own guide concedes that Bolt Cloud now offers native databases, auth and hosting, so calling Bolt frontend-only is outdated.1 Bolt's clearest advantage is flexibility: it supports React, Vue, Svelte and Astro, while Lovable uses a fixed stack.2

The complaints about Bolt are similar to the complaints about Lovable. Reviewers describe apps that work in preview and break in production. They also describe painful token limits on complex projects, and fixes that create new bugs.5 Several write-ups place Bolt in the middle of the spectrum. It feels more like a cloud code editor than Lovable does, but it doesn't match Cursor's depth.9 Pricing claims vary between write-ups. Bolt's entry tier is listed at anywhere from $10 to $29 a month, which suggests frequent plan changes and careless secondary reporting.389

Cursor now centers on managing agents

Cursor's 2026 product releases have changed it more than any comparison table shows. Cursor 3.0 shipped on April 2 with an Agents Window for running many agents in parallel across local machines, git worktrees, the cloud and remote SSH.35 New commands such as /worktree and /best-of-n let several models attempt the same task in isolated checkouts, so the developer compares the diffs and keeps the best one.35 Version 3.1 added tiled layouts for managing parallel agents.37 Version 3.3 in May added "Build in Parallel" subagents, Composer 2.5 and native Jira integration.35 Cursor 3.5 on May 20 introduced Cloud Agents, which run in sandboxed VMs with their own terminal, browser and desktop access and are billed separately from the seat price.3532

Other additions push the same way. Bugbot reviews pull requests and, since February, can start a cloud agent to test and propose fixes. One guide puts its resolution rate at around 80%.35 Subagents arrived in v2.4 in January, each with its own context window.32 Composer, Cursor's in-house coding model, reduced its reliance on outside model providers.33 Reviewers have described the product as less "VS Code with AI" and more an agent workbench that also happens to be an editor.33

The consensus on Cursor is unusually consistent. It is rated the safest and most production-ready option because a developer reviews every diff, and the strongest on context because it indexes the whole codebase.456 The downside is just as clear: it is close to useless if you can't code, and it provides no backend, auth or hosting.62 The corporate picture is less settled. Cursor's last priced round reportedly valued it at $29.3 billion, with more than $2 billion in ARR by February. Several outlets report a pending sale to SpaceX/xAI at around $60 billion, though those reports remain unconfirmed.1126

Windsurf has been absorbed into Cognition's Devin

The Windsurf angle is now mostly a story about a brand disappearing. Accounts disagree on the timeline. Several say Cognition acquired Windsurf in December 2025 for about $250 million.212225 Others place the acquisition in July 2025, right after Google paid a reported $2.4 billion to license Windsurf's technology and hire its CEO into DeepMind, following a collapsed OpenAI deal.2329 Both versions agree that Cognition took over a company with about $82 million in ARR and more than 350 enterprise customers.2325

The sources agree on what came next. In March 2026, Cognition replaced credits with daily and weekly quotas and raised Pro from $15 to $20 a month, which matches Cursor and ends Windsurf's budget positioning.2126 On June 2, an over-the-air update renamed the editor "Devin Desktop," and windsurf.com now redirects to devin.ai.2326 The Cascade agent reached end-of-life on July 1. Devin Local replaced it, and an Agent Command Center now coordinates local and cloud agents in a Kanban view.28 Cognition's proprietary SWE-series models, now at SWE-1.6, remain its main technical differentiator.2428

Cognition and Anysphere are now building the same thing: an IDE designed to run a fleet of agents. Spring coverage still described Windsurf as "closing the gap."27 By summer, reviewers noted that anyone who liked Windsurf as a cheaper, simpler Cursor alternative had lost that option.29

What buyers should take from this

The category has split in two. Prompt-to-app builders like Lovable and Bolt compete on how fast a non-developer can reach a working demo. They charge through credit or token meters that reviewers repeatedly say get expensive during debugging.811 Agent-native editors like Cursor and Devin Desktop compete on how many supervised agents a professional can run safely at once, and increasingly they charge for cloud compute on top of seats.3528

The sensible way to buy is to pick the lane first, then look at features. Founders validating an idea should expect to rebuild or harden a Lovable app before scaling. One reviewer with a commercial interest estimates that 40% to 60% of such projects need partial or full rewrites. Treat that as a warning, not a measured figure.6 Engineering teams should judge Cursor and its rivals on governance: privacy modes, review tooling and how agent costs are metered.3537 Going by valuation, the market currently pays more for Lovable's growth. Going by the reviews, the code that holds up in production still comes mostly from tools where a developer approves every change.