Anthropic built Cursor's foundation. Now it's dismantling it.

Cursor is four years old and already in the trap that successful tech companies sometimes cannot code their way out of.

The AI coding company — founded by MIT undergrads in 2022, valued at $2.5 billion at the start of 2025 and nearly $30 billion by year's end — is raising a new round at $50 billion per Fortune. It crossed $2 billion in annualized revenue in February. Sixty-seven percent of the Fortune 500 uses it. By any conventional metric, this is one of the fastest enterprise software companies ever built.

The structural problem underneath is what makes the valuation interesting — and fragile. Cursor runs on Anthropic's models. Anthropic sells those models to Cursor at retail. Anthropic also competes with Cursor via Claude Code, which launched in February 2025 and has been eating Cursor's lunch in developer mindshare. As one VC told Fortune: Anthropic is trying to drown out Cursor. The company that made AI coding popular is paying rent to the company that is now using that position to undercut it.

The dependency trap made explicit

Cursor built its empire on a simple idea: embed AI deeply into the IDE, the application where developers already live. The tool reads your codebase, suggests completions, lets you chat with your code in a sidebar. It was code assistance — Tony Stark wearing the Iron Man suit.

Claude Code took a different approach: give the AI the task and let it execute. No GUI, no sidebar, no IDE integration. It runs in the terminal, reads your entire codebase, and ships. The AI wears the suit while you move on to the next problem.

By early 2026, Claude Code had captured significant developer mindshare — a developer survey on dev.to put Claude Code at 46% versus Cursor at 19% in a showdown poll, a reversal that played out in under a year. The figures come from an informal community poll rather than formal research, but the directional signal is consistent with what the broader discourse has reflected. Claude Code reached $2.5 billion in run-rate revenue by early 2026, with over 300,000 business customers, per CNBC's reporting on Anthropic's metrics.

Cursor's president Oskar Schulz acknowledged the shift directly: the IDE is not the right form factor anymore for a world where you can produce 10 times more code. Ninety-five percent of Cursor users are now agent users — meaning the company is essentially trying to bolt an agentic architecture onto an IDE-first product.

Can Cursor build its way out?

Cursor's response is to build its own model. Its in-house model, Composer, has outperformed Anthropic's Opus 4.6 on some benchmarks, per Fortune — a genuine signal that the company is not helpless. But training and running frontier models is expensive, and the talent market is dominated by the very labs Cursor is competing against. Key engineer departures to xAI — Jason Ginsberg, Andrew Milich — suggest the company is struggling to retain the researchers it needs to close the gap.

The unit economics question is also sharper than the revenue headline suggests. Cursor's $2 billion annualized revenue is impressive in isolation. Whether it constitutes a durable business depends on margins — and on whether Anthropic raises model prices further as Claude Code scales.

Cursor president Oskar Schulz pushed back on the doom framing in Fortune's piece: "We knew from the start the best solution was going to change every six months, every year." CEO Michael Truell keeps a photo of biographer Robert Caro above his desk as a reminder that consequential work takes time. That philosophy built a $30 billion company in two years. Whether it can sustain one at $50 billion is the open question.

The valuation signal

The bigger read from Cursor's trajectory is what it tells us about the AI coding tooling market. The category is projected to hit $30 billion by 2032, up from $4.9 billion in 2024, per SaaStr's analysis. Cursor's current revenue suggests it is capturing significant share — but the competitive field has gotten crowded in under a year: Anthropic's Claude Code, OpenAI's Codex, Cognition's Windsurf, and GitHub Copilot all compete for the same developers and enterprises.

Cursor at $30 billion is a bet that the AI-augmented developer workflow is worth that much as a standalone business. That may be true. Or it may be that venture-backed growth metrics got conflated with durable business value. At $50 billion, the company needs to sustain growth rates no B2B software company has maintained at that scale in such a compressed timeframe.

The acquisition path remains live. OpenAI reportedly approached Cursor about acquisition in early 2025, per CNBC. That deal did not happen. If unit economics remain challenging and public market pressure on AI companies intensifies, the option may look more attractive to both sides.

The IDE era for AI coding may be ending. Cursor — if it can navigate the agentic transition, the model dependency problem, and the unit economics trap — is far from dead. It is just in a corner that its own success helped build.