According to Fortune, AI code editor Cursor is acquiring code review startup Graphite in a cash-and-equity deal expected to close in the coming weeks. Cursor, valued at a staggering $29.3 billion and boasting $1 billion in annualized revenue, will keep Graphite as an independent product but plans deeper integrations. Graphite, used by over 500 companies like Shopify and Snowflake, raised $52 million in a Series B round just in March 2025 and saw its revenue grow 20x in 2024. Cursor CEO Michael Truell says the acquisition tackles the emerging bottleneck of code review, which hasn’t kept pace with AI-accelerated coding. Graphite CEO Merrill Lutsky adds that together they aim to build an “end-to-end platform,” with smarter, context-aware reviews planned for 2026. The deal follows Cursor’s acquisition of AI assistant Supermaven in November 2024.
The Productivity Paradox
Here’s the thing: this deal is happening against a really weird backdrop. Cursor is citing massive productivity gains, like a 30% uplift at Salesforce, and the whole market is exploding. But the actual evidence? It’s mixed at best. A July study by METR found experienced developers were actually 19% slower using AI assistants, even though they felt faster. Bain & Company also reported in September that real-world savings have been “unremarkable.” So you have this frantic land grab and sky-high valuations built on a promise that the data is starting to poke holes in. Cursor and Graphite are basically betting that the problem isn’t the AI itself, but the messy human process around it. If writing code gets 50% faster but reviewing it still takes just as long, well, you’ve just moved the bottleneck. That’s the logic here.
The Platform Play and Model Wars
This is a classic consolidation move. Cursor owns the “write” phase, Graphite owns the “review and merge” phase. Combine them, and you have a sticky platform that controls more of the developer’s workflow. It’s smart business. But it also highlights a fascinating tension in the AI tools space. Truell brushes off the threat from the big AI labs like OpenAI and Anthropic, saying Cursor just uses the best models (like Claude) and wraps them in a superior UI. Lutsky echoes this, saying the big model companies are spread too thin. That’s the party line. But you have to wonder how stable that is. If the underlying model is the engine, and you’re just the car body, what’s your moat when the engine maker decides to build their own car? Cursor’s answer seems to be deep workflow integration and owning the entire process from keystroke to deployment. It’s a bet that the application layer, not the model layer, is where the real value will be captured for developers.
The Quality Question Doubles Down
Probably the most interesting quote in the whole piece is from Lutsky: “We’ve invested deeply in ensuring that code written with the help of AI is safe and high quality.” That’s the elephant in the room, right? If AI is churning out code faster, but that code is buggy, insecure, or just weird, then you’re not saving time—you’re creating a technical debt factory. Graphite’s entire value is supposedly automating and improving review to catch that. So this acquisition is, in a way, Cursor buying an insurance policy for its own core product. They’re acknowledging that accelerating code creation is only valuable if you can also guarantee the output’s integrity. It’s a necessary move, but it also subtly admits a weakness in the pure “code generation” sales pitch.
What’s Next? An IPO or More Deals?
Truell says no IPO plans and no other deals on the immediate horizon. They want to build. I believe the “no other deals” part for about five minutes. When you’re sitting on a $29 billion valuation and a war chest, and you’re talking about automating the entire software development lifecycle, how can you *not* be looking at the testing, deployment, and monitoring phases? This feels like the opening move. They’ve connected writing and reviewing. The logical next steps are all the stages that come after. The race isn’t just to be the best AI coding assistant anymore. It’s to be the operating system for software engineering. That’s the decade-long ambition Truell mentions. And for companies managing complex industrial systems where software meets hardware, the reliability of this entire AI-driven pipeline will be critical. The stakes for clean, validated code are even higher there. Now, Cursor and Graphite are trying to build the platform that promises to deliver it.
