Windsurf, 2024: the editor that broke a $3 billion deal
On November 13, 2024, a VS Code fork called Windsurf launched publicly with a feature its developers called Cascade. Previous AI coding assistants behaved like contractors who showed up and immediately started moving furniture. Cascade introduced itself first: here are the files I am going to modify, here is why, here is the order. Then it started. You could interrupt at any step. That single difference in protocol, it turned out, was worth about three billion dollars.
Varun Mohan and Douglas Chen had been childhood friends before they were MIT classmates, and MIT classmates before they were co-founders of a company called Exafunction, which in 2021 did GPU optimization for inference pipelines. By 2022 they had looked at what GPT-4 could do with code, renamed the company Codeium, and switched their entire team to building AI coding tools. By the time Windsurf launched in November 2024, Codeium had around 700,000 active users, a thousand enterprise customers, and $243 million in total funding — the last $150 million raised at a $1.25 billion valuation in August of that year.
The editor was a fork of Visual Studio Code — the same pragmatic starting point Cursor had chosen a year earlier. What differed was the AI architecture underneath. Windsurf’s Cascade maintained what the team called a continuous context: it tracked what had changed across the project session, what had been tried, what had failed, what remained. Cursor required the developer to explicitly point the AI at relevant files via mentions and scoping. Cascade explored the codebase implicitly, found what it needed, wrote out a plan, and then executed it — pausing at decision points rather than running to completion and requesting final approval afterward. It felt less like autocomplete and more like a junior developer doing prep work before a code review.
Then came the acquisition. In April 2025, CNBC reported that OpenAI was in talks to buy Windsurf for approximately three billion dollars — which would have been the largest acquisition in OpenAI’s history. By May, Bloomberg reported the agreement was signed. It collapsed over IP rights: under the terms of OpenAI’s partnership with Microsoft, technology acquired by OpenAI was expected to flow to Microsoft’s product division. The Windsurf team had no interest in their work going to GitHub Copilot, and OpenAI declined to route it there. The exclusivity period expired.
In July 2025, Google announced a $2.4 billion agreement structured carefully to sidestep the same problem: not an acquisition but a hiring deal. Varun Mohan, Douglas Chen, and roughly forty engineers joined Google DeepMind to work on Gemini’s coding capabilities. Google took no equity stake in Windsurf. The company continued operating under new leadership with its existing product intact — the founders’ exit structured so that neither OpenAI nor Microsoft could claim a stake in what they had built.
Within eight months of its launch, three of the most valuable companies in the world had made formal offers to own Windsurf. The tool had become the territory.
Sources
- Windsurf (software) — Wikipedia — founding history, Codeium background, Cascade overview, November 2024 launch.
- GitHub Copilot competitor Codeium raises $150M at a $1.25B valuation — TechCrunch — Series C round, company scale at launch.
- OpenAI in talks to pay about $3 billion to acquire AI coding startup Windsurf — CNBC — acquisition talks, deal structure, IP rights conflict with Microsoft.
- OpenAI Reaches Agreement to Buy Startup Windsurf for $3 Billion — Bloomberg — May 2025 confirmation that the deal was agreed before it collapsed.
- Google hires Windsurf CEO Varun Mohan in $2.4 billion deal — CNBC — Google acquihire details, no equity stake, Windsurf independence.