Windsurf (formerly Codeium) vs Alphabet Inc. (Google)

Varun Mohan and Douglas Chen met in middle school in Sunnyvale, California, reconnected at MIT, and co-founded Exafunction in June 2021 — a company that optimised GPU utilisation for large-scale deep learning workloads. It was genuinely useful: Mohan had managed deep learning infrastructure at Nuro (autonomous vehicles), Chen had built VR software tools at Meta. They raised $28 million and scaled Exafunction to $2 million in annual revenue. For most founders, this would have been the beginning of a comfortable company-building journey. In 2022, they recognised a fatal flaw: generative AI would make GPU optimisation software a commodity product within years. They killed the business and started over.

The pivot became Codeium: an AI-powered code completion tool that launched in October 2022. The strategic differentiation was a deliberate free tier for individual developers, building grassroots adoption within engineering teams before converting to paid enterprise contracts — a bottom-up SaaS motion similar to Figma and Slack. Codeium's free tier was genuinely unlimited, with no paywalled features that mattered for individual use. Enterprise customers paid for self-hosted deployments and compliance features. By early 2025, Codeium had reached approximately $40 million in annualised recurring revenue and attracted 350+ enterprise customers including Anduril, which required a self-hosted deployment for national security reasons.

In November 2024, Codeium launched the Windsurf Editor — an AI-native integrated development environment built on a fork of Visual Studio Code. The defining feature was Cascade: an AI agent that could understand entire codebases, coordinate changes across multiple files simultaneously, execute terminal commands, and maintain persistent context about a project's architecture. Where GitHub Copilot and other AI coding tools suggested completions within a single file, Cascade could be told "add authentication to this app" and execute the changes across every affected file autonomously. The launch generated 10,000 signups in the first two days and hundreds of thousands within weeks. ARR grew from $40 million to approximately $100 million by April 2025.

In May 2025, OpenAI agreed to acquire Windsurf (Codeium had rebranded in April 2025) for approximately $3 billion — what would have been OpenAI's largest acquisition. The deal collapsed over intellectual property. Under Microsoft's partnership agreement with OpenAI, Microsoft had contractual rights to intellectual property that OpenAI acquired through any transaction. Since Microsoft's GitHub Copilot competed directly with Windsurf, Varun Mohan was unwilling to allow Microsoft access to Windsurf's technology. The exclusivity period on the acquisition expired without resolution. Within hours of the collapse, Google moved.

On July 11, 2025, Google announced a $2.4 billion licensing and talent acquisition deal: Varun Mohan, Douglas Chen, and approximately 40 senior research engineers joined Google DeepMind to work on Gemini's agentic coding capabilities. Google received a non-exclusive technology licence but not equity or control of the company. Three days later, Cognition AI — the startup behind the Devin autonomous coding agent — acquired the remaining Windsurf business: the product, brand, IP, and customer base. Windsurf had raised $243 million total at a $1.25 billion valuation in August 2024; investors received approximately 4x their capital in the two-transaction outcome. The two MIT friends who had killed a profitable GPU business to chase a bigger opportunity had generated billions in value — just not in the way any conventional acquisition model had anticipated.

Larry Page and Sergey Brin met at Stanford in 1995. By 1996 they had built a search engine called BackRub — named after its method of analysing backlinks. It ran on Stanford's servers and consumed so much bandwidth that the university repeatedly asked them to take it down.

Sun Microsystems co-founder Andy Bechtolsheim wrote Google a cheque for $100,000 in August 1998 after a brief demo in a Stanford parking lot. There was one problem: Google Inc. didn't exist yet. Page and Brin had to incorporate the company before they could cash it.

In 1999, Page and Brin tried to sell Google to Excite CEO George Bell for $1 million. Bell turned them down — the search was too good, it would send users away from Excite's portal too quickly. Bell later called it the worst decision of his career. Excite filed for bankruptcy in 2001. Google became worth more than $2 trillion.

When OpenAI launched ChatGPT in November 2022, Google declared an internal "code red." The company that had invented the transformer architecture — the technology that made ChatGPT possible — had published its research openly and was now scrambling to compete with the system it had inadvertently enabled. Google rushed Bard to market. It made a factual error in its first public demo. The stock dropped $100 billion in a single day. Google's search share fell from 80% to under 70% as users migrated to ChatGPT, Perplexity, and Claude.

In September 2025, a U.S. federal court ruled that Google had illegally maintained its monopoly in search, ordering the company to end exclusive distribution contracts with Apple, device manufacturers, and browser makers. The Apple deal alone had been worth approximately $20 billion per year to keep Google as the default search engine on Safari. The court stopped short of forcing Google to sell Chrome or Android, but the ruling was the most consequential antitrust action against a technology company since the Microsoft case of 1998. Google announced it would appeal. The case is expected to reach the Supreme Court.