The Equation Series
Menu

Part III: Building What Works

Chapter 14

Distribution and Platform Risk

On 3 June 2025 Varun Mohan, the chief executive of Windsurf, posted that Anthropic had cut "nearly all" of Windsurf's first-party capacity for its Claude 3.x models with less than five days' notice (TechCrunch, 3 June 2025). Windsurf, formerly Codeium, made an AI code editor whose Cascade agent had been built around Claude Sonnet, the model most developers then preferred for coding. Anthropic had launched its Claude 4 models on 22 May without giving Windsurf direct access to them. In April and May, Bloomberg and others had reported that OpenAI, Anthropic's chief rival, had agreed to buy Windsurf for ~$3 billion (reported; never confirmed by either company). Windsurf claimed ~$100 million of annual recurring revenue that April (reported).

Anthropic said it was "prioritizing capacity for sustainable partnerships" (spokesperson, reported). Its co-founder Jared Kaplan later said it would be "odd" to sell Claude to OpenAI, and added that Anthropic was short of compute (reported). Windsurf moved its free and trial users off Claude by default, bought what capacity it could through third-party providers, let users bring their own Anthropic keys and promoted its own SWE-1 models and Google's Gemini 2.5 Pro (reported). Within six weeks the OpenAI deal had lapsed, Google had hired the founders, and Cognition had bought what remained for an undisclosed price. Anthropic's models returned to the product only after that sale (VentureBeat, July 2025, reported).

This chapter is about the platforms an AI product does not control, the model it calls, the device it runs on and the channel through which customers find it, and about why distribution, more than technology, decides who keeps the value an AI application creates.

The preview ends here. The full chapter is in the book.

In this chapter

  1. 14.1Three platforms underneath every product
  2. 14.2Windsurf: when the supplier picks a side
  3. 14.3Jasper: when the supplier becomes the distributor
  4. 14.4Bundles and defaults
  5. 14.5A worked example: pricing a supplier shock
  6. 14.6How an expert reads distribution and platform risk