The Anthropic revealed an annualized run-rate revenue of $30 billion and plans to consume 3.5 gigawatts of next-generation Google TPU chips — and that single paragraph already captures well the stage we're at in the AI race: absurd numbers announced with the casualness of someone talking about the weather. The most revealing detail in the story, however, isn't Anthropic's numbers. It's who felt the need to protect themselves: Broadcom included an explicit note in its regulatory filing stating that the consumption of all that computing capacity depends on Anthropic's "continued commercial success." In corporate language, that's the equivalent of saying "we bet big on them, but, let's be honest, we're not guaranteeing anything." It's rare to see a supplier flag the risk posed by its own customer in a regulatory document — which says a lot about the scale of the commitments involved and the uncertainty that still surrounds even the industry's biggest bets. Anthropic responded to the skepticism by disclosing that its run-rate revenue grew from approximately $9 billion at the end of 2025 to over $30 billion now — more than 3x growth in a matter of months. The number of enterprise customers spending over $1 million annualized doubled to more than 1,000 in under two months. These are metrics that, if accurate, justify the optimism. If inflated by the current hype cycle, they justify exactly the caution Broadcom is showing. At its core, what this story illustrates is the structure of mutual dependency taking shape: Anthropic needs Google's and Broadcom's infrastructure to scale; Broadcom needs Anthropic to succeed for the investment to make sense; and Google profits in every scenario — as a cloud provider, as an investor, and as a competitor with Gemini. The game is far from simple, and the $30 billion run-rate is as much an argument as it is a promise.