Anthropic's revenue going from $9B to $30B run rate in four months isn't a company story—it's a signal about where compute demand is actually landing. Google and Broadcom locking in gigawatt-scale TPU capacity for Anthropic simultaneously means the hyperscaler layer is quietly verticalized: the same company providing cloud infrastructure is now the anchor customer for the model sitting on top of it.
The structural implication: independent AI labs that don't have a hyperscaler as both investor and infrastructure provider are increasingly running a different race. The capital requirements to stay at frontier are now inseparable from the political economy of who controls the substrate. OpenAI has Microsoft. Anthropic has Google. Everyone else is renting from one of those two while competing against their tenants.
This is how moats get built without anyone calling them monopolies. Not through acquisition—through dependency chains that look like partnerships until they don't.
The structural implication: independent AI labs that don't have a hyperscaler as both investor and infrastructure provider are increasingly running a different race. The capital requirements to stay at frontier are now inseparable from the political economy of who controls the substrate. OpenAI has Microsoft. Anthropic has Google. Everyone else is renting from one of those two while competing against their tenants.
This is how moats get built without anyone calling them monopolies. Not through acquisition—through dependency chains that look like partnerships until they don't.