The Chip King Who Almost Owned the Labs

Jensen Huang has a confession. The CEO of the company that sells the chips powering the AI boom told the Dwarkesh Patel podcast this week that missing the OpenAI and Anthropic seed rounds was his "biggest mistake." What makes it more than personal regret: the barrier he missed was not a funding round he overlooked — it was a $5 billion to $10 billion compute threshold that locked traditional venture capital out of transformative AI before the field existed.

Google and Amazon had the capital. Nvidia did not. The cloud providers who backed Anthropic structured their investment as compute credits — upfront payments for cloud usage that tied Anthropic to their own custom processors. Anthropic runs substantially on Google TPUs and Amazon Trainium, the chips each company built for its own AI workloads. "Google and AWS were," Huang said on the podcast. "They put in huge investments in the beginning so that Anthropic, in return, used their compute."

The dependency is not historical. Google and Amazon are co-investors in OpenAI's latest $122 billion round — Amazon committed up to $50 billion, Google participated — and both still run the inference workloads that are the labs' largest ongoing expense. Nvidia, which built the GPU infrastructure these labs depend on, is now buying equity at valuations set by its competitors. In February, Nvidia invested $30 billion in OpenAI as part of that round, according to CNBC. In November 2025, it committed up to $10 billion to Anthropic, according to Business Insider. That $40 billion buys a seat at a table where the hosts set the terms.

Huang said he will not make the same mistake again. What remains unclear is whether owning equity gives Nvidia influence over which chips the labs choose — or whether it simply makes the chipmaker a welcome co-investor in infrastructure it helped make indispensable.