A new report claims that before Amazon or Google reveals anything about their latest artificial intelligence chips, they first notify Nvidia CEO Jensen Huang. The practice, touted by sources in The Information’s recent report on Nvidia’s internal dealmaking, reflects the quiet reality of the AI hardware market: Nvidia is still the dominant supplier of training compute, and its customers are trying not to get cut off.
According to the report, Amazon and Google each provide advance notice to Huang before unveiling updates to their custom silicon. The reason, sources say, is that Nvidia is still deeply embedded in their cloud operations and neither wants to surprise the person who effectively controls their AI infrastructure supply. Nvidia accounts for the overwhelming majority of the accelerators used to train large language models, and its GPUs also handle a growing share of inference tasks in the public cloud.
The scale and timing of those investments show how aggressively Nvidia is trying to preempt the rise of non-GPU accelerators. Companies like Amazon, Google, and OpenAI are all pursuing in-house silicon efforts designed to reduce dependency on Nvidia hardware and improve performance or cost at scale. But even with years of effort and tens of billions of dollars behind them, their platforms are still heavily reliant on Nvidia’s CUDA ecosystem, software tooling, and manufacturing pipeline.
According to The Information, Nvidia’s dominance has created a dynamic where it acts like a financial backstop to the entire AI supply chain. The company can now fund suppliers, rent out capacity, and underwrite long-term purchases to support continued demand for its hardware. That makes it harder for any individual customer to walk away, even as they build competing products.
