- OpenAI CFO Sarah Friar said the company is diversifying its chip supply chain and treating Nvidia as one of multiple qualified suppliers, calling it good CFO risk mitigation.
- Friar split AI workloads into episodic training, where Nvidia excels, and inference, which rewards purpose-built chips optimized for OpenAI's models with lower cost per token.
- OpenAI is commissioning inference-specific silicon, a move that pressures Nvidia's premium pricing and gross margins as the fastest-growing workload category shifts.
- Nvidia fell 5.9% and Broadcom dropped 7.95% over five sessions, while AMD was roughly flat, as markets priced exposure to OpenAI's supplier diversification.
- OpenAI's $122 billion cash reserve and no IPO deadline remove urgency and dependency that normally support a supplier's pricing power.
Quotes
“We also have a strategy for diversifying our supply chain. And that is just good CFO risk mitigation. You never know when someone's supply chain is going to get gummed up. So you need to have multiple providers.”
“Nvidia is still an incredible platform of accelerators for training. But in jalapeno's case, that is a chip very focused on inferencing because it is set up exactly for our models. So therefore it is very efficient for you.”
“An IPO is just a milestone in the journey. I'm going to keep reiterating that.”