OpenAI reportedly seeking alternatives to Nvidia chips, unhappy with inference performance

Reuters reports that OpenAI is “unsatisfied” with Nvidia’s latest AI chips and has been seeking alternatives since last year, citing a whopping eight sources familiar with the matter.

This news comes on the heels of a recent report from The Wall Street Journal that Nvidia’s plan to invest $100 billion in OpenAI had stalled.

Nvidia CEO Jensen Huang seemingly confirmed the WSJ reporting in comments to the press over the weekend, but still struck a positive public tone on OpenAI, indicating that the chip designer would be participating in its upcoming funding round.

Sources inside OpenAI appear to be choosing a more combative response.

Per Reuters, the specific shortcoming OpenAI sees in Nvidia’s offering involves inference, or the “thinking” being done by AI models.

Now, the idea that OpenAI is seeking alternatives to Nvidia, or at least additional sources, is well known: the ChatGPT maker struck highly publicized deals in October with Advanced Micro Devices that the chip designer said would “deliver tens of billions of dollars in revenue” as well as custom chip specialist Broadcom to develop and deploy 10 gigawatts of custom AI accelerators.

So this really has the feel of, “I dumped her, she didn’t dump me!”