OpenAI is said to have been seeking alternatives to some of Nvidia’s latest artificial intelligence chips since last year due to dissatisfaction with their performance for specific inference tasks, whilst a planned $100 billion investment from the chipmaker has stalled after months of negotiations, according to people familiar with the matter.
Seven sources told Reuters that OpenAI is not satisfied with the speed at which Nvidia’s hardware processes certain types of problems, particularly inference computing when AI models respond to user queries. The company needs new hardware that would eventually provide about 10 per cent of its inference computing needs, one source said.
The performance issues became especially visible in Codex, OpenAI’s product for creating computer code, with staff attributing some of the tool’s weakness to Nvidia’s GPU-based hardware, one source told Reuters. Sam Altman, OpenAI chief executive, told reporters on 30 January that customers using the company’s coding models would “put a big premium on speed for coding work”.
OpenAI has held discussions with startups including Cerebras and Groq to provide chips for faster inference, two sources told Reuters. However, Nvidia struck a $20 billion licensing deal with Groq that ended OpenAI’s talks with that company, one source said. OpenAI subsequently announced a commercial deal with Cerebras last month.
The Wall Street Journal reported that the September memorandum of understanding for Nvidia to invest up to $100 billion in OpenAI and build at least 10 gigawatts of computing power has not progressed beyond early stages. Jensen Huang, Nvidia chief executive, has privately expressed concern about what he described as a lack of discipline in OpenAI’s business approach and competition from Google and Anthropic, people familiar with the matter told the Journal.
Huang told reporters in Taipei on Saturday that suggestions of tension were “nonsense” and that Nvidia would invest “a great deal of money, probably the largest investment we’ve ever made”, though not $100 billion. Altman responded on social media platform X, writing that Nvidia makes “the best AI chips in the world” and that OpenAI hoped to remain “a gigantic customer for a very long time”.
Nvidia said in a statement that customers continue to choose the company for inference because it delivers the best performance and total cost of ownership at scale.
