Overview
- The Wall Street Journal, citing people familiar, reports Nvidia plans a new processor focused on inference to be unveiled at next month’s GTC in San Jose.
- The platform is said to incorporate technology designed by startup Groq, but Reuters has not verified the claims and Nvidia and OpenAI did not comment.
- Sources previously told Reuters that OpenAI was dissatisfied with certain inference speeds on Nvidia hardware and explored alternatives including Groq and Cerebras.
- One source said a reported $20 billion Groq licensing agreement by Nvidia ended OpenAI’s direct talks with the startup.
- Investor reports note Nvidia shares gained ahead of GTC on expectations of an inference-focused announcement.