OpenAI has sought alternatives to Nvidia's inference chips since last year, Reuters reports, after finding some Nvidia hardware too slow for specific inference tasks such as code generation; it has struck deals with AMD and Cerebras and held talks with Groq. Negotiations over Nvidia's proposed up-to-$100 billion investment in OpenAI have stalled as OpenAI shifts toward SRAM-heavy chips for faster inference, and Nvidia's $20 billion licensing deal with Groq reshaped options.
Key Points
- 1Shifts hardware: OpenAI seeks non‑Nvidia inference chips (AMD, Cerebras, Groq) since last year
- 2Cites inference latency and memory limits; SRAM‑heavy chips promise faster responses for coding workloads
- 3Indicates procurement and architecture shifts; practitioners should evaluate SRAM-based accelerators and latency benchmarks
Scoring Rationale
Strong sourcing and major industry implications drive score; remaining uncertainty around deal finalization tempers full clarity.
Sources
Public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
