OpenAI's Jalapeño inference chip beats Blackwell by up to 1.9x per watt
shashib · x · 2026-08-30
Following Google, Amazon, Microsoft and Meta, OpenAI now has its own AI chip: Jalapeño, built with Broadcom and designed for inference — answering queries from already-trained models — not training.
OpenAI posted the first numbers last week: versus Nvidia Blackwell, it delivers 1.5–1.9x more work per watt and 1.7–3.6x faster answers. SemiAnalysis saw some of the lab runs and confirmed it is faster on those tests.
For now Jalapeño only runs inside OpenAI's data centers and cannot be bought; training stays on Nvidia. If it matters to you, the impact will show up in OpenAI's prices and reply speed.
More from Infra
- SK Hynix ties new US plant to looser chip equipment exports to China, sources say — pstAsiatech · 2026-10-15
- Perplexity's new embedding models: index with a 9B model, search with 0.6B — antoine_chaffin · 2026-10-15
- Unsloth ships GGUF quants for Nemotron 3.5: Bolt-74B runs in 24GB RAM — danielhanchen · 2026-10-15
- vLLM-Omni ships Day-0 serving for FLUX 3 Image Klein with BF16, FP8, NVFP4 support — vllm_project · 2026-10-15
- NVIDIA's Nemotron 3.5 Super VL lands training and inference support on Applied Compute — ypatil125 · 2026-10-15
- CoreWeave bond yield tops 16% as loans outlast customer contracts by ~2 years — tengyanAI · 2026-10-15