Deep Dive: OpenAI's Custom Chip Jalapeño May Outperform Cerebras
bookwormengr · x · 2026-08-26
SemiAnalysis suggests OpenAI's custom chip Jalapeño significantly outperforms Cerebras in cost and power. An 8-chip Jalapeño node holds 1.7TB memory, enough for a 1.6T parameter FP4 model at 11kW. In contrast, Cerebras requires 20 wafers, costing over $20M and consuming 1MW. Cerebras faces immense pressure as Jalapeño ramps up.
More from Infra
- Every x86 boot replays 40 years of computing history in milliseconds, in 4 stages — lauriewired · 2026-10-16
- engy.ai processes 50B+ tokens in a day as open-model inference demand surges — const_reborn · 2026-10-16
- New inference engine 'e' hits 10,500 tok/s on a single DGX Station — brandon_galang · 2026-10-16
- SGLang ships Day-0 support for FLUX 3 Image Klein: 10 reference images, BF16/FP8/NVFP4 variants — ying11231 · 2026-10-16
- NVIDIA's Nemotron 3.5 Super VL hits CoreWeave: 12B-active multimodal MoE at $0.25/M input tokens — wandb · 2026-10-16
- Chip equipment beats bull case: 2026 WFE seen at $153B, $278B by 2030 — BenBajarin · 2026-10-16