Well, Nvidia should pay attention. OpenAI just released the first benchmarks for Jalapeño, their custom inference chip. The results are wild. Let me break them down. Tested against Nvidia's GB200 and GB300 systems on three models: GPT-OSS 120B, DeepSeek R1 670B, and Kimi K2.5 1T. Using SemiAnalysis's InferenceX benchmark. Not OpenAI's own numbers. A public, third-party benchmark. The results: – 1.5-1.9 more AI work per watt at peak throughput. – 1.7-3.6x lower end-to-end latency. –…