OpenAI's first chip runs at 700 watts and just posted benchmarks beating Nvidia's 1,400 watt flagship. The results dropped one week after Nvidia agreed to backstop $105 billion of OpenAI's data center financing. Jalapeño delivered 1.5 to 1.9x more throughput per kilowatt than Nvidia's GB200 and GB300 racks, and up to 3.6x lower latency, on a public benchmark suite anyone can rerun. The timeline is the part chip people are losing their minds over. 16 months from first hire to tape-out. The…