Washington Weighs AI Inference Chips as China Militarizes Tech

OpenAI’s reveal of a custom inference chip—codenamed “Jalapeño” and built with Broadcom—did more than move the goalposts; it exposed the new battlefield. The AI race has pivoted from model horsepower to control over compute, supply chains, and real-time deployment. Washington now faces a hard question: who governs the silicon that governs the decisions?

In parallel, briefings across D.C. warn that Beijing’s militarized AI push is accelerating—not in the lab, but at the edge: autonomous targeting, ISR fusion, and decision loops measured in milliseconds. Proprietary inference stacks are the fulcrum; export regimes and cloud gateways are the levers. The policy window just snapped open. The thesis: The next 60 days decide whether Washington locks a compute-control standard—aligning industry, allies, and arsenals—or cedes tempo to Beijing’s grey-zone inference surge.