OpenAI just unveiled its first custom AI chip. It's called Jalapeño. Built from scratch by OpenAI and manufactured by Broadcom on TSMC's 3nm process. It went from initial design to fabrication readiness in nine months, which OpenAI says is the fastest ASIC development cycle ever achieved in high-performance semiconductors. They used their own AI models to accelerate parts of the chip design process. This is not a general-purpose GPU. Jalapeño is an Application-Specific Integrated Circuit (ASIC) purpose-built for LLM inference, the workloads that power ChatGPT, Codex, the API, and future agentic products. Bloomberg reports it cuts inference costs by roughly 50% compared to current Nvidia GPUs. OpenAI says early testing shows performance per watt "substantially better than current state-of-the-art," and engineering samples are already running ML workloads in the lab at production target frequency and power, including GPT-5.3-Codex-Spark. The strategic significance here is massive. OpenAI burned through $34 billion in operational expenses in 2025 while generating $13 billion in revenue. R&D costs alone hit $19 billion, and the company paid Microsoft over $10 billion just for compute infrastructure. Custom silicon that cuts inference costs in half is not a vanity project. It's a survival play ahead of a planned IPO. OpenAI is now building the full stack from products to models to chips, joining Google (TPUs), Amazon (Trainium), Microsoft (Maia), and Meta (MTIA) in the custom silicon race. The company plans to deploy Jalapeño at gigawatt scale with data center partners by end of 2026, with multiple chip generations planned. This does not replace OpenAI's existing chip partnerships. Nvidia invested $30 billion into OpenAI in February as part of a $110 billion funding round. Amazon committed $50 billion and 2 gigawatts of Trainium capacity. AMD and Cerebras deals remain in place. Jalapeño gives OpenAI its own seat at the silicon table rather than replacing the chairs it already has. Greg Brockman on CNBC this morning: "This is a real performance improvement on performance per watt and performance per dollar." The AI infrastructure arms race just added another player building their own weapons.