OpenAI Jalapeño chip signals in-house AI compute push vs Nvidia

OpenAI has unveiled its first in-house AI chip, **Jalapeño**, developed with **Broadcom** for **LLM inference** workloads powering ChatGPT, Codex, and future agentic products. OpenAI says the **Jalapeño** chip was completed in **nine months** as part of a “full-stack” strategy to reduce reliance on third-party hardware, cut operating costs, and gain more control as AI demand scales. The move intensifies the competitive pressure on **Nvidia**, which still dominates AI training and inference accelerators. While OpenAI does not claim Jalapeño replaces every accelerator, the message is clear: the company wants greater flexibility in AI compute procurement as it scales. Separately, OpenAI expanded **ChatGPT Enterprise** with **BBVA**, increasing rollout from 11,000 to 120,000 employees across 25 countries, targeting customer service, risk analysis, software development, and internal operations. It also deepened ties with **Visa** via a strategic agentic-commerce partnership. For markets, the chip launch strengthens **IPO speculation**. Sam Altman has suggested a potential offering within a year, and Coinbase introduced OpenAI-linked **pre-IPO futures**, giving traders a way to express valuation expectations. Overall, **Jalapeño** is a tech-sector/AI-infra story with potential sentiment spillovers, not a direct crypto catalyst.