On June 24, 2026, OpenAI and Broadcom jointly announced the Jalapeño inference chip, a custom ASIC designed for large...

On June 24, 2026, OpenAI and Broadcom jointly announced the Jalapeño inference chip, a custom ASIC designed for large language model inference (Source 1, Source 2, Source 3). The chip, built by Broadcom and manufactured by TSMC, achieved tape-out in just nine months using OpenAI's models to accelerate design (Source 4). Early testing shows performance per watt substantially better than current state-of-the-art, with initial deployment targeted by end of 2026 (Source 7, Source 8). This design win is a significant positive for Broadcom, which had seen its stock drop nearly 20% since early June due to a sub-optimal quarter and concerns over Alphabet and ByteDance exploring alternative chip partners (Source 5). The Jalapeño chip validates Broadcom's custom silicon capabilities and provides a strong counter to competitive threats, potentially reversing negative sentiment. Broadcom shares stabilized on the announcement day ([Source 5]). The data center inference market is a key growth area, and this partnership positions Broadcom as a leading provider of custom AI accelerators alongside hyperscaler customers.