OpenAI unveils Jalapeño, its first LLM inference chip with Broadcom

On June 24, 2026, OpenAI and Broadcom (NASDAQ: AVGO) unveiled Jalapeño, OpenAI's first "Intelligence Processor" — an accelerator architected from scratch for large language model inference rather than adapted from earlier AI workloads.

Key points

• Built for LLM inference: the architecture is optimized around the kernels, memory movement, networking and serving patterns of OpenAI's models; engineering samples already run ML workloads in the lab, including GPT-5.3-Codex-Spark.
• Nine-month tape-out: OpenAI calls it what it believes is the fastest ASIC development cycle ever achieved in high-performance semiconductors, partly accelerated by using OpenAI's own models in the design process.
• Performance per watt: early testing shows performance per watt substantially better than current state-of-the-art; a detailed technical report is promised later.
• Partners & scale: Broadcom provides silicon and networking (including Tomahawk), Celestica handles board and rack systems; initial deployment targets the end of 2026, expanding to gigawatt scale with Microsoft and other partners.
• Why devs care: owning more of the stack aims to cut cost and latency for ChatGPT, the API and Codex.