Broadcom and OpenAI Unveil Jalapeño Chip and Move Toward Gigawatt AI Deployments

Overview

• OpenAI and Broadcom unveiled the Jalapeño LLM inference accelerator on June 24, which the companies say went from design to manufacturing tape‑out in nine months and delivered substantially better performance‑per‑watt in early testing.
• Broadcom’s business model is to co‑design bespoke ASICs for single customers rather than sell general‑purpose GPUs, earning design and manufacturing fees for custom chips like Jalapeño.
• Broadcom launched the AI XPV financing platform with Apollo and Blackstone earlier in June, providing an initial $35 billion tranche intended to underwrite multi‑gigawatt AI builds including Anthropic’s plan for more than 1 gigawatt of capacity.
• The companies plan to begin gigawatt‑scale deployments of Jalapeño‑based systems starting later in 2026, but analysts and investors flag execution risks from supply constraints, timing of future chip ramps, and revenue concentration among a few large AI customers.
• Market response has been mixed with share volatility and institutional flows favoring some peers, so the near‑term proof point to watch is Broadcom’s production ramp and customer rollouts that will verify the chip’s claimed efficiency and shape multi‑year revenue growth.