OpenAI and Broadcom Launch Jalapeño Chip for LLM Inference
Event summary
- OpenAI and Broadcom unveiled Jalapeño, an LLM-optimized Intelligence Processor designed for AI inference.
- The chip was developed in nine months, with engineering samples already running ML workloads including GPT-5.3-Codex-Spark.
- Early testing indicates performance per watt substantially better than current state-of-the-art solutions.
- Jalapeño is the first step in a multi-generation compute platform, with deployment at gigawatt scale planned by end of 2026.
The big picture
This collaboration marks a significant step in OpenAI's strategy to build a full-stack infrastructure for AI, reducing reliance on third-party hardware. The rapid development cycle and focus on LLM-specific optimization reflect the growing trend of vertical integration in the AI semiconductor space. The deployment at gigawatt scale could reshape data center economics, making advanced AI more accessible and affordable.
What we're watching
- Performance Validation
- How Jalapeño's promised performance per watt will compare to competitors once detailed technical reports are released.
- Scalability Challenges
- The pace at which Broadcom and OpenAI can deploy gigawatt-scale data centers with partners like Microsoft.
- Industry Adoption
- Whether Jalapeño's architecture will be adopted by other AI models beyond OpenAI's ecosystem.
Related topics