OpenAI unveils its first custom chip, built by Broadcom
4 hours ago
- #AI Hardware
- #Chip Design
- #OpenAI
- OpenAI unveiled its first custom-built inference processor named Jalapeño, developed with Broadcom.
- Jalapeño is designed for AI inference tasks, showing better performance-per-watt in early tests.
- The chip aims to reduce OpenAI's dependence on Nvidia GPUs, similar to custom AI accelerators from Google and Amazon.
- Jalapeño focuses on low-cost operation for real-time coding models, though pre-training may still rely on Nvidia hardware.
- OpenAI's vertical integration across chip design, infrastructure, and models aims to optimize speed, reliability, and affordability.