OpenAI unveils Jalapeño chip to cut inference costs

Key takeaway

• OpenAI has unveiled Jalapeño, a custom inference chip developed with Broadcom, which early testing shows delivers better performance-per-watt than existing options.

• By building its own chip alongside models and data centers, OpenAI aims to lower the cost of serving AI responses to users and reduce its dependence on Nvidia.

3 Key Points

FAQ

When will Jalapeño be available to use?

The chip is currently still being tested; no release date or availability timeline is stated in the announcement.

What is Jalapeño designed to do?

Jalapeño is specifically designed for inference—the process of running pre-built AI models in response to user commands. The company has emphasized its low operating cost when running real-time coding models.

Why is OpenAI building its own chip?

OpenAI says it has developed a deep understanding of its inference workload and wants to reduce dependence on Nvidia's GPUs. By controlling the entire stack—models, data centers, and hardware—the company can optimize each layer around the same goal: making its models faster, more reliable, and more affordable for users.