OpenAI and Broadcom announced today their first dedicated artificial intelligence chip, called Jalapeño, marking ChatGPT maker's entry into AI chip design. The move is part of OpenAI's strategy to control more layers of the infrastructure behind its models, with Broadcom building the chip to OpenAI's specifications.
The chip will be used by OpenAI and its chatbot for the complex computation behind AI model operations, including answering user queries and other applications. OpenAI President Greg Brockman said, "By independently designing larger parts of our infrastructure, we can deliver more AI capabilities with greater efficiency and continue advancing AI to a wider audience."
The companies said in October, after 18 months of joint work, that they planned to develop custom chips together and build server systems based on OpenAI-designed chips starting at the end of 2026. The new chip is an ASIC, or application-specific integrated circuit, which experts say is less flexible than Nvidia's GPUs but cheaper and better suited to defined AI tasks.
OpenAI said development was completed in just nine months, while it also designed major parts of the computing system where the chip will be installed. The companies describe it as an "Intelligence Processor" and say it is the first AI accelerator in a broader platform intended to make advanced AI faster, more reliable, and accessible to more users.
Since launching ChatGPT in 2022 and helping trigger the generative AI boom, OpenAI has been one of Nvidia's biggest buyers of expensive GPUs for training AI models and running large workloads. Rapid growth in users and infrastructure demand is pushing OpenAI to diversify its chip supply, including deals with Amazon Web Services for Trainium chips and with AMD and Cerebras, alongside continued purchases from Nvidia.
