OpenAI has unveiled its first custom AI accelerator, named Jalapeño. This chip is a product of a partnership with Broadcom and is designed specifically for inference workloads. The initial physical sample of the Jalapeño chip was received on June 24, with plans for its first deployments set for late 2026.
#What Does the Partnership Entail?
The collaboration between OpenAI and Broadcom spans around 18 months and was publicly announced in October 2025. The ambitious aim of this partnership involves the development of 10 gigawatts of bespoke AI accelerators. OpenAI focuses on chip and system design, while Broadcom takes charge of the development, deployment, and networking aspects. The actual production of the chips will be handled by TSMC, a leading semiconductor manufacturer that also produces chips for major companies including Apple, AMD, and Nvidia.
#What is the Rollout Timeline?
The rollout of the Jalapeño chip is planned in phases, with the first deployment expected by late 2026. After the initial launch, scaling operations are anticipated to continue through 2027 and 2028, culminating in a full rollout by late 2029.
#Why is Jalapeño Significant for the Industry?
The Jalapeño chip is not created for training AI model development but is optimized for inference. Training is known to be resource-intensive, requiring substantial computational power to create AI models. On the other hand, inference refers to the stage where these trained models are put to use, such as answering questions or generating content. As AI applications begin to serve hundreds of millions of users, the costs associated with inference are becoming increasingly significant.
Application-specific integrated circuits, or ASICs, like Jalapeño, tackle these challenges by offering specialized performance that is both cheaper and more efficient. Google has successfully implemented this strategy with its Tensor Processing Units, while Meta and Amazon have also developed custom chips for their respective needs.
#What Impact Will This Have on Market Players?
The release of the Jalapeño chip has already made waves in the stock market, particularly for Broadcom, which experienced a notable increase in its share price following reports of a substantial order potentially worth around $10 billion from an undisclosed client. Amid rising demand for AI computational capabilities, there has been a tighter supply of necessary hardware components, affecting production costs along with influencing market dynamics.
The landscape for AI hardware is shifting, with Nvidia facing competition not just from its traditional rivals but also from its own customer base. The future performance of the Jalapeño chip will only become evident after its deployment, which is set for late 2026.