OpenAI and Broadcom unveil Jalapeño AI chip for next-generation language models

OpenAI has introduced its first custom artificial intelligence processor, developed in partnership with Broadcom, marking a significant step in the company’s efforts to build more efficient infrastructure for its AI services. Named Jalapeño, the new processor is designed specifically to support the growing demands of large language model (LLM) inference, which powers products such as ChatGPT.

The announcement follows the companies’ October 2025 decision to work together on a custom AI accelerator. The new chip marks OpenAI’s first move into hardware design, giving the company greater control over the technology powering its services.

Jalapeño aims to improve AI performance and efficiency

According to OpenAI, Jalapeño is its “first Intelligence Processor: an accelerator architected around OpenAI’s vision for the future of LLM inference.” The company explained that the chip has been purpose-built to handle the complex processing required to run advanced language models more efficiently.

OpenAI said the processor currently delivers “performance per watt substantially better than current state-of-the-art” AI chips. While the company highlighted these early results, it also noted that final performance testing is still underway. A more detailed technical report outlining the chip’s specifications and benchmark results is expected to be released in the coming months.

The company plans to begin deploying Jalapeño in data centres towards the end of 2026. This initial rollout will allow OpenAI to evaluate the processor in real-world environments while supporting the growing computational needs of its AI platforms.

By developing its own specialised hardware, OpenAI aims to improve the efficiency of operating its language models. Greater control over both software and hardware could help reduce costs, optimise performance and expand the availability of AI services as demand continues to increase.

Nine-month development highlights close collaboration

OpenAI revealed that Jalapeño progressed from its initial design to manufacturing tape-out in only nine months, an unusually short development period for a processor of this complexity. The company credited the rapid timeline to close cooperation between its engineering teams and Broadcom’s semiconductor specialists.

OpenAI stated, “Jalapeño was co-developed from initial design to manufacturing tape-out in just nine months,” adding that the “speed reflects deep software-hardware co-development with OpenAI’s engineering teams, Broadcom’s silicon implementation expertise, and the use of OpenAI models to accelerate parts of the design and optimisation process.”

The use of OpenAI’s own AI models during parts of the chip’s design and optimisation process also demonstrates how artificial intelligence is increasingly being used to assist engineering tasks. This approach may help shorten development cycles for future hardware projects while improving design efficiency.

Broadcom’s experience in semiconductor implementation played an important role in bringing the project from concept to production. The partnership combines OpenAI’s expertise in AI software with Broadcom’s established capabilities in chip design and manufacturing.

Partnership lays the foundation for future AI hardware

Jalapeño is the first processor to emerge from the partnership between OpenAI and Broadcom, but both companies have indicated that the collaboration extends well beyond a single product. They described the project as the beginning of a “multi-generation compute platform” intended to support increasingly powerful AI systems in the years ahead.

The long-term goal of the partnership is to develop hardware that enables advanced AI to be “faster, more reliable, and more accessible to more people.” As demand for AI applications continues to grow across industries, custom-designed processors are becoming increasingly important for delivering high-performance, energy-efficient computing.

For OpenAI, the launch of Jalapeño represents more than a new chip. It signals a strategic shift towards building a larger portion of the technology stack that supports its AI services. Rather than relying entirely on third-party processors, the company is investing in specialised hardware designed to meet the specific requirements of its language models.

If the processor performs as expected during testing and deployment, Jalapeño could help OpenAI improve the speed and efficiency of future AI services while supporting continued expansion of ChatGPT and other AI-powered applications. It also strengthens the company’s position in an increasingly competitive AI infrastructure market, where hardware performance is becoming just as important as advances in software.