OpenAI Introduces Its First In-House AI Chip Developed With Broadcom

OpenAI on Wednesday officially introduced its first custom-built inference processor, developed in partnership with Broadcom. The new chip, named Jalapeño, has been purpose-built to meet the specific requirements of OpenAI’s inference infrastructure. According to the company, OpenAI’s own artificial intelligence models also played a role in assisting with the design and development of the processor.

Although Jalapeño is still undergoing testing, OpenAI says initial results indicate that the chip delivers significantly improved performance per watt compared with today’s leading inference hardware.

The collaboration between OpenAI and Broadcom was formally announced in October, though speculation surrounding OpenAI’s ambitions to develop proprietary chips had circulated for much longer. The move is widely viewed as part of the company’s strategy to reduce its reliance on Nvidia’s graphics processing units (GPUs). Other major technology companies, including Google and Amazon, have already developed their own AI-focused processors—commonly referred to as AI accelerators—to optimise machine learning workloads and reduce dependence on third-party hardware.

Shortly after the Broadcom partnership was announced, OpenAI President Greg Brockman discussed the company’s chip strategy during OpenAI’s in-house podcast.

“We have a deep understanding of the workload,” Brockman said during the episode. “We’ve really been looking for specific workloads that are underserved, [and asking] how we can build something that will be able to accelerate what’s possible?”

Jalapeño has been specifically engineered for inference—the stage where already-trained AI models process user requests and generate responses. In its announcement, OpenAI highlighted the processor’s ability to deliver lower operating costs while running real-time coding models. More computationally demanding workloads, such as training new foundation models, are still expected to rely largely on Nvidia hardware. However, even modest improvements in inference efficiency could substantially reduce operating expenses and strengthen the company’s long-term economics.

Improving inference performance is expected to become one of the most important competitive factors across the AI industry. OpenAI is already developing AI agents such as Codex, creating increasingly advanced models, and building the data centre infrastructure required to support them. By expanding into custom chip development, the company is extending its control over another critical layer of the AI technology stack.

As OpenAI explained in its announcement, “OpenAI is not only developing frontier models or building products on top of them; it is designing the infrastructure underneath them: chip architecture, kernels, memory systems, networking, scheduling, deployment systems, and product experience.”

The company added that controlling multiple layers of its technology stack allows it to optimise the entire system toward a common objective.

“Because OpenAI operates across the stack, each layer can be optimised around the same goal: making its models faster, more reliable, and more affordable for users,” the company said.