OpenAI has got its hands on its first-ever custom chip, with its architecture specifically designed to reduce data movement and balance compute, memory, and networking resources.

Dubbed "Jalapeño," the chip marks the first step in the ChatGPT maker's efforts for chip-dependence, having brought Broadcom on board last October to co-design the hardware. Broadcom CEO Hock Tan and president Charlie Kawwas presented the first of its new accelerator line to OpenAI’s Sam Altman and Greg Brockman this week.

Billed as a “blank-slate design” for inference hardware, it’s specifically optimized for AI workloads and is believed to have gone from initial design to manufacturing tape-out in just nine months.

“Jalapeño is part of our long-term full-stack infrastructure strategy to make compute more abundant, resulting in AI which is faster, more reliable, more affordable for people and businesses, and can be used to solve more important problems,” Altman said in a statement. “By designing more of the stack ourselves, we can serve more intelligence with greater efficiency and keep pushing advanced AI toward broader access.”

OpenAI leaders have long desired their own silicon lines, having previously mulled buying Cerebras – a tidbit only revealed as part of Elon Musk’s spat with the AI lab – while also with Microsoft on its custom AI chip project, codenamed Athena.

According to OpenAI, its product lines were used to inform the design for Jalapeño, including ChatGPT, Codex, and “future agentic products.”

Richard Ho, the former Google TPU lead turned OpenAI hardware chief, said: “We optimized the architecture around the kernels, memory movement, networking, and serving patterns that matter most for frontier AI models. Based on early testing, Jalapeño will efficiently execute our most important workloads close to the hardware’s theoretical limits”

The firm referenced a “full-stack advantage” upon unveiling the chip, with OpenAI having helped develop the Multi-Path Reliable Connected (MRC) protocol for improved AI-centric XPU networking alongside Broadcom, and other big names in the chip space.

That stack looks set to evolve further, with OpenAI describing Jalapeño as “the first step in a multi-generation compute platform” with Broadcom.

“Our collaboration with OpenAI represents a fundamental commitment to scaling the physical infrastructure required for the next decade of AI,” said Tan. “This is just the beginning of a multi-generation roadmap. By co-developing our industry-leading silicon directly with OpenAI, we are enabling the deployment of gigawatt-scale data centers with Microsoft and other partners beginning in 2026.”