OpenAI Ships Jalapeño, Its First Custom AI Inference Chip

OpenAI has launched its first custom-designed AI chip, a move that signals the company's intent to control more of its computing infrastructure and reduce dependence on Nvidia's market-leading processors.

The chip, called Jalapeño, was developed in partnership with Broadcom and completed in just nine months. OpenAI designed the processor specifically for inference workloads—the process of running AI models to generate outputs—and claims early testing shows it outperforms current state-of-the-art chips.

A strategic shift toward vertical integration

OpenAI president Greg Brockman framed the chip as part of a broader "full-stack infrastructure strategy" aimed at making compute more abundant and AI services faster, more reliable, and more affordable. By designing more of its own hardware stack, OpenAI can optimize systems for its specific models and reduce the bottlenecks that come from competing with the entire AI industry for access to Nvidia's GPUs.

The company said Jalapeño is the first in a multi-generation computing platform that will roll out later this year and in subsequent years. While OpenAI remains one of Nvidia's largest customers, developing proprietary chips gives the company an alternative path to securing the massive computing power its models require.

Why it matters

OpenAI's chip debut reflects a broader industry trend: major AI companies are no longer willing to rely solely on Nvidia for their computing needs. Amazon, Google, Microsoft, and Meta have all developed or deployed custom AI processors, with Amazon and Google now offering those chips to third-party customers. This shift creates new competitive pressure on Nvidia from its own client base, even as rivals like AMD, Qualcomm, and Cerebras push into the AI chip market. For OpenAI, owning its silicon means greater control over costs, performance optimization, and supply—critical advantages as AI inference demands scale exponentially.

The competitive landscape

OpenAI's move follows a pattern set by cloud hyperscalers and AI-native companies seeking alternatives to Nvidia's high-powered processors. Amazon and Google have begun renting their custom chips to external customers, and Meta has hinted at potentially entering the cloud computing market with its own hardware.

Meanwhile, Nvidia faces intensifying competition not only from its customers but also from traditional chip rivals. AMD is working to capture AI data center market share, while newer entrants like Qualcomm and Cerebras are carving out niches in the AI chip space.

Broadcom shares rose more than 1% following the announcement, according to Yahoo Finance, which first reported the details of the chip launch.