OpenAI has unveiled its first in-house AI chip, a custom processor developed alongside Broadcom and named "Jalapeno."
According to mezha.net, the chip is designed to help OpenAI cut costs and reduce its reliance on Nvidia, whose processors have become the default hardware for training and running large AI systems. By designing silicon tailored to its own needs, OpenAI joins a growing list of major technology companies seeking more control over the expensive chips that power their products.
The partnership with Broadcom is a notable part of the story. As The Globe and Mail reports, the launch has raised questions among investors about whether the collaboration means more upside for Broadcom's stock, which trades under the ticker AVGO. Broadcom specializes in helping large customers build custom chips, making it a natural partner for a company like OpenAI moving into hardware design.
The debut comes with a significant caveat. According to commentary from Wedbush cited by 富途牛牛, large-scale commercial deployment of the chip will require multiple design iterations. In other words, unveiling a first chip is not the same as having one ready to run at massive scale — refining the design through several generations is typically necessary before it can reliably handle real workloads.
The sources provided do not include technical specifications, production timelines, or manufacturing details for the chip.
Why it matters: If OpenAI can eventually deploy its own chips at scale, it could lower the enormous costs of running AI and loosen Nvidia's grip on the market — but Wedbush's warning signals that payoff is still several design cycles away.