OpenAI to Produce First AI Chip in Partnership with Broadcom, Targeting 2026 Release

OpenAI is preparing to produce its first artificial intelligence chip in collaboration with US semiconductor leader Broadcom, according to a report by the Financial Times citing sources familiar with the matter. The new chip is expected to be ready by next year and will be used for internal purposes, rather than being sold commercially.

This move signifies OpenAI’s effort to reduce its heavy reliance on Nvidia, whose GPUs have long been the dominant hardware standard for training and running large language models. Reports from Reuters earlier indicated that OpenAI is also working with Broadcom and Taiwan Semiconductor Manufacturing Co. (TSMC) on the chip’s development while additionally using processors from AMD alongside Nvidia to meet soaring demand.

An insider close to the project confirmed that OpenAI’s first-generation chip design is nearing completion and will soon be sent to TSMC for fabrication. Broadcom’s CEO Hock Tan stated during a recent earnings call that the company anticipates significant AI revenue growth in fiscal 2026 after securing over $10 billion in infrastructure orders from a new customer widely believed to be OpenAI.

Broadcom has also revealed it is working with multiple new customers on custom AI chips alongside its existing major clients. OpenAI’s decision to create proprietary hardware aligns with other tech giants like Google, Amazon, and Meta, who have developed their own chips to better optimize performance and manage costs related to AI workloads.

The launch of its own AI chip will allow OpenAI to scale future models with greater flexibility and tighter cost control. This move comes amid intensifying competition in the AI industry, where training increasingly powerful models demands vast computing resources. OpenAI has emphasized its long-term strategy to diversify supply chains to avoid dependency on a single chip provider.