OpenAI to Launch First AI Chip with Broadcom in 2026

According to insiders, OpenAI is set to partner with U.S. semiconductor giant Broadcom in 2026 to produce its first in-house AI chip. The chip is intended for internal use only and will not be sold externally.

OpenAI is focused on commercializing generative AI capable of responding to queries with near-human accuracy, relying on massive computing power to train and operate its systems.

Last year, reports indicated that OpenAI had been collaborating with Broadcom and TSMC to develop its first self-designed chip, while also leveraging AMD and NVIDIA solutions to meet surging infrastructure demands. The company explored multiple strategies to diversify its chip supply and reduce costs.

Earlier this February, it was reported that OpenAI was advancing plans to cut dependency on NVIDIA chips by rolling out its first-generation custom AI chip. Sources reveal that OpenAI will finalize the chip design in the coming months, with production slated at TSMC.

Broadcom CEO Hock Tan stated that the company expects "significant growth" in AI revenue for fiscal 2026, following over $10 billion in AI infrastructure orders from new clients—though specific names were not disclosed. Tan also noted that a confirmed order from a new customer last quarter qualified the company as a major AI partner.

Earlier this year, Tan hinted that, beyond its three existing major clients, four potential new customers are "deeply collaborating" on developing their own custom chips.

OpenAI's move joins other tech giants like Google, Amazon, and Meta in creating proprietary AI chips. As demand for the computing power required to train and run AI models continues to skyrocket, companies increasingly turn to custom-designed chips to handle these workloads efficiently.