Rumor

OpenAI develops inference AI ASICs with Broadcom while maintaining Nvidia training partnership

Thursday, January 15, 2026 at 07:57 AM

OpenAI continues its relationship with Nvidia for training requirements while collaborating with Broadcom on the development of custom AI ASICs specifically for inference tasks.

Context

Broadcom has secured a landmark deal with OpenAI to develop its first custom AI infrastructure chip, which will be manufactured using TSMC’s 5nm process technology. This partnership marks a strategic pivot for OpenAI as it seeks to reduce its heavy reliance on Nvidia GPUs and lower the massive operational costs associated with training and running large language models. By collaborating with Broadcom, OpenAI joins other tech giants like Google and Meta in leveraging custom silicon to optimize performance for specific generative AI workloads. For investors, this development reinforces Broadcom’s position as the primary beneficiary of the AI ASIC market, where it already co-designs chips for Google’s TPU and Meta’s internal hardware. The project has already secured manufacturing capacity at TSMC, with volume production expected to begin by 2026. This timeline underscores a maturing AI supply chain where major labs are increasingly prioritizing long-term hardware independence and cost efficiency to sustain the next generation of scaling.

Related Companies

TSMC

2330

Broadcom

AVGO