OpenAI is working with TSMC and Broadcom to build its in-house AI chips, and has reportedly started using AMD chips alongside Nvidia chips to train its AI.
According to Reuters, OpenAI has abandoned plans to establish a network of factories for chip manufacturing, at least for now. Instead, the company plans to focus on in-house chip design. OpenAI has been working with Broadcom for several months to develop an AI chip to run the models, which could arrive as early as 2026, Reuters reports.
OpenAI, on the other hand, plans to use AMD chips through Microsoft's Azure cloud platform to train its models. According to Reuters, the company had previously relied almost entirely on Nvidia GPUs for training, but chip shortages and lag, as well as the high cost of training, led OpenAI to explore alternatives. That's what it means.