As it has been reported, OpenAI is working with TSMC and Broadcom so as to design an AI chip on its own, and besides Nvidia's, now makes use of AMD chips too in order to train its AI.

According to Reuters, OpenAI has quashed plans to launch a network of factories dedicated to chip manufacturing at least for now and will instead move chips in-house. OpenAI has been working with Broadcom for months to develop an AI chip for running models, which could launch as early as 2026, reports Reuters.

The opposite is true for OpenAI; it plans to train the models using AMD chips, through Microsoft's Azure cloud platform. The company nearly exclusively used Nvidia GPUs before for training; however, the shortages and delays in delivering chips along with very high cost of training have led OpenAI to find alternative solutions, according to Reuters.