Artificial intelligence company OpenAI is considering developing its own AI chips, and has evaluated a potential acquisition target.

Reuters reports that the company has not yet decided whether to make the move, and is also looking at working more closely with other chipmakers including Nvidia.

71_OpenAI_DALLE_2-_A_Stained_glass_window_of_a_semiconductor_chip_2.png
– DCD/DALL·E 2

The company, known for ChatGPT and Dall-E, is seeking to reduce the huge cost of Nvidia GPUs and diversify from the supply-strained company.

OpenAI is also believed to be testing out an early version of Microsoft's unannounced 'Athena' AI chip, with the company relying on Azure for its compute needs.

Google has developed its successful TPU family of AI chips, while Amazon Web Services has its Graviton and Inferentia chips. Meta is also developing its own hardware, but the project has been hit by delays and a prior version was canceled. Intel last month announced that it would build a supercomputer with Gaudi2 AI hardware accelerators that Stability AI would use.

But, beyond small deployments, none of the alternative processors have been able to break Nvidia's stranglehold on the generative AI market, particularly for training models.