So far OpenAI used a supercomputer made by Microsoft which has 10000 Nvidia’s GPUs for developing generative AI technology. ChatGPT has high operational costs. Each query costs about 4 cents to the company. If these queries grow even to a tenth scale of Google search, OpenAI will require $48.1 billion worth of GPUs to begin with, and about $16 billion worth of chips to remain operational.

Against this backdrop, OpenAI is planning to make its own AI chips. Alternatively, it can acquire any company that facilitates this task. All this is still at the discussion stage.

There is shortage of expensive AI chips.

If OpenAI goes into chip manufacturing, it may seek a tie-up with Nvidia and may tap suppliers beyond Nvidia.

Nvidia commands 80 per cent market share of the chips best suited to run AI applications. There are two considerations for OpenAI — a shortage of the advanced processors to run OpenAI’s software and very expensive operational costs to run the hardware.