OpenAi intends to start manufacturing of its own chip in the coming months [RUMOR]

OpenAi would be close to finishing the design for its own acceleration chip. According to Reuters sources, the company should enter the “Tape Out” phase in the coming months – that is, send the project for manufacturing. And in this case, TSMC will be responsible for the production of chips, according to the publication.

Like every Big Tech involved in AI, OpenAi is concerned with reducing its dependence on Nvidia for technology processing. Rumors have been circulating for a long time that the company would be investing in creating its own chip to achieve this.

Now we have information that there is a dedicated team within the company for this purpose, led by Richard Ho. In recent months, this team would have doubled in size, reaching 40 people and working in collaboration with Broadcom in developing the new hardware.

Expert estimates point out that completely new hardware for AI -scale – of the level that OpenAI would need – could cost around $ 500 million only for the first version of the chip. Software implementation and development costs could double this budget.

TSMC will manufacture the OpenAi chip in the 3NM process

The unofficial information still points out that the OpenAi chip will be manufactured in the TSMC 3NM process. If the “Tape Out” actually happens in the coming months, there are predictions that at the end of the year the mass production of the component will begin. This is if not unforeseen and the first version of the design does not have problems.

Related News:

As might be expected, the OpenAi accelerator will use HBM memories, which have been the favorite for AI -facing components.

Possibly the company did not plan a chip for the TSMC 2NM process due to the costs of the manufacturer’s most advanced lithography.

Via: CNBC