Elon Musk has told investors that his artificial intelligence stratagem xAI aims to create a supercomputer to power the next version of the Grok chatbot. Reuters writes this with reference to a report from The Information.

According to Musk, approximately 20 thousand NVIDIA H100 GPUs were used to train the second version of the Grok chatbot, while the third and subsequent versions will require more than 100 thousand of these chips,

In order for such a chatbot to fully reveal itself, xAI intends to release a supercomputer, which is currently planned for the fall of 2025.

It is reported that xAI may work on this computer in partnership with Oracle, one of the largest companies engaged in the development of database management systems and tools for database development.

When the computer is ready, NVIDIA’s chip groups will be at least four times the size of the largest GPU clusters that exist today. The H100 chips are the most popular in the AI market, so there is a lot of demand for them right now, which could cause problems for such ambitious plans.

Elon Musk launched xAI in July 2023 as an alternative to OpenAI and other companies that have rapidly started developing in the field of artificial intelligence.