According to PANews, Elon Musk's artificial intelligence startup, xAI, has concluded negotiations with Oracle to expand their existing agreement and lease Nvidia chips. Musk confirmed that xAI is building a supercomputer with 100,000 Nvidia H100 GPUs, which is expected to be delivered and start training later this month. He stated that this would be the most powerful training cluster globally, with a significant lead. 'Our core competitiveness is to be faster than any other AI company, which is the only way to narrow the gap,' Musk said.
Previously, xAI had leased 24,000 H100 chips from Oracle and used them to train Grok 2. Musk stated that Grok 2 is currently in the final stages of polishing and only needs minor adjustments and bug fixes before it can be released, which is expected to happen as early as next month. In May, there were reports that the two companies were close to reaching an agreement to expand their cooperation, with xAI expected to spend approximately $10 billion to lease cloud servers from Oracle for several years.