Elon Musk reveals plans to expand AI compute power with 1.1 million nvidia gb300 chips by year-end

Elon Musk, CEO of SpaceXAI, announced on September 25, 2026, that his company plans to deploy 1.1 million Nvidia GB300 chips by year-end, with 220,000 units scheduled for deployment in November and December 2026. The GB300 chips, part of Nvidia’s Blackwell architecture, are expected to significantly enhance AI training and inference capabilities. SpaceXAI currently operates two major clusters, Colossus 1 and Colossus 2, with the latter set to receive the additional GB300 chips. Musk stated that if deployment proceeds smoothly, the company could achieve industry-leading AI capabilities within six months. He noted that SpaceXAI, with only three years of AI development, faces competition from established firms like Anthropic and OpenAI, which have 6 to 10 years of experience. Musk emphasized that the expanded compute power will enable the development of more advanced AI products, such as Grok, with improved accuracy, reasoning, and natural language interaction.