Elon Musk Plans to Scale xAI Cluster to 1.1 Million NVIDIA GB300 GPUs

Elon Musk revealed plans to scale SpaceXAI's Colossus 2 computing cluster to 1.1 million NVIDIA GB300 GPUs by the end of the year. He stated that despite starting only three years ago, xAI aims to reach an industry-leading position within six months. This unprecedented infrastructure scale underscores the intensifying AI compute arms race, where raw hardware power drives frontier model capabilities. If successful, xAI could rapidly close the multi-year advantage held by competitors like OpenAI and Anthropic while accelerating development of future Grok models. The existing Colossus 1 cluster operates 150,000 H100, 50,000 H200, and 30,000 GB200 GPUs, while Colossus 2 currently houses 110,000 GB200 and 440,000 GB300 GPUs. The deployment plan calls for adding 220,000 GB300 GPUs next week, followed by two additional batches of 220,000 units to reach the target.