xAI’s aggressive infrastructure buildout could more than double Colossus 2 computing capacity by year-end, highlighting continued demand for Nvidia’s latest AI accelerators.
Elon Musk is preparing another major expansion of artificial intelligence computing capacity, with xAI’s Colossus 2 supercomputer potentially reaching roughly 1.21 million Nvidia chips by the end of the year.
According to Musk, Colossus 2 currently operates with 110,000 Nvidia GB200 chips and 440,000 GB300 accelerators. Another 220,000 GB300s are expected to become operational next week, followed by an additional 220,000 units in November. Musk said another 220,000 chips could come online by late December if deployment remains on track.
The expansion would more than double the approximately 550,000 chips currently installed and further establish xAI as one of the world’s most aggressive builders of AI infrastructure.
The company is already operating Colossus 1, which includes 150,000 Nvidia H100s, 50,000 H200s and 30,000 GB200s.
For Nvidia, the scale of xAI’s expansion provides another indication of strong demand for its newest AI hardware, particularly the GB300 generation. Hundreds of thousands of additional accelerators could be deployed within just months.
However, the unprecedented buildout also brings significant costs. Operating infrastructure at this scale requires enormous amounts of capital and power, putting greater emphasis on utilization and investment returns.
Investors will now be watching whether xAI can meet its aggressive year-end deployment schedule and continue expanding beyond the million-chip milestone.
