SpaceXAI to Double Nvidia Chips at Colossus 2 by Year-End

Elon Musk said SpaceXAI plans to nearly double the number of advanced Nvidia AI chips operating at its Colossus 2 data center by the end of 2026, significantly expanding the computing power behind its artificial intelligence operations.

In a post on X early Friday, Musk said the Memphis, Tennessee-based Colossus 2 facility currently operates with 110,000 Nvidia GB200 chips and 440,000 Nvidia GB300 chips.

The SpaceX CEO, whose company includes SpaceXAI as a unit, outlined an aggressive expansion schedule. An additional 220,000 Nvidia GB300 chips are expected to become operational by next week, followed by another 220,000 units in November.

“If we get lucky, yet another 220k GB300 by late December,” Musk said while responding to a post discussing xAI’s computing capacity.

If the additional December deployment materializes, the facility would substantially increase its stock of Nvidia’s high-performance AI processors, highlighting the enormous computing requirements associated with developing and operating advanced artificial intelligence models.

Musk has described the Colossus complex as the “world’s largest AI supercomputer.” Colossus 1 was initially constructed in 2024 to provide the computing infrastructure needed to train Grok, xAI’s flagship generative AI model.

Since then, the Memphis complex has expanded beyond supporting Musk’s own AI projects. Earlier this year, xAI began renting computing capacity from the facility to other major artificial intelligence companies, including Anthropic and Alphabet-owned Google.

The planned expansion also underscores continued demand for Nvidia’s latest AI hardware as technology companies race to build increasingly powerful data centers capable of training and running large AI models.

However, Musk did not provide details about whether Colossus 2 currently has sufficient power and supporting infrastructure to operate all of the additional Nvidia GB300 chips. Power availability has become an increasingly important constraint for large-scale AI data center projects as computing requirements continue to rise.