TL;DR
On September 25, 2026, Elon Musk detailed on X the expansion plan for SpaceXAI's Colossus 2 data center: the Memphis cluster currently runs 550,000 Nvidia GPUs and is planned to grow in three phases to roughly 1.21 million by the end of the year (as disclosed by Musk and reported by multiple media).
Event overview
On September 25, 2026, Elon Musk published on X the latest expansion progress and plans for SpaceXAI's Colossus 2 data center. Per CNR and Sina Finance, Colossus 2 is located in Memphis, Tennessee, USA, and is among the world's largest AI supercomputing clusters; Musk said he plans to push total GPU count to about 1.21 million by the end of the year.
Key facts
- Current scale: Colossus 2 has deployed 110,000 Nvidia GB200 and 440,000 GB300 chips, 550,000 GPUs in total (per Musk's disclosure, relayed by ZOL).
- Expansion plan: 220,000 GB300 chips coming online next week, another 220,000 in November 2026, and another 220,000 by the end of December; total GPU count would reach about 1.21 million by year-end (per CNR and Sina Finance).
- Target: Musk says the expanded Colossus 2 will exceed the original 2026 goal of 1 million GPUs, targeting OpenAI's compute scale (per Sina Finance and ZOL).
- Note on sourcing: the figures and plans above are data disclosed by Musk on social media (per CNR, Sina Finance and ZOL).
Official sources
- CNR (September 26, 2026): Musk: SpaceX data center GPU count to double by year-end
- Sina Finance (September 25, 2026): Musk shows off compute: 660K new GB300 chips
- ZOL (September 27, 2026): SpaceXAI expands Colossus 2 to 1.21M GPUs