Jump to content

Colossus (supercomputer)

From Wikipedia, the free encyclopedia

Colossus is a supercomputer developed by xAI, an artificial intelligence company owned by Elon Musk, located in Memphis, Tennessee. It is believed to be the world's largest AI supercomputer.[1] Its purpose is to train the company’s AI language model Grok and to serve as a hub for Musk’s other companies including X (formerly Twitter) and SpaceX.[2]


Background

[edit]

The supercomputer was launched in September of 2024, at the former Electrolux site on 3231 Paul R. Lowry Road in Southern Memphis with the purpose of training the AI language model Grok.[2] Reports state that it took Elon Musk and his xAI team 19 days to go from the concept phase to the being ready for Nvidia’s gear. In comparison, other data centers have taken an average of four years to complete the planning of the project, ship the equipment, and have it installed.[3] Both Dell Technologies and Supermicro partnered with xAI to build the computer. It was originally powered by 100,000 GPUs (Graphic Processing Units) and was built in 122 days.[4] These chips were supplied mainly by the company Nvidia, which currently makes the most powerful processing chips on the market. Each GPU is housed within one of Supermicro’s 4U Universal GPU Liquid Cooling Systems to ensure that they do not overheat.[5] 92 days after the first 100,000 GPUs were built,[4] xAI announced that they had increased the size of the computer to 200,000 GPUs and that they intended to continue expanding its size to 1 million GPUs. Colossus is currently the largest AI training platform in the world.[4]



It operates a cluster of more than 200,000[6] interconnected Nvidia GPUs (graphics processing units). It is planned to eventually include more than 1 million GPUs.[7][8]

Its power comes from 150MW of electricity from the Memphis grid, and 420MW from 35 methane gas burning turbines[9], as of April 2025.

See also

[edit]

References

[edit]
  1. ^ Wicks, Myracle; Haywood, Tarvarious; Bolden, Bria (2024-06-05). "Elon Musk's xAI to build multi-billion-dollar supercomputer project in Memphis". https://www.actionnews5.com. Retrieved 2025-04-25. {{cite web}}: External link in |website= (help)
  2. ^ a b "How Memphis became a battleground over Elon Musk's xAI supercomputer". NPR. Archived from the original on 2025-04-13. Retrieved 2025-04-25.
  3. ^ published, Aaron Klotz (2024-10-14). "Elon Musk set up 100,000 Nvidia H200 GPUs in 19 days - Jensen says process normally takes 4 years". Tom's Hardware. Archived from the original on 2025-03-30. Retrieved 2025-04-25.
  4. ^ a b c "Colossus | xAI". x.ai. Archived from the original on 2025-04-18. Retrieved 2025-04-25.
  5. ^ published, Dallin Grimm (2024-10-28). "First in-depth look at Elon Musk's 100,000 GPU AI cluster — xAI Colossus reveals its secrets". Tom's Hardware. Retrieved 2025-04-25.
  6. ^ Hetzner, Christiaan. "Elon Musk's just fired up 'Colossus'—the world's largest Nvidia GPU supercomputer". Fortune. Archived from the original on 2025-02-04. Retrieved 2025-04-23.
  7. ^ Morris, Stephen; Kinder, Tabby (December 4, 2024). "Elon Musk plans to expand Colossus AI supercomputer tenfold". Financial Times. Archived from the original on December 4, 2024.
  8. ^ "Elon Musk's xAI Memphis Supercomputer Eyes Expansion to 1 Million GPUs". PCMAG. Archived from the original on 2024-12-27. Retrieved 2024-12-27.
  9. ^ Kerr, Dara. "Elon Musk's xAI powering its facility in Memphis with 'illegal' generators". Retrieved 22 April 2025.