Elon Musk’s xAI has reached a major milestone, powering its Colossus supercomputer with 300 MW and scaling up to 200,000 Nvidia GPUs in just over seven months. Built far faster than industry norms, Colossus positions xAI—and potentially Tesla—for a significant AI edge. The company now aims to expand to 1 million GPUs, challenging leading AI infrastructure players like OpenAI, Meta, and Oracle, according to Abhishek Bhardwaj at Interesting Engineering. He writes:
Elon Musk’s intelligence company xAI has achieved a significant milestone with its Colossus supercomputer operating 200,000 Nvidia graphics processing units (GPUs).
The company says that Colossus has been built in just 122 days (around four months), as compared to the estimates of 24 months or two years. xAI then took 92 more days to bring the number of GPUs to 200,000.
It is now planning to increase the number of GPUs in the world’s biggest supercomputer to 1 million GPUs. xAI has a goal to compete with its rivals like the Oracle Cloud Infrastructure, which has approximately 1,31,072 GPUs. […]
Read more here.
