Startup xAI Ilona Maska launched an AI cluster with 100 thousand. NVIDIA H100 accelerators

Elon Musk announced that the startup he oversees, xAI, has launched the Colossus cluster, designed for AI training. To date, this computing complex includes 100 thousand NVIDIA H100 accelerators, and its capacity will be expanded in the future.

Let us remind you that xAI is implementing a project to create a “gigafactory” for AI tasks. It is expected that this supercomputer will eventually have up to 300 thousand of the latest NVIDIA B200 accelerators. Equipment for the platform is supplied by Dell and Supermicro, and the huge xAI data center is located in the vicinity of Memphis, Tennessee.

«This weekend, the xAI team launched the Colossus AI training cluster with 100,000 H100 cards. From start to finish, everything was done in 122 days. Colossus is the most powerful AI training system in the world,” Musk wrote on the social network X.

Image Source: WebProNews

According to him, the platform’s computing power will double in the coming months. In particular, 50 thousand NVIDIA H200 products will be added. Musk emphasizes that Colossus is not just another AI cluster, it is a leap into the future. The main focus of the project will be on using the power of Colossus to push the boundaries of AI: it is planned to develop new models and improve existing ones. As it scales and matures, the system is expected to become an important resource for the broader AI community, offering unprecedented opportunities for research and innovation.

Launching such a productive cluster in just 122 days is a significant achievement for the entire AI industry. “It’s amazing how quickly this has been accomplished, and Dell Technologies is honored to be part of this important AI training pipeline,” said Michael Dell, CEO of Dell Technologies.