Elon Musk set up 100k Nvidia H200 GPUs in 19 days; normally takes 4 years

Elon Musk and the xAI team have successfully set up a supercluster of 100,000 Nvidia H200 GPUs in just 19 days, a feat that Nvidia CEO Jensen Huang claims typically takes around four years to accomplish. Huang expressed admiration for Musk's ability to coordinate the complex installation process, which included building a new factory equipped with liquid cooling and power systems necessary for the GPUs. The rapid deployment involved transitioning from the concept phase to operational status, including the first AI training run on the new supercluster. Huang noted that the extensive planning and logistics usually required for such a project would take an average data center three years, followed by an additional year for installation and testing. The achievement is considered unprecedented in the industry, with Huang stating that it is unlikely to be replicated soon.

- Elon Musk's xAI team set up 100,000 Nvidia H200 GPUs in 19 days.

- Nvidia CEO Jensen Huang highlighted that this process usually takes four years.

- The project involved building a new factory and complex networking for the GPUs.

- The rapid deployment included the first AI training run on the supercluster.

- Huang described the achievement as unprecedented and unlikely to be duplicated soon.