Elon Musk's xAI installed 100,000 Nvidia H200 GPUs in 19 days, vastly exceeding typical 4-year deployment expectations.

Elon Musk's xAI team accomplished the rapid installation of 100,000 Nvidia H200 GPUs in only 19 days, a timeline that Nvidia CEO Jensen Huang says normally stretches to four years in data center projects. In just over two weeks, the xAI team completed the transition from "concept" to "gear" compatible with Nvidia, including training the first AI on the freshly constructed supercluster. Jensen Huang characterized Musk's achievement as "superhuman" when sharing the accomplishment with Tesla Owners Silicon Valley.

The entire procedure began with the construction of an enormous X factory to house the GPUs, followed by the installation of electricity and liquid cooling systems across the entire facility to ensure that all 200,000 GPUs could function properly. Getting the hardware and infrastructure supplied and installed in a precise and coordinated manner required a great deal of cooperation between the engineering teams at Nvidia and Elon Musk's organization.

According to Huang, a typical data center would require four years to accomplish what the xAI team completed in nineteen days. The first three years would be devoted to planning, while the last year would be occupied with shipping, installing, and getting everything up and running. Huang describes in great detail the intricate nature of networking on Nvidia's hardware, noting that "the number of wires that goes in one node...the back of a computer is all wires." He emphasizes that networking Nvidia's gear isn't the same as networking regular servers in a data center.

No other firm is likely to replicate this integration of 100,000 H200 GPUs for quite some time, as it has "never been done before" according to Jensen Huang.