Nvidia CEO Jensen Huang lauded Elon Musk for his remarkable achievement of building a supercomputer in just 19 days through his startup xAI. This feat, typically requiring years, highlights Musk's exceptional understanding of engineering, construction, and resource management.
The supercomputer, named Colossus, was built with 100,000 Nvidia H100 chips. Musk had previously mentioned that it took only 19 days from hardware installation to initial training, making it the fastest in the industry. The entire project took 122 days from start to finish, emphasizing its complexity and scale.
Huang noted that creating such a supercomputer generally involves three years for planning and construction, especially given its need for a vast factory equipped with cooling systems and power. The integration of 100,000 GPUs makes Colossus the fastest supercomputer cluster on the planet. It's anticipated to aid Tesla in advancing its autonomous driving technology.
This unprecedented integration of Nvidia GPUs sets a benchmark difficult for others to replicate soon. Despite this success, even Musk faced challenges in acquiring the necessary GPUs. In September, Oracle's co-founder and chairman, Larry Ellison, revealed that he and Musk had to persuade Nvidia to supply them with the GPUs by directly appealing to Huang, which they successfully managed.
