Xai enhances its Colossus Supercluster with 200,000 Nvidia Hopper GPUs | 5 basic components of cpu | Gpu vs cpu vs graphics card | Gpu hardware list | Turtles AI

xAI is expanding its Colossus Supercluster, bringing the total to 200,000 NVIDIA Hopper GPUs. This supercomputer, already the largest in the world, was built in record time through a collaboration between xAI and NVIDIA. Musk continues to astonish with extraordinary ambitions in the field of AI.

Key points:

• xAI doubles the Colossus Supercluster to 200,000 NVIDIA Hopper GPUs.
• The supercomputer was up and running in just 122 days.
• Jensen Huang praises Musk as “superhuman” in engineering.
• Sales of NVIDIA’s Hopper GPUs are expected to increase.

xAI, the company founded by Elon Musk, is taking a significant step in its technology development with the expansion of the Colossus Supercluster to an impressive 200,000 NVIDIA Hopper GPUs. This supercomputer, already considered the largest in the world, is currently training the Grok language model series. The news of the expansion comes on the heels of xAI’s rapid success in getting Colossus up and running, a process that took only 122 days, an extraordinary achievement compared to the usual timeframe that can extend for months or even years for systems of such complexity. NVIDIA confirmed that xAI was able to start training the Grok LLM in a surprisingly short amount of time: only 19 days after the first rack of H100 GPUs was installed. This achievement is a testament to the efficiency of the collaboration between xAI and NVIDIA, which enabled them to optimize every aspect of the supercomputer implementation. The company’s communication also emphasizes the high reliability of the system, which maintained 95 percent of data throughput, with no problems with latency or packet loss. In a recent interview, NVIDIA CEO Jensen Huang publicly praised Musk, calling him “superhuman” for his exceptional ability to manage large-scale projects, highlighting his unique understanding of the engineering and logistics required to build complex, liquid-cooled systems in record time. In the context of GPU sales, NVIDIA is expected to reach a volume of about 1.5 million units of its Hopper GPUs in the fourth quarter of 2024, before a projected decline in the first quarter of 2025, when sales of the new Blackwell GPUs begin to take off.

In this scenario, xAI and NVIDIA’s innovation and speed of execution could mark a decisive change in the AI landscape.