1 million GPUs! Musk plans to expand the Colossus AI supercomputer cluster tenfold, and the AI infrastructure arms race continues

The "burning money" competition in AI does not seem to be slowing down, and now Musk has added fuel to the fire. According to the media, Musk's AI startup xAI plans to expand its Colossus supercomputer tenfold, ultimately integrating over 1 million GPUs to surpass competitors like Google, OpenAI, and Anthropic. The Colossus, built in just three months earlier this year, is considered the largest supercomputer in the world, containing over 100,000 interconnected NVIDIA GPUs. These chips are used to train Musk's chatbot Grok. Given the enormous number of GPUs, the cost will be extremely high, with the latest generation of NVIDIA GPUs typically priced at tens of thousands of dollars. Therefore, the investment required for Musk's plan to expand Colossus is expected to reach hundreds of billions of dollars, while also incurring significant costs for construction, power supply, and cooling. However, xAI has raised about $11 billion from investors this year. Compared to the collaboration model of startups and major AI companies, such as OpenAI with Microsoft and Anthropic with Amazon, Musk prefers to rely on his own power and influence in the tech field to directly build supercomputing capabilities. Musk started later, with xAI being established just over a year ago. But it is undeniable that xAI is developing rapidly, currently valued at $45 billion, and recently raised $5 billion in funding. Jensen Huang stated in October that "there is only one person in the world who can do this." He described Colossus as "without a doubt the fastest supercomputer" and noted that data centers of such scale typically take three years to build. However, it is worth noting that Colossus has also faced controversy for its rapid construction speed. Opponents argue that it bypasses planning permissions and criticize the burden it places on the local power grid. In response, xAI's senior manager for site construction and infrastructure, Brent Mayo, stated: We are not just leaders; we are pushing progress at an unprecedented speed while ensuring grid stability