Elon Musk has unveiled a new supercomputer for his artificial intelligence startup, xAI. Named Colossus, the system is powered by 100,000 Nvidia AI chips, making it the most powerful AI training system on the planet.

The data center, built in Tennessee, came online over the Labor Day weekend, just 122 days after construction began — a record, according to Nvidia. “Colossus is the most powerful AI training system in the world,” Mr. Musk wrote in a post on X.

The supercomputer utilizes Nvidia H100 graphics processing units, which are among the most coveted hardware for training and running generative AI systems, including AI chatbots and image generators.

Mr. Musk said Colossus is set to expand even further. “In a few months, Colossus will double in size to 200,000 AI chips,” he said, including 50,000 H200 GPUs, a newer version with nearly double the memory capacity and 40 percent more bandwidth than its predecessor.

Founded just last summer, xAI’s premier product is Grok, an AI chatbot integrated into the X social media platform. The rapid assembly of Colossus highlights xAI’s ambition to match the hardware capabilities of established tech leaders like OpenAI and Microsoft.

According to Fortune, Nvidia regards Mr. Musk as one of its best customers. Prior to xAI, Mr. Musk had already purchased tens of thousands of GPUs for Tesla — worth approximately $3 billion to $4 billion. Some of those chips, intended for Tesla’s Full Self-Driving system, were repurposed to train an early version of Grok.

Colossus’s title as the most powerful AI training system is already under threat. Other AI leaders, like OpenAI, Microsoft, Google, and Meta, are not likely to rest on their laurels. Microsoft, for one, aims to amass 1.8 million AI chips by the end of the year, though this goal seems optimistic. Meanwhile, Mark Zuckerberg has signaled that Meta intends to purchase an additional 350,000 Nvidia H100s by the same deadline.