xAI's Colossus 2 could reach 1.21 million Nvidia chips by year-end

xAI could more than double the number of Nvidia chips at its Colossus 2 computing site before the end of 2026, if an aggressive installation schedule holds. Elon Musk wrote that the Memphis-area cluster currently runs 110,000 GB200 and 440,000 GB300 chips, for a combined 550,000.

As quoted in Invezz's report, Musk projected another 220,000 GB300s becoming fully operational the following week and a second batch of the same size in November. He made a third 220,000-chip wave in late December explicitly conditional: “if we get lucky.”

If all three additions arrive and become operational on time, they would add 660,000 chips and bring Colossus 2 to roughly 1.21 million. TokenPost's calculation reaches the same total. That number is a year-end scenario based on Musk's schedule, not the site's current installed count.

The newer chips dominate the expansion

All of the proposed additions are Nvidia GB300 systems, the newer Blackwell generation used for training and running large AI models. Musk separately listed Colossus 1 at 150,000 H100, 50,000 H200 and 30,000 GB200 chips, or 230,000 total.

The scale also raises practical constraints beyond obtaining the processors. Bringing hundreds of thousands of accelerator chips online requires power, cooling, networking and construction to move together. Musk's own qualification on the final batch makes the December total uncertain even before any independent verification of the operating counts.