The infrastructure for the Colossus 2 cluster will undergo a series of hardware expansions starting next week to increase training capacity for xAI. Elon Musk plans to add 220,000 Nvidia GB300 GPUs next week, followed by another 220,000 in November and a final 220,000 in late December if the timeline holds. This Q4 expansion of up to 660,000 units joins an existing fleet of 110,000 GB200s and 440,000 GB300s in Colossus 2, and a first cluster consisting of 150,000 H100s, 50,000 H200s, and 30,000 GB200s.
The final quarter rollout represents an investment of over $50 billion, bringing the total value of the combined GPU fleet to more than $100 billion with 2 GW of compute power. To minimize the facility footprint, the newest hardware is being deployed at a Minihard GPU facility using a denser configuration. xAI is building this internal capacity to power the Grok model, avoiding the cost and constraints of renting compute from third parties.
Key sources
- SOURCE@elonmusk“Another 220k GB300 will be fully operational next week and another 220k in November”x.com
- SUPPORT@sawyermerritt“possible total of 660,000 GB300s coming online in Q4 alone”x.com
- SUPPORT@sawyermerritt“over $100 billion of GPUs combined, and brought online in record time. Over 2 GW of compute”x.com
- SUPPORT@jun_song“The biggest difference? They do not rent their compute. They own it.”x.com
- SOURCEmarketbrief.now