Musk: SpaceXAI to Add 660K GB300 GPUs by Year-End
Elon Musk says SpaceXAI will bring 220,000 Nvidia GB300 GPUs online next week, another 220,000 in November, and a further 220,000 by late December, pushing its GB300 total to 1.1 million.
Elon Musk said SpaceXAI will bring 220,000 Nvidia GB300 GPUs online by next week, with another 220,000 set to follow in November. A further 220,000 units could come online by late December if things go well, he added.
The additions would raise SpaceXAI's GB300 count to 1.1 million and its total GPUs in operation to 1,440,000. Colossus 2 already runs 110,000 GB200 and 440,000 GB300 GPUs, while Colossus 1 has 150,000 H100, 50,000 H200, and 30,000 GB200 GPUs.
Colossus 2 uses only Blackwell GPUs, and Colossus 1 was rented to Anthropic for inference because it is inefficient for training Grok. SpaceXAI is three years old, compared with six years for Anthropic and ten for OpenAI.
The company is building a 1.2-GW power plant and has promised to remove unpermitted gas turbines over a year, but faces a lawsuit over running unpermitted gas turbines. Musk said SpaceXAI will grow data center capacity sevenfold by 2027 and aims for 50 million H100-equivalent GPUs by 2030. SpaceX also plans an Orbital Data Center System with a million satellites.
Broadcom said in 2024 that three hyperscale customers aim for a million-GPU target by 2027, according to Tom's Hardware.
Quick answers
How many GB300 GPUs will SpaceXAI add by the end of the year?
Musk said 220,000 GB300 GPUs will be operational by next week, another 220,000 will come online in November, and a further 220,000 could come online by late December if things go well.
How many GPUs does SpaceXAI operate now?
Colossus 2 already operates 110,000 GB200 and 440,000 GB300 GPUs. Colossus 1 has 150,000 H100, 50,000 H200, and 30,000 GB200 GPUs.
What are SpaceXAI's long-term GPU and data center targets?
Musk said SpaceXAI will grow data center capacity sevenfold by 2027 and aims for 50 million H100-equivalent GPUs by 2030.