🚀 Musk Frenziedly Hammers “Compute Power”! SpaceXAI Colossus 2 Plans to Add 660,000 NVIDIA Blackwell GPUs by Year-End
Elon Musk has once again raised the ceiling of the AI compute arms race. SpaceXAI’s Colossus data center in Memphis is undergoing a massive expansion. Musk said that Phase 2, Colossus 2, will roll out 660,000 of NVIDIA’s latest Blackwell GB300 GPUs ahead of year-end.
📌 Key highlights at a glance
The empire of compute surges: Musk calls Colossus a compute “super factory.” Phase 1, Colossus 1, is currently running roughly 230,000 GPUs (mainly H100, with a small number of GB200). Phase 2, Colossus 2, will fully shift to the more powerful Blackwell GB300 (equipped with Blackwell Ultra GPU cores).
Phased deployment plan:
Current base: 110,000 GB200 and 440,000 GB300 already in operation.
Sprint milestone: This week, 220,000 GB300 will be launched; add another 220,000 by the end of November. It’s expected that the final batch of 220,000 will be fully deployed by the end of December.
By then, the total number of GB300 in Colossus 2 alone will reach 1.1 million—an unrivaled scale of compute power globally.
Benchmarking industry giants: Musk said plainly that SpaceXAI has been around for just 3 years, while Anthropic and OpenAI have 6 years and 10 years of history, respectively. He predicts that top-tier models (such as GPT-6-level) will be released within 2–3 months. If the current acceleration momentum is maintained, SpaceXAI could overtake within about 6 months and take the lead position in the industry.
💡 Crypto/AI perspective: As the scarcest “hard currency” in today’s AI wave, compute power not only determines how quickly top large models iterate, but also serves as a core anchor driving the strong growth of decentralized compute (DePIN) and the AI + Web3 narrative. By pushing data center construction at an “extreme speed,” Musk is reshaping the competitive landscape across AI and computing infrastructure.
$SPCX.US
$NVDA.US
$META.US
Elon Musk has once again raised the ceiling of the AI compute arms race. SpaceXAI’s Colossus data center in Memphis is undergoing a massive expansion. Musk said that Phase 2, Colossus 2, will roll out 660,000 of NVIDIA’s latest Blackwell GB300 GPUs ahead of year-end.
📌 Key highlights at a glance
The empire of compute surges: Musk calls Colossus a compute “super factory.” Phase 1, Colossus 1, is currently running roughly 230,000 GPUs (mainly H100, with a small number of GB200). Phase 2, Colossus 2, will fully shift to the more powerful Blackwell GB300 (equipped with Blackwell Ultra GPU cores).
Phased deployment plan:
Current base: 110,000 GB200 and 440,000 GB300 already in operation.
Sprint milestone: This week, 220,000 GB300 will be launched; add another 220,000 by the end of November. It’s expected that the final batch of 220,000 will be fully deployed by the end of December.
By then, the total number of GB300 in Colossus 2 alone will reach 1.1 million—an unrivaled scale of compute power globally.
Benchmarking industry giants: Musk said plainly that SpaceXAI has been around for just 3 years, while Anthropic and OpenAI have 6 years and 10 years of history, respectively. He predicts that top-tier models (such as GPT-6-level) will be released within 2–3 months. If the current acceleration momentum is maintained, SpaceXAI could overtake within about 6 months and take the lead position in the industry.
💡 Crypto/AI perspective: As the scarcest “hard currency” in today’s AI wave, compute power not only determines how quickly top large models iterate, but also serves as a core anchor driving the strong growth of decentralized compute (DePIN) and the AI + Web3 narrative. By pushing data center construction at an “extreme speed,” Musk is reshaping the competitive landscape across AI and computing infrastructure.
$SPCX.US
$NVDA.US
$META.US
