Key Points
Elon Musk announced on September 25 that SpaceX’s xAI unit is set to significantly expand its AI training capabilities through its Colossus 2 supercomputer, which currently houses 110,000 Nvidia GB200 chips and 440,000 GB300 chips. This deployment could double Nvidia GPUs by year-end, as an additional 220,000 GB300 chips are expected to come online next week, with further orders planned for November and December.
SpaceX’s original target of scaling to 1 million GPUs by the end of 2026 now appears achievable, especially with an investment of approximately $18 billion for 550,000 GPUs. Furthermore, Musk stated that SpaceX anticipates ending the year with around 2 gigawatts of computing power, aiming for 10 gigawatts by 2027. Each gigawatt of compute power has an estimated cost of $50 to $60 billion, with Nvidia components accounting for about $35 billion.
Nvidia has also secured a partnership with SpaceX to develop GPUs designed for future orbital data centers, addressing challenges posed by cosmic radiation. This collaboration underscores Nvidia’s role as a critical AI infrastructure provider and solidifies its position in the burgeoning AI sector.
5 Stocks Our Experts Predict Could Double In the Next Year
By submitting your email, you'll also get a free pivot & flow membership. A free daily market overview. You can unsubscribe at any time.
