SpaceXAI is preparing another major expansion of its Colossus AI infrastructure, with hundreds of thousands of Nvidia Blackwell GPUs scheduled to come online before the end of 2026.

Elon Musk said September 25 that Colossus 1 currently contains 150,000 H100, 50,000 H200 and 30,000 GB200 GPUs. Colossus 2 contains 110,000 GB200 and 440,000 GB300 GPUs, according to Musk. He said another 220,000 GB300s are expected to be fully operational next week and another 220,000 in November.

A third 220,000-GPU addition could arrive in late December if the schedule holds. That last batch is a target, not deployed hardware. If all three additions arrive, reporting from Bloomberg and Tom’s Hardware calculates that the combined Colossus installations would approach 1.44 million GPUs. The enormous buildout also illustrates why power delivery is becoming a defining engineering problem for high-density AI data centers.

Power remains a central constraint. SpaceXAI is transitioning its Memphis-area infrastructure toward permanent generation while expanding the compute campus. The GPU schedule also depends on hardware delivery, electrical capacity, networking and cooling arriving together, so future deployment numbers should be treated as plans until the systems are actually operational.

Sources: Elon Musk’s September 25 X disclosure, Bloomberg via Yahoo Finance, Tom’s Hardware, and SpaceXAI.

Follow BitcoinVersus.tech on X

Editorial note and disclaimer: BitcoinVersus.tech covers Bitcoin, technology, computing, energy and emerging infrastructure. Future deployment figures are plans until the hardware is confirmed operational. This article is for informational purposes and is not financial or investment advice.

Related BitcoinVersus.tech coverage: xAI 660,000 GB300 expansion · Giga Computing GB300 factory · Supermicro Vera Rubin racks · Armenia NVIDIA AI hub