Elon Musk announced that SpaceX aims to reach at least 13 gigawatts of AI compute capacity by the end of 2027. This target represents a significant acceleration from the current capacity of 1.7 gigawatts recorded as of September 30, 2026. To achieve this, the company plans to add approximately 10 gigawatts of capacity throughout 2027.
The company is currently installing about 200,000 Nvidia GB300 GPUs every month through the end of 2026. Future deployments are expected to transition to Nvidia’s Vera Rubin architecture. A key milestone is the launch of a 440,000-GPU training cluster, which is scheduled to become operational by mid-2027.
Beyond compute capacity, SpaceX is developing power and cooling infrastructure designed to reach 15 to 20 gigawatts by the end of 2027. These facilities are intended to support the broader AI buildout. The company’s Super Intelligence segment, covering xAI cloud services, reported $7.56 billion in revenue for Q3 2026, marking an 830% year-over-year increase.
