SpaceX Aims for 10 Gigawatt AI Capacity with Microsoft Partnership on Horizon

SpaceX has rapidly become a major player in the artificial intelligence (AI) infrastructure landscape. The company spent $15.8 billion on AI-related capital expenditures during the second quarter, bringing its total online capacity to 1.4 gigawatts of nameplate power. This is just the beginning: SpaceX aims to reach 10 gigawatts by the end of next year.

To achieve this ambitious goal, SpaceX will focus exclusively on Nvidia's Vera Rubin architecture. The company intends to shift from internal training for its AI model, Grok, to renting additional capacity for external training and inference applications.

Much of this expansion is being driven by demand from large cloud hyperscalers such as Microsoft. According to a report from SemiAnalysis, Microsoft may be looking to partner with SpaceX due to its own growing needs. With existing contracts valued at over $300 billion, Microsoft is likely to require significant additional capacity in the near future.

SpaceX's data center expansion plans have already attracted major players such as Anthropic and Google Cloud. The company has secured agreements worth more than $70 billion with these partners, validating SpaceX's ability to convert its underutilized data center capacity into high-margin recurring revenue.