SpaceX Aims for 10 GW of AI Capacity as Microsoft Eyes Partnership

SpaceX, founded by Elon Musk, has quietly become a major player in the artificial intelligence infrastructure market. In the second quarter, the company spent $15.8 billion on capital expenditures for AI compute, delivering 1.4 gigawatts of capacity online.

The goal is to reach 10 gigawatts by the end of next year, with plans to use Nvidia's Vera Rubin architecture in its data centers. This infrastructure expansion will not only support SpaceX's own AI model, Grok, but also provide external training and inference applications for other companies.

Microsoft could be a key partner in this endeavor, as it needs significant additional capacity by 2027. The company has already contracted for 10 gigawatts of capacity elsewhere, but that won't be fully online until late 2027 or early 2028. SpaceX's ability to rapidly deploy massive compute clusters and its existing agreements with AI developers Anthropic and Google Cloud make it an attractive option.