SpaceX 计划到 2027 年实现 10 吉瓦算力目标

spacex plans to increase its compute capacity more than fivefold to reach approximately 10 gigawatts by 2027, a move that would require procuring potentially over two million nvidia vera rubin gpus. the company currently operates 540,000 gpus across h100, gb200, and gb300 architectures and is merging xai into spacex to consolidate these operations. this aggressive buildout suggests the merged entity anticipates massive demand for training compute, likely serving both internal grok models and external cloud clients like anthropic. the strategy positions spacex as a dominant infrastructure provider, challenging competitors who diversify across amd, google, and aws hardware.