Elon Musk estimated at a SpaceX employee meeting that the power capacity of SpaceX AI (formerly xAI) data centers will increase by about seven times before the end of 2027, reaching 10GW. This means that this company, which is famous for rockets, is rapidly advancing its AI computing infrastructure at an almost exponential pace. Musk also claimed that SpaceX AI has already built the "strongest AI training cluster in the world" and plans to further expand the cluster size to about 10 times its current scale by the end of next year.

It should be clarified that GW (gigawatts) is a unit of power, with 1GW equaling 1 billion watts. It is used to describe the power supply capacity or rated power consumption of data centers, but power supply capacity does not equate to actual computing power. Systems such as cooling, lighting, power distribution, networking, storage, and servers all consume electricity, so 10GW more often represents the energy consumption ceiling of the infrastructure rather than pure computing output.

30-50 Dollars per Watt, Annual Revenue Targeting 500 Billion Dollars

Musk calculated the stakes of this computing power gamble: if each watt is worth about 30 to 50 dollars, and the 10GW AI computing power can be online by the end of next year, xAI could generate annual revenue of 300 billion to 500 billion dollars. At current exchange rates, this range is approximately 2.03 trillion to 3.38 trillion Chinese yuan, almost equivalent to the GDP of a medium-sized country.

The basis for achieving this goal is the existing layout: the two data centers of SpaceX AI located in Memphis and Southaven have a combined rated power consumption of 1.4GW, and are planned to increase to 10GW by the end of 2027, meaning a growth of about seven times within about one and a half years. This aligns with Musk's previous prediction that "about 99% of SpaceX's value will come from AI five years later." When the rocket business is overtaken by the wave of computing power, SpaceX has essentially transformed into an AI infrastructure company. However, the power consumption scale of 10GW will also bring enormous pressure on the power grid and carbon footprint, and the competition between computing expansion and energy constraints may be more decisive for the success or failure of this gamble than the model itself.