In a recent public speech, Elon Musk, CEO of Tesla, predicted that AI computing power in space will become the most cost-effective option within the next 4 to 5 years. He believes this is due to the "free" solar energy in space and convenient radiative cooling. Musk mentioned that as AI computing clusters continue to grow in scale, the power and cooling demands on Earth will reach levels that current infrastructure cannot handle. He estimates that future AI computing power may require 200 to 300 gigawatts, or even close to 1 terawatt of continuous power output.

To better understand this number, it can be compared with existing power supplies. A typical nuclear power plant can provide about 1 gigawatt of power, while the total continuous electricity generation in the United States is only 490 gigawatts. Since building power plants of this scale on Earth is almost impossible, Musk believes the only solution is to develop into space.

However, Huang Renxun, CEO of NVIDIA, opposes this idea, stating that Musk's plan for space-based AI data centers is currently just a "dream." Huang pointed out that even in the seemingly ideal geostationary orbit, building large AI data centers faces many technical challenges. For example, a GPU cluster at the megawatt level requires tens of thousands of square meters of cooling equipment, far exceeding the deployment capacity of any current spacecraft. In addition, launching such massive equipment into orbit would require thousands of high-cost launches, which is almost impossible to achieve within Musk's timeframe.

Furthermore, the harsh environment of space presents challenges for technology. Current high-performance AI chips cannot withstand the strong radiation of geostationary orbit without modification. To make these chips work properly, they must be heavily shielded or designed with new radiation-resistant solutions, which usually affect the performance of the chips. At the same time, technologies such as high-bandwidth connections to Earth, autonomous services, debris avoidance, and robotic maintenance are still in early development and not yet mature.

Key Points:

🌌 Musk predicts that AI computing power in space will be more cost-effective than on Earth in the next five years.

🚀 Huang Renxun expresses doubt about Musk's space plan, stating that there are still many technical challenges.

💡 The construction of large-scale AI data centers in space involves complex technical issues that need further solutions.