Masayoshi Son: Nuclear fusion energy will be the key to the future development of AI

Kuai Technology News on July 14, according to media reports, SoftBank Group founder Masayoshi Son said at the SoftBank World annual event on Tuesday that nuclear fusion technology will become the most realistic solution to meet the growing power needs of artificial intelligence data centers in the near future. Son also emphasized that natural gas will still take on most of the power supply tasks, but nuclear fusion - the process of generating energy from the sun and stars - is expected to gradually replace it in 15 years. Masayoshi Son predicts that by 2040, the power capacity demand for global data centers will reach as high as 3 terawatts. He pointed out that nuclear fusion will provide a cheaper, cleaner and safer new energy source, but admitted that the technology still faces many technical and economic challenges before it can be put into commercial use. "Is it really appropriate to use large amounts of natural gas to generate electricity now? I think that in 15 years, nuclear fusion will replace it." Son clearly refuted doubts about whether hundreds of billions of dollars in investment in AI infrastructure can produce returns. He said: "Asking whether artificial intelligence is a bubble is an extremely stupid question." He believes that only those who do not understand or fully utilize AI will raise such questions, and emphasizes that the high cost of building data center infrastructure will eventually be reasonably rewarded. It is worth noting that Masayoshi Son has been strongly opposed to traditional nuclear fission power generation since the Fukushima nuclear accident in 2011. But he sees nuclear fusion as a completely different technological path: In theory, nuclear fusion has extremely low carbon emissions, produces far less persistent radioactive waste than nuclear fission, and has no risk of runaway chain reactions, giving it significant safety advantages. In recent years, global investment in nuclear fusion startups has increased significantly, and many companies have moved from theoretical research to prototype manufacturing. The U.S. Department of Energy has set the mid-2030s as the target node for large-scale application of this technology.