SoftBank's Son: Fusion power will fuel AI data centers by 2040

Son's remarks reflect a broader shift in how technology executives are grappling with AI's explosive energy demands. Natural gas currently supplies more than one-fifth of global power generation and accounts for the bulk of data center needs today, but its continued use contradicts the industry's climate commitments. Nuclear fusion—a process that has long been pursued as a theoretical solution to clean energy—is moving from pure research into prototype development, with several companies now advancing toward commercialization and government backing accelerating timelines.

Son's 15-year prediction aligns roughly with U.S. Energy Department goals to scale fusion in the mid-2030s, though BloombergNEF's caveat about remaining technological and financial hurdles suggests the timeline remains uncertain. Son's position is noteworthy given his historical opposition to conventional nuclear power following Japan's 2011 Fukushima disaster; fusion's theoretical absence of chain-reaction risk and its lower long-term waste burden appear to address his prior concerns. His confidence in fusion's viability stands in contrast to the complexity of the technology, which only a few countries have demonstrated even at an experimental level.