SoftBank Group (SFTBY) chief executive Masayoshi Son told shareholders this week that Elon Musk’s plan to build data centers in orbit is a losing bet, Fortune reported.Son, who has committed billions to on-the-ground AI infrastructure, framed the dispute as a question of timing and cost, not of engineering capability.The 68-year-old investor, whose company recently surpassed Toyota as Japan’s most valuable corporation, according to the Japan Times, offered a rare, direct feedback to Elon Musk.The remark carries weight because both men are deploying tens of billions toward AI infrastructure, yet their strategies could not look more different.Son argues chips cost far more than kilowatts in the data-center equationAt SoftBank’s wireless division shareholder meeting, Son responded to a question about orbital computing with a cost breakdown that challenges Musk’s space-based strategy, Bloomberg reported.Musk and SpaceX have pitched orbital data centers as a way to harness cheap solar energy in space, cutting the electricity bills that weigh on terrestrial operations.Son countered that electricity accounts for only a small share of total operating expenses, compared with the far higher costs of semiconductors and hardware.SoftBank CEO Masayoshi Son argued that the next few years of AI development carry decisive weight, Bloomberg noted.In the battle for AI, the next few years will be far more important than what might happen a decade or so from now.He also cited the steep cost of launching equipment into orbit, the difficulty of maintaining hardware in space, and communication latency between ground and orbit.Son acknowledged Musk as an “extraordinary agent of change,” but drew a firm line on timing and priorities in the current AI race, Bloomberg confirmed.SoftBank’s ground-first AI bets: from Stargate to artificial superintelligenceSoftBank committed $19 billion to the Stargate project alongside OpenAI in January 2025, according to Fortune, an initiative that plans to invest $500 billion over four years in U.S.-based AI infrastructure.Son has also invested heavily in chips through SoftBank’s dominant stake in Arm Holdings, the British designer whose blueprints power processors across the technology industry.At SoftBank Group’s parent company annual shareholder meeting the following day, Son set a target to grow the firm’s net asset value to 1 quadrillion yen over the next 16 years, Fortune reported.That figure, equivalent to roughly $6.4 trillion, would be driven by the company’s pursuit of what Son calls “artificial superintelligence.”
SoftBank doubles down on AI with massive infrastructure, chip, and superintelligence investments, aiming for a $6.4 trillion future vision.SOPA Images/Getty Images
The Son and Musk rivalry over AI runs deeper than orbital data centersThe orbital critique is only the latest flashpoint between two billionaires with a history of public disagreements stretching back nearly a decade.Elon Musk and Son met in 2017 to discuss a potential SoftBank investment in Tesla, but the talks collapsed over disagreements about ownership structure, Bloomberg reported.More AI:Goldman Sachs has blunt message for AI stock investorsMicrosoft CEO sends a blunt warning on AI and the tech ecosystemThe next AI infrastructure race has nothing to do with chipsAfter Stargate was announced in January 2025, Musk questioned its financial footing on X (the former Twitter), claiming SoftBank had secured well under $10 billion.SoftBank has since backed its commitment with acquisitions, including a $5.4 billion purchase of ABB Robotics and a $4 billion DigitalBridge deal.SpaceX’s record IPO puts serious money behind Musk’s orbital ambitionsSpaceX went public on June 12 in the largest initial public offering in financial history, raising approximately $75 billion at a valuation of about $1.77 trillion, according to CNBC.Morningstar analyst Nicolas Owens valued SpaceX at just $63 per share in a June report, well below the $135 offering price. Owens estimated that investors were effectively paying a $72-per-share premium for the chance to profit from speculative future projects such as orbital data centers.Morningstar assigned a 43% probability to a scenario in which orbital data centers never become commercially viable, and just 7% to its most optimistic outcome.Global energy demand gives both sides of the data center debate real urgencyGlobal data centers consumed roughly 415 terawatt-hours of electricity in 2024, about 1.5% of global electricity consumption, according to the International Energy Agency, and Gartner estimates the figure rose to about 448 terawatt-hours in 2025.That figure is projected to more than double by 2030, reaching roughly 945 terawatt-hours annually, driven by the rapid expansion of AI computing workloads.Son’s position is that the urgency of this energy crisis favors his approach, as terrestrial construction can deliver new capacity within months rather than years. “He who strikes first wins,” Son told shareholders, as Japan Times reported, condensing his strategic thesis into a single sentence that functions as both a commitment and a warning. Related: SoftBank CEO doubles down on AI, stock market message for 2026