Son slams Musk's orbital AI bet: $19B Stargate trumps space-based dreams

In a rare direct confrontation between two of the world's most ambitious technology investors, SoftBank Group CEO Masayoshi Son delivered a stinging critique of Elon Musk's plan to build orbital AI data centers, calling it a 'losing bet' at SoftBank's wireless division shareholder meeting in late June 2026. The rebuke, reported by Bloomberg and Fortune, pits two fundamentally different infrastructural strategies against one another: Musk's vision of harnessing cheap solar energy in space versus Son's relentlessly terrestrial, immediate-scale approach. Son, whose company recently surpassed Toyota as Japan's most valuable corporation, argued that the cost equation for AI computing is dominated by semiconductors and hardware, not electricity. He noted that while SpaceX pitches orbital data centers as a way to slash electricity bills, power accounts for only a small share of total operating expenses. The far greater expense lies in the chips themselves, making the solar-energy argument a red herring. Beyond the cost fallacy, Son cited three operational barriers: prohibitive launch costs for moving heavy computing equipment into orbit, the near-impossible task of maintaining hardware in a space environment, and unacceptable communication latency between ground and orbit that would hamstring AI training and inference workloads. His dismissal carries particular weight given that SoftBank has committed $19 billion to the Stargate project alongside OpenAI, an enormous terrestrial AI infrastructure buildout announced in January 2025. The contrast in timelines is central: Son emphasized that the next few years will decide AI supremacy, making decade-long orbital projects irrelevant to the current race. Yet he acknowledged Musk as an 'extraordinary agent of change,' signaling that the critique is strictly about timing and economics, not engineering capability. For the space industry, this public disagreement illuminates the gap between visionary space-based infrastructure and the near-term capital allocation realities that dominate boardrooms. Son's argument may cool enthusiasm for orbital computing ventures, potentially diverting investment back toward ground-based data center expansion, which benefits chipmakers like Nvidia, real estate trusts, and terrestrial energy providers. Conversely, it underscores the immense technical and financial hurdles SpaceX must overcome to make orbital AI commercially viable. Looking ahead, the debate highlights how the AI boom is forcing a reckoning on infrastructure deployment speed; as Son bluntly put it, the next few years are far more important than what might happen a decade from now. That urgency may accelerate capital flows into immediate, scalable ground solutions, while leaving orbital computing as a long-shot concept awaiting not just technological breakthroughs but a fundamental shift in cost structures.