SoftBank Group chief executive Masayoshi Son told executives at the company's annual event in Tokyo on Tuesday that expanding artificial intelligence will require almost $5 trillion in investment annually and globally, and he waved away fears of a bubble as absurd [1]. "To ask whether AI is a bubble is a foolish question," Son said. "AI will transform our lives completely, and do so in a way that generates profits" [1].

He put a date and a share of the world economy on the claim. "In 2040, approximately 20% of the world's GDP will be replaced by AI-related industries, the world of superintelligence," Son said [1]. The annual $5 trillion, he added, would go to expanding data centers, increasing production of computer chips, and providing the energy systems and other infrastructure AI depends on [1].

What Son did not supply is the arithmetic. The $5 trillion and 20% figures arrived without a disclosed model, financing plan, return threshold, or allocation across chips, power, and buildings — the operative questions for anyone deciding whether the buildout pays. That gap matters because Son is not a neutral forecaster: he founded SoftBank more than four decades ago, was an early backer of AI, and has poured tens of billions of dollars into related companies through the Vision Funds he oversees [1]. AP made the interested-party point plainly, reporting the numbers as Son's estimates and cataloguing his stake in the outcome [1].

That framing is where the story splits. On X, AI bulls lifted the $5 trillion out of Tokyo and recirculated it as an inevitability — the scale itself offered as proof the spending is destined and safe. Skeptics ran the other way, treating the same figure as a valuation-and-return warning from a man whose funds profit if the market believes him. The mainstream account keeps the caveat attached; the social version strips it, and a promotional line at a shareholder-facing event becomes a neutral map of the future.

Son's timing sharpened the contrast. His remarks landed as markets have been swept by concern that the surge in shares of companies like Nvidia, and the flood of money into data centers, may not return the profits investors are counting on [1]. SoftBank itself has ridden the wave up: the group earlier reported profit for the fiscal year through March soaring nearly five-fold to 5 trillion yen, about $32 billion, on the back of its AI bets [1]. A single strong year is evidence the trade has worked so far. It is not the disclosed model, financing plan, or return threshold that would show a $5 trillion annual bill can be paid every year to 2040.

-- Theo Kaplan, Tokyo