SoftBank’s Billion-Dollar Bet , Why Masayoshi Son is Going All-In on an AI-Powered Future, Despite the Hype

Masayoshi Son’s public appearances have a certain aspect that distinguishes him from the cautious, circumspect communication style of other leaders running businesses the size of SoftBank. He doesn’t use qualified projections or scenarios when speaking. He uses civilizational certainties in his speech. Son discusses artificial superintelligence—machines that will be ten thousand times smarter than a human brain—at investor presentations in Tokyo, in interviews with Bloomberg and the Financial Times, and in talks with heads of state and tech entrepreneurs. He does this with the cool specificity of someone who has already determined that this is inevitable and has focused his entire professional life on making it happen. For those viewing all of this, the question is whether Masayoshi Son is, once again, the most confident person in the room on a wager that hasn’t yet been put to the test by reality or if that certainty represents true prescience.

In this case, context is crucial. Launched with great fanfare in 2017 and supported by Saudi sovereign cash and his own convictions, Son’s Vision Fund provided some of the most costly investing lessons in venture capital history. Son’s $47 billion firm, WeWork, fell into bankruptcy in a matter of months, making it the headline disaster, but it wasn’t the only one. Numerous portfolio companies that received cash from the Vision Fund at high valuations went on to fail, restructure, or refund a portion of their investment. Tens of billions of dollars were written off in total. At one shareholder meeting, Son displayed a PowerPoint of himself with the Japanese words “I was an idiot” on it. It was noteworthy for its candor and implied that the author had some understanding of the discrepancy between vision and result.

Key Biographical & Company Information

It is worthwhile to consider that background while assessing Son’s current actions, not because past errors foretell future ones but rather because the structural patterns are identifiable. He has completed a $41 billion investment in OpenAI, which is the largest single investment SoftBank has made in any company. This gives the group an estimated 11 to 13 percent stake in a company whose valuation has been rising due to enthusiasm for AI, but whose path to the kind of profitability that would justify that valuation at scale remains genuinely uncertain.

SoftBank took on a roughly $40 billion bridge loan to fund this wager, a move so big that S&P Global changed its credit outlook for the business to negative. In order to free up funds for AI initiatives that might or might not yield equivalent profits, Son separately sold SoftBank’s Nvidia stock, which would have been worth an incredible multiple of its acquisition price if held. Given how the company’s valuation has changed since, the Nvidia decision in particular has taken on a particularly bitter resonance.

Together with OpenAI, Oracle, and other partners, Son is spearheading the Stargate effort, which calls for a $500 billion commitment to construct AI data center infrastructure across the US. It is an aspiration at a size that is actually hard to imagine—a figure that needs to be considered for a while before it becomes clear. If carried out, it would be among the biggest infrastructure investments in the history of American technology.

The parties involved will suffer significant financial repercussions if it freezes or creates stranded assets, such as data centers designed for an AI demand curve that develops more slowly than expected. Son has stated in public that he has “no financial constraints” on his aspirations. His credit rating companies and lenders seem to have a slightly different opinion.

The component of Son’s AI strategy with the most structural rationale is Arm Holdings, the chip manufacturer that SoftBank went public and kept a controlling share in. Hardware truly limits AI computing, and regardless of which AI software business ends up winning, the firm that creates the architecture that powers energy-efficient AI inference chips holds a position with genuine, long-lasting value.

Son is pushing Arm farther into the creation of AI accelerators, expanding the company’s relevance into the particular workloads that are now using the most capital in the industry: model training and inference at scale. Compared to some of the others, this wager is more realistic since it is based on tangible goods and well-established intellectual property rather than predictions about which AI model will work best.

The risk of competition is genuine and deserving of frank mention. The current wave of AI funding and deployment was sparked by OpenAI’s ChatGPT, but during the past year, Google’s Gemini and Anthropic’s Claude have significantly reduced the capability gap. Son is placing a wager that OpenAI will continue to be the industry leader in AGI development. To make this wager, OpenAI must continue to have a technical and financial advantage over rivals with far bigger balance sheets. That’s not impossible, but it’s hardly the kind of moat that usually costs $41 billion.

There’s a sense that what’s happening is either the beginning of something historically significant or an extraordinarily costly repetition of a pattern that Masayoshi Son has already experienced once, given the debt accumulation, the bold infrastructure commitments, and the talk of one billion AI agents replacing human labor inside his own company. He thinks it’s the former. The market is still unsure and is becoming a little more wary about AI values.