TOKYO—The loan has no collateral in the usual sense; its security is a stake in a company that did not exist five years ago. SoftBank Group has borrowed $10 billion in a margin loan secured by its shares of OpenAI, according to people familiar with the matter, a transaction that lets the Japanese conglomerate convert paper gains in its AI portfolio into cash for new investments. The deal, first reported by The Wall Street Journal, is the latest chapter in Masayoshi Son’s campaign to place the largest possible bet on artificial intelligence without selling the assets he believes will pay off.
The mechanics of the loan are straightforward. SoftBank holds a large stake in OpenAI, acquired through a combination of investments that have made it one of the company’s biggest backers. Rather than sell those shares, SoftBank borrowed against them, using the market value of its OpenAI position as security. Margin loans are common in the financial world, but the size of this one—$10 billion—and the volatility of the underlying asset make it a bet on both directions: that OpenAI’s value will keep rising, and that SoftBank’s other investments will generate the returns to service the debt.
The purpose of the money is where the story gets interesting. SoftBank has been on a spending spree across the AI economy, from chipmakers to data centers to model developers, and the loan gives it fresh firepower for the next round. Son has said repeatedly that he intends to be the largest investor in the AI transition, and his strategy has been to use SoftBank’s balance sheet as a vehicle for bets that are too large for most funds. The OpenAI loan fits that pattern: borrow against the crown jewel, reinvest the proceeds in the wider AI complex.
The structure echoes moves Son made during the dot-com era, when SoftBank borrowed heavily against its technology holdings to fund further acquisitions. That strategy ended badly, and Son has acknowledged the scars, but he has also argued that this cycle is different, that AI is a genuine transformation rather than a mania. The difference is embodied in the collateral: OpenAI shares, which have appreciated sharply as the company’s valuation has climbed to hundreds of billions of dollars. The equity cushion is the argument for the loan’s safety.
The risks are real and measurable. Margin loans carry the danger of a downward spiral: if the value of the collateral falls, the lender can demand more cash or more shares, and if SoftBank cannot meet the call, the position is sold at the worst possible moment. OpenAI’s valuation is among the most volatile in the market, and a reset in AI stocks would hit the loan’s terms directly. SoftBank has other assets to draw on, including stakes in Arm and Alibaba, but those, too, are subject to market swings. The structure works as long as the AI boom continues; it becomes uncomfortable the moment it does not.
The transaction also says something about the state of the AI capital markets. The willingness of lenders to advance $10 billion against OpenAI shares is a measure of how far the company has come, and how much credibility it now has with the world’s biggest banks. Margin lending of this size requires the lender to believe the collateral is worth what the borrower says it is, and the fact that a consortium of banks agreed suggests the market has accepted OpenAI’s valuation as real. That acceptance is itself a form of endorsement, one that OpenAI can use in future fundraising.
For SoftBank, the loan is part of a broader pattern of financial engineering that has turned the conglomerate into one of the most aggressive players in the AI economy. The company has used debt, equity and special-purpose vehicles to fund its investments, and its balance sheet has become a machine for converting market confidence into capital. The OpenAI loan is the latest and largest example: a transaction that takes a paper stake and turns it into cash without diluting anyone. Son’s willingness to use such structures reflects his conviction that the AI opportunity will outrun the cost of the borrowing.
The deal also highlights the increasingly intertwined finances of the AI industry. SoftBank is simultaneously OpenAI’s largest backer, a major investor in AI infrastructure, and a borrower whose collateral is the value of a single startup. Those threads can pull in different directions in a downturn, and the industry’s exposure to a single point of failure has grown with the size of the bets. For now, the market is betting that the threads hold; the $10 billion loan is the latest confirmation that the capital is available for the AI buildout at almost any scale.
Son has described the AI transition as the defining opportunity of his lifetime, and the loan is consistent with that framing. The money will likely fund investments that reinforce SoftBank’s position across the AI stack, deepening the web of holdings that already spans the industry’s most important companies. Whether the strategy ends in triumph or in another painful lesson about borrowing, the $10 billion loan is a statement of intent: SoftBank intends to be the biggest player in the biggest technology story of the decade, and it is willing to borrow against its most valuable asset to stay in the game.
