This is not merely an investment; it is an all-in gamble, an act of faith. In Masayoshi Son's dictionary, fear of risk does not exist. He is perpetually driven, ready to deploy capital at a moment's notice. Having missed out on the premier AI stock, Nvidia, Son is now preparing to expend all of Softbank's ammunition, liquidating assets and securing colossal loans to place the entirety of his funds on OpenAI. The outcome of this gamble is binary: either legendary success or catastrophic failure. If OpenAI achieves its ambition of becoming a multi-trillion dollar enterprise, Son would secure an unprecedented investment return, potentially multiples of his legendary gain from Alibaba, thereby erasing the regret of missing Nvidia. However, should the AI bubble burst, Softbank could face severe damage, with losses potentially exceeding the massive writedown from its investment in WeWork.
The plan involves a $40 billion loan specifically to fund further investment in OpenAI. This dollar-denominated loan would be the largest in Softbank's history, with an approximate 12-month term, jointly underwritten by four top-tier financial institutions including JPMorgan. With the Federal Reserve's benchmark rate between 4.25%-4.5%, large corporate syndicated loans are typically priced at a spread over SOFR (Secured Overnight Financing Rate), which is currently around 4.3%. Softbank's credit rating is speculative grade (junk status: S&P BB+, Moody's Ba3). This rating implies a significantly higher loan spread compared to investment-grade companies, typically adding 150-300 basis points (1.5%-3%) or more. While the involvement of major banks like JPMorgan might offer some pricing benefits, Softbank's rating and the loan's purpose—betting on a single, unlisted company—will inflate the risk premium. Interest payments alone could amount to billions of dollars.
The scale of a $40 billion short-term loan is historic. Last year's total U.S. leveraged loan volume was approximately $600 billion. Typically, only massive leveraged buyouts (LBOs) involve loans of this magnitude, such as Elon Musk's $25.5 billion loan for Twitter or the recent Ellison family's $54 billion loan for Warner Bros. The critical difference is that such colossal bank loans usually finance the acquisition of cash-flowing companies with tangible assets for collateral. In contrast, Softbank intends to use this funds as a short-term bridge loan, placing a massive bet on OpenAI—a young, unprofitable, private AI company lacking hard assets.
So, what collateral does Softbank offer? Its stake in Arm is the primary security. Softbank holds roughly 90% of Arm, which has a Nasdaq market cap of about $1600 billion, valuing Softbank's holding at approximately $1400 billion. This represents its most liquid hard asset. Additionally, the OpenAI equity acquired with the loan would be partially used as collateral. Essentially, banks are lending based on the value of Softbank's Arm shares, not on OpenAI's intrinsic worth. The loan fundamentally hinges on banks betting that Arm's stock price remains stable and the AI bubble does not burst. This is the loan's most striking aspect. Softbank's core risk is extreme asset concentration; Arm alone underpins most of the collateral value. If Arm's share price plummets, eroding collateral value, banks could demand additional security or early repayment—the core logic behind S&P's recent revision of Softbank's credit outlook.
In reality, Son is using Arm as a chip to bet on OpenAI's future. This wager is not impulsive but a years-long obsession. Since ChatGPT's launch in 2022, Son has relentlessly pursued the industry leader that ushered in the AI era. In 2023, Son revealed he uses ChatGPT heavily daily and maintains regular, almost daily, brief communication with OpenAI CEO Sam Altman. At the time, Microsoft was the strategic investor, and Son waited patiently for an opening. His patience paid off in September 2024, when Softbank made a tentative $500 million investment in OpenAI at a $1500 billion valuation. This was merely an entry point; Son awaited a larger opportunity.
In February 2025, Softbank led a $400 billion funding round for OpenAI at a $3000 billion valuation—nearly double the $1570 billion valuation from October 2024. Softbank contributed $300 billion, with co-investors like Microsoft, Coatue, Altimeter, and Thrive adding $100 billion. Notably, the agreement included a protective clause: Softbank's commitment could be reduced to $200 billion if OpenAI failed to transition to a for-profit entity by end-2025. OpenAI later successfully completed this transition despite pressure from Musk, securing the full investment. By end-2025, cumulative investment exceeded $346 billion for an approximate 11% stake, making OpenAI Softbank's second-largest holding after Arm. To raise these funds, Softbank sold $5.8 billion of its Nvidia stake, divested $4.8 billion in T-Mobile shares, and borrowed $11.5 billion using Arm stock as collateral.
Recently, OpenAI completed a $1100 billion funding round at an $8400 billion valuation. Softbank and Nvidia each committed $300 billion, with Amazon investing $500 billion. Softbank plans to invest its share in three $100 billion tranches in April, July, and October. This will bring Son's total investment in OpenAI to $646 billion, with Softbank's stake expected to rise from 11% to approximately 13%. The revelation of the $40 billion bridge loan plan mere days after this announcement indicates Softbank is not only liquidating assets but also taking on massive debt to fuel this gamble. The only explanation: Son believes OpenAI's future value far exceeds current imagination.
Softbank is now OpenAI's third-largest shareholder and the second-largest external investor after Microsoft. The OpenAI non-profit foundation and Microsoft hold comparable stakes, but the foundation controls the board. Microsoft and Softbank collectively own over 40% of OpenAI yet hold no board seats or governance rights, unable to influence strategy or product direction. Essentially, Son has purchased a pure financial investment ticket, betting on post-IPO share price appreciation. He has no say in the AI leader's operations, path to AGI, or business model decisions. OpenAI's unique structure is designed to keep investor control at bay. Furthermore, Softbank's investment differs from those of Microsoft, Nvidia, and Amazon. These tech giants have strong strategic partnerships and supply-chain ties with OpenAI; their investments enhance their own commercial interests and future industry positioning. Softbank is almost purely a financial investor.
Understanding Son's current obsession requires revisiting a painful chapter: his regret over Nvidia. In 2017, Softbank's Vision Fund invested $700 million for a nearly 5% stake in Nvidia. According to CEO Jensen Huang, Softbank was Nvidia's largest institutional shareholder at the time, as ownership was highly fragmented among index funds. The AI wave had not yet crested; Nvidia's primary business was gaming GPUs, a stable but limited-growth market. Its potential was underestimated, with a market cap around $80-90 billion and a declining stock price. Perhaps perceiving limited upside, and needing to bolster Vision Fund's books after massive losses from a $10+ billion investment in WeWork, Son reluctantly sold the Nvidia stake in early 2019, netting $3.6 billion—a $2+ billion profit over two years that initially seemed satisfactory.
The subsequent story is well-known. Post-ChatGPT, Nvidia's GPUs became strategic necessities, fiercely demanded by tech giants. Nvidia's stock soared, eventually surpassing Apple to become the world's most valuable company with a current market cap exceeding $4.3 trillion. The sold stake would theoretically be worth over $210 billion today. Son later publicly admitted selling Nvidia was a decision he "cried over," one of his most painful business regrets. At a Nvidia event in Tokyo in 2024, Huang told the audience, "Many may not know, Softbank was once Nvidia's largest shareholder." Son, speechless, jokingly pretended to cry on Huang's shoulder, with Huang consoling him, "It's okay, we can cry together." This experience shapes Son's近乎偏执的逻辑 regarding OpenAI: having missed the first AI champion, he cannot afford to miss the winner in the large model arena.
Ironically, to fund the initial $300 billion investment in OpenAI, Softbank again sold approximately 32 million Nvidia shares last year, raising about $5.8 billion, and immediately redirected the proceeds into OpenAI. History's cycle is dramatic: the previous Nvidia sale was forced by financial pressure; this recent divestment was a deliberate choice to bet on what he perceives as greater than Nvidia.
Son's ambitions extend beyond being OpenAI's largest financial investor. He aims to position Softbank as an indispensable infrastructure provider for the entire AI ecosystem. In January 2025, Son collaborated with OpenAI, Oracle, and MGX to launch "Stargate," a project planned to invest $5000 billion in building AI data centers and配套 power infrastructure across the U.S. Son served as project chairman, appearing alongside former President Trump at a White House announcement. Musk, an OpenAI competitor, publicly mocked the plan as an empty promise, claiming the consortium lacked the funds. Altman retorted, and Arm's CEO stated funding was "quite sufficient."
A year later, the三方似乎还没有搞定结构框架. OpenAI desires to own the data centers itself, a point of contention with Oracle and Softbank. The Stargate joint entity has not hired staff or developed data centers uniformly; the parties have signed bilateral agreements, bypassing the Stargate framework itself. Furthermore, significant funding gaps have emerged, partially validating Musk's skepticism. Bloomberg reported that OpenAI and Oracle abandoned plans to expand the Abilene, Texas campus due to financing issues and OpenAI's inaccurate demand forecasts. Nvidia subsequently mediated, with Meta potentially taking up the unused capacity.
How much has Softbank actually invested? At the project's launch, Softbank and OpenAI each committed $19 billion as primary contributors to the initial $100 billion, with Softbank responsible for financing and OpenAI for operations. Softbank's committed capital to Stargate was $19 billion for a 40% stake; Oracle and MGX committed $7 billion each for 10% stakes; remaining funds were to come from limited partners and debt financing. However, the $19 billion is a "commitment," not an immediate cash transfer. Moreover, Softbank's actual equity contribution is only 10%, with the rest heavily reliant on debt and external borrowing—meaning most of the $19 billion commitment is leveraged. This is another leveraged bet: minimal own capital amplifying a massive wager on the future.
In essence, before even raising substantial funds for the contentious Stargate project, Softbank is seeking a $40 billion loan primarily for direct equity investment in OpenAI. Compared to the三方扯皮 of Stargate, leading OpenAI's funding round is the tangible priority. Son's leverage has reached unprecedented levels. Combining the $64.6 billion equity investment and the $19 billion compute project commitment, Softbank plans to invest a total of $83.6 billion in OpenAI over two years. This is not Son's sole AI bet; he is also investing elsewhere.
Softbank acquired chip designer Ampere Computing for $6.5 billion, made a $5.4 billion offer for ABB's robotics business, invested $2 billion in Intel, maintains control over Arm (holding ~90% of global chip design IP), and led a $1 billion funding round for autonomous driving company Wayve. From chip design to models, robotics, and autonomous vehicles, from compute power to applications, Son's full-stack AI investment portfolio is taking shape.
Not everyone applauds Son's冒险精神. Concurrent with the $40 billion loan announcement, S&P revised Softbank's credit outlook to negative, citing potential damage to liquidity and asset credit quality from the massive OpenAI investment. OpenAI's commonly cited valuation ranges between $7300-$8400 billion, with some analysts predicting an IPO valuation reaching $1 trillion. Critics point out that OpenAI continues to burn cash heavily, with an unclear path to profitability; its high valuation relies largely on faith in AGI's potential rather than current cash flows.
Softbank's fate is now deeply intertwined with OpenAI's commercial success or failure—prosperity and ruin shared. WeWork was Son's most famous Waterloo, a $10+ billion debacle that severely wounded Softbank. The current OpenAI gamble is orders of magnitude larger. Fairly, Son's current hand is strong. In Q4 2025, Softbank reported a net profit of approximately $1.6 billion, marking four consecutive quarters of profitability, a stark contrast to substantial losses a year prior. The core driver of this reversal was the significant appreciation of its OpenAI stake—by end-December 2025, cumulative unrealized gains from the investment reached about $19.8 billion.
But investors know well: paper profits are not real profits. OpenAI remains private; should the AI sector cool, Softbank's equity could become illiquid. Trillion-dollar valuation gains are built on future expectations, not realized cash. Son, however, appears unconcerned. He is not making an ordinary venture capital investment; he is using commercial action to interpret his judgment about humanity's future.
In his judgment, AGI will be realized within the next decade, with AI intelligence surpassing humans by thousands or tens of thousands of times,彻底重塑 all industries in human history. He has repeatedly stated publicly: "AI can create incremental value equivalent to 10% of global GDP. Investing trillions of dollars is worthwhile." Son compares himself to when he invested in Alibaba—$20 million yielding thousand-fold returns, creating one of business history's most successful VC bets. He believes OpenAI is his second Alibaba, albeit with stakes thousands of times larger.
At the peak of the dot-com bubble, Son briefly became the world's richest person, only to see Softbank's stock plunge 98% after the bust, wiping $60 billion from his personal wealth and hitting rock bottom. Yet, the impulsive $20 million investment in Jack Ma's Alibaba, made before the crash, later yielded an unbelievable return exceeding $72 billion for Softbank. This time, however, Softbank's risk is unprecedented, with nearly all eggs placed in one basket: $64.6 billion in direct investment, a $40 billion bridge loan, plus his push for the $5000 billion Stargate project. Son's冒险 is on a scale never seen before. He refuses to label the current AI fervor a "bubble." In his view, those sounding warnings are simply "not smart enough" to see the direction of history.
Comments