SoftBank Issues $11 Billion Bond to Invest in OpenAI, Investor Roadshow in New York from September 14 to 17
SoftBank Group has completed a bond issuance totaling $11 billion, which includes $10 billion in U.S. dollar bonds and €1 billion in euro-denominated bonds. The funds will be used for ongoing investments in OpenAI.
This bond issuance is led by SoftBank's CFO Yoshimitsu Goto, with an investor roadshow taking place in New York from September 14 to 17. Citigroup, Goldman Sachs, JPMorgan, and Morgan Stanley are the lead underwriters, and the bonds are privately issued to institutional investors under Rule 144A. The previously reported issuance range was between $10 billion and $20 billion, with the final pricing at the lower end of the range.
The raised funds will mainly be used to refinance a $40 billion bridge loan, which was jointly underwritten by JPMorgan, Goldman Sachs, Mizuho Bank, Sumitomo Mitsui Banking Corporation, and Mitsubishi UFJ Financial Group, specifically to finance SoftBank's investment in OpenAI. So far, $30 billion of this bridge loan has been drawn, with the third tranche of $10 billion set to mature on October 1, while the original loan maturity date is March 2027. SoftBank's total investment commitment to OpenAI is approximately $64.6 billion, corresponding to about 13% equity stake.
In terms of credit ratings, S&P has assigned SoftBank a BB+ rating with a negative outlook (this rating has been maintained since March this year); Moody's has given a Ba2 non-requested rating (SoftBank has not actively sought ratings since 2020). Both rating agencies classify it as speculative grade, reflecting market concerns over its aggressive capital expenditure strategy.
Just before this dollar/euro bond issuance, SoftBank completed a ¥1 trillion (approximately $6.3 billion) seven-year retail bond issuance in the Japanese domestic market at the beginning of September, setting a record for the largest retail bond issuance in Japanese corporate history, with a coupon rate set at 4.75%, nearly double the average coupon rate of 2.3% for yen-denominated retail corporate bonds this year. The funds raised will mainly be used to refinance maturing bonds and to pay for the $5.375 billion acquisition of ABB's robotics business announced last October.
The core reason investors are willing to accept SoftBank's series of high-yield bond issuances is that the current interest spread and coupon are viewed as the risk premium necessary for SoftBank's simultaneous bets on OpenAI equity, the ABB robotics acquisition, Arm's control, and data center expansion, which is a "multi-line betting" strategy. So far this year, large tech companies have cumulatively issued about $220 billion in bonds in the bond market, making the overall high-yield bond market relatively crowded, with credit spreads widening. This is also the direct reason why SoftBank's current dollar/euro bonds are priced at the lower end of the range rather than the maximum of $20 billion.
Market analysis compares SoftBank's current high-leverage multi-line betting strategy to its Vision Fund's earlier investments of "tens of billions" in companies like WeWork and Wag, which ultimately performed poorly, causing significant impairments. This history remains an important reference for the market when assessing SoftBank's risk appetite. The current financing arrangement for OpenAI continues a similar approach—first quickly fulfilling investment commitments through a $40 billion bridge loan (underwritten by JPMorgan, Goldman Sachs, Mizuho, Sumitomo Mitsui, and Mitsubishi UFJ), and then gradually replacing the bridge loan with retail and institutional bonds, essentially following the consistent path of "using short-term funds to seize market share, then rolling over with long-term capital markets financing."
The funding path is clear: $30 billion has been drawn from the bridge loan, and the third tranche of $10 billion will mature on October 1. SoftBank will use the ¥1 trillion (approximately $6.3 billion) retail bond issued in early September and the current $11 billion institutional bond to fill the gap, bringing its total investment in OpenAI to approximately $64.6 billion for about a 13% equity stake. Meanwhile, at the same time, SoftBank is also advancing the $5.375 billion acquisition of ABB's robotics business, holding control of Arm, and expanding its data center business—leveraging multiple fronts simultaneously, motivated by not wanting to miss any key entry points in the AI cycle (model layer of OpenAI, hardware layer of Arm and robotics, infrastructure layer of data centers), using a "full track betting" strategy to secure multiple positions in the AI industry chain.
This "debt-financed bet on a single super platform" model is most closely analogous to SoftBank's own $32 billion acquisition of Arm in 2016—at that time also seen as a high-stakes gamble, and it took the market nearly a decade to gradually recognize its strategic value. The current industry phase SoftBank is in is the "capital concentration betting period" of the AI infrastructure investment cycle: unlike companies like Nvidia and Microsoft that rely on their own cash flow to invest in AI, SoftBank chooses to leverage through debt with a junk credit rating, which significantly increases its risk exposure compared to its peers in the AI sector.
Essentially, this is a form of capital concentration, but the method of concentration is worth examining—SoftBank transfers the group's credit risk through the bond issuance tool to global fixed-income investors willing to pay a premium for BB+ rated junk bonds, in exchange for concentrated holdings in assets like OpenAI, Arm, and robotics. Mechanically, this structure can exist because the current market still believes that "the track SoftBank is betting on is large enough and the returns high enough to cover the cost of debt"—once this belief wavers (for example, if OpenAI's valuation growth slows or interest rates rise further), SoftBank's financing costs and refinancing pressures will rise simultaneously, forming a negative feedback loop of "weakening story → rating pressure → rising financing costs → increased balance sheet pressure," which is the greatest structural risk of the high-leverage concentrated betting strategy.
ABAB News · Cognitive Law
To capture all tracks, one must pay for all risks.
Betting on debt is not about probability, but about how long the belief can hold.
The lower the rating, the bigger the story must be told.
ABAB AI Insight
SoftBank's practice of financing aggressive bets through high-yield bond issuances has a historical precedent. Under Masayoshi Son, the Vision Fund's first phase invested tens of billions in companies like WeWork and Wag, most of which performed poorly, leading to significant impairments. This history remains an important reference for the market when assessing SoftBank's risk appetite. The current financing arrangement for OpenAI continues a similar approach—first quickly fulfilling investment commitments through a $40 billion bridge loan (underwritten by JPMorgan, Goldman Sachs, Mizuho, Sumitomo Mitsui, and Mitsubishi UFJ), and then gradually replacing the bridge loan with retail and institutional bonds, essentially following the consistent path of "using short-term funds to seize market share, then rolling over with long-term capital markets financing."
The funding path is clear: $30 billion has been drawn from the bridge loan, and the third tranche of $10 billion will mature on October 1. SoftBank will use the ¥1 trillion (approximately $6.3 billion) retail bond issued in early September and the current $11 billion institutional bond to fill the gap, bringing its total investment in OpenAI to approximately $64.6 billion for about a 13% equity stake. Meanwhile, at the same time, SoftBank is also advancing the $5.375 billion acquisition of ABB's robotics business, holding control of Arm, and expanding its data center business—leveraging multiple fronts simultaneously, motivated by not wanting to miss any key entry points in the AI cycle (model layer of OpenAI, hardware layer of Arm and robotics, infrastructure layer of data centers), using a "full track betting" strategy to secure multiple positions in the AI industry chain.
This "debt-financed bet on a single super platform" model is most closely analogous to SoftBank's own $32 billion acquisition of Arm in 2016—at that time also seen as a high-stakes gamble, and it took the market nearly a decade to gradually recognize its strategic value. The current industry phase SoftBank is in is the "capital concentration betting period" of the AI infrastructure investment cycle: unlike companies like Nvidia and Microsoft that rely on their own cash flow to invest in AI, SoftBank chooses to leverage through debt with a junk credit rating, which significantly increases its risk exposure compared to its peers in the AI sector.
Essentially, this is a form of capital concentration, but the method of concentration is worth examining—SoftBank transfers the group's credit risk through the bond issuance tool to global fixed-income investors willing to pay a premium for BB+ rated junk bonds, in exchange for concentrated holdings in assets like OpenAI, Arm, and robotics. Mechanically, this structure can exist because the current market still believes that "the track SoftBank is betting on is large enough and the returns high enough to cover the cost of debt"—once this belief wavers (for example, if OpenAI's valuation growth slows or interest rates rise further), SoftBank's financing costs and refinancing pressures will rise simultaneously, forming a negative feedback loop of "weakening story → rating pressure → rising financing costs → increased balance sheet pressure," which is the greatest structural risk of the high-leverage concentrated betting strategy.
ABAB News · Cognitive Law
To capture all tracks, one must pay for all risks.
Betting on debt is not about probability, but about how long the belief can hold.
The lower the rating, the bigger the story must be told.