Author: Dongcha Beating
Son Masayoshi is borrowing again, committing another $10 billion to bet on OpenAI
Son Masayoshi has gone to borrow money again.
This time it's $10 billion plus €1 billion, all used to buy OpenAI.
On September 21, Reuters uncovered SoftBank’s latest bond offering documents. Citibank and JPMorgan Chase are listed side by side as bookrunners, tasked with selling these high-risk notes on the open market. SoftBank’s corporate credit rating remains at BB+, a more polite term for high-yield bonds—commonly known as junk bonds.
The timeline for the bond issuance is extremely tight: the interest rate was finalized on September 24, funds were received on the 29th, and by October 1st, the full amount had to be transferred unchanged into OpenAI’s account. This is the third installment of the agreed-upon investment.
Billions of dollars pass through, lingering on SoftBank's books for only two days.
A tech visionary who claims to see thirty years into the future is calculating his cash flow day by day.
He never rushed like this when he was doing big deals before. In 1999, he met Jack Ma and invested $20 million. At the time, everyone thought he was crazy, but he used his own spare money—he could afford to wait fifteen full years, win or lose.
At that time, time was his ally.
Now it’s completely reversed. On the surface, it’s still the grand narrative of changing the world; beneath the surface, he’s racing against the calendar before every settlement deadline.
Chips locked in a drawer
On February 27, SoftBank signed the final agreement with OpenAI.
A total of $30 billion in additional investment has been split into three tranches of exactly $10 billion each, scheduled to be delivered on April 1, July 1, and October 1 respectively. Once the final payment is made, SoftBank’s total cash investment in the company will reach $64.6 billion, equivalent to approximately 13% equity.
$64.6 billion is twice the amount SoftBank spent to acquire Arm in 2016.
In this round, OpenAI’s pre-money valuation surged to $730 billion, far surpassing SoftBank Group’s own market capitalization. The installment payments were a compromise: OpenAI secured funding for the next several months, while SoftBank gained breathing room.
But the cost is that SoftBank must raise $10 billion among financial institutions every three months.
What convinced Son Masayoshi was a pitch deck prepared for top-tier investors.
Documents show that OpenAI projects $36 billion in revenue for 2026, swelling to $350 billion by 2030, with cumulative earnings of $840 billion over five years. The recent $122 billion funding round in March pushed its valuation to $852 billion. Its rival Anthropic is preparing for an IPO, and Masayoshi Son, closely watching the rapidly inflating valuation, believes he has secured a pivotal seat in the era of superintelligence.
But behind this document are nothing but bottomless holes.
The same document forecasts that over the five-year period from 2026 to 2030, OpenAI will accumulate negative free cash flow of $278 billion. The costs for computing power and data center infrastructure alone will amount to $856 billion, exceeding the total projected revenue over the entire five-year period.
Every step of technological iteration is losing money in the billions.
This is why Son Masayoshi is needed. A company with grand ambitions but persistent losses urgently requires a buyer willing to back it with the entire conglomerate’s credit.
The assets SoftBank regained are currently stagnant.
The agreement clearly states that all subscriptions are for preferred shares, which cannot be converted into freely tradable shares until the public listing and bell ringing. Outside the exchange, there are almost no institutions willing to take on a position worth hundreds of billions of dollars. Converted into $64.6 billion in holding certificates, these are merely a stack of papers locked in a safe until the actual listing and bell ringing.
Masayoshi Son is no stranger to waiting.
In 1995, he bet on Yahoo, and his paper gains once surged over three hundredfold, briefly propelling him to the top of the world’s richest list; his $20 million investment in Alibaba lay dormant for over a decade before finally being rewarded by the iconic bell ring at the New York Stock Exchange that echoed around the globe.
Two famous battles instilled in him a belief in major trends, but also caused him to overlook a key premise: back then, whether it was Yahoo or Alibaba, the money at the bottom was always their own.
You can afford to lose your own spare money; the worst-case scenario is simply cutting your losses and stepping away. As long as you stay at the table, there’s always a chance that a miracle might happen.
In the official announcement on February 27, he spoke without hesitation:
AI is changing the world at an unprecedented pace. OpenAI is a clear leader, with world-class technology and an unparalleled global user base, and we are confident in its continued growth.
Grand visions belong to the era; the cost of realizing them is written in another contract full of loan terms.
Four billion floating bridges
On March 27, SoftBank signed an unsecured bridge loan totaling $40 billion.
The lead arrangers are JPMorgan Chase, Goldman Sachs, Mizuho, Sumitomo Mitsui, and Mitsubishi UFJ, backed by a syndicate of more than twenty other major international banks. The entire loan is unsecured by any physical assets and has a one-year term, maturing on March 25, 2027.
Wall Street approved this massive exposure based solely on SoftBank’s corporate credit and Arm, which sits at the bottom of the balance sheet. In the credit market, a bridge loan is inherently a temporary instrument, designed only to fill the gap until long-term funding is secured.
But SoftBank's withdrawal speed is extremely fast.
On April 1, $10 billion was withdrawn to deliver the first tranche, and another $10 billion was withdrawn on July 1 to deliver the second tranche. SoftBank also withdrew an additional $10 billion in April as a liquidity reserve. Of the $40 billion facility, $30 billion has now been drawn down.
More subtly, September.
On September 9, SoftBank announced it would repay $25.9 billion of its outstanding balance ahead of schedule on September 15. The loan has not been fully settled, leaving approximately $4.1 billion still outstanding on the books; the official announcement made no mention of the source of funds to cover this over $20 billion.
The swap is not a hasty move. In the initial investment announcement on February 27, SoftBank clearly stated that the funds would first be covered by a bridge loan, followed by replacement with existing assets and long-term financing.
At the end of August, market reports indicated that Mizuho Bank was leading a $10 billion two-year loan with a spread of approximately 275 basis points, directly intended to refinance the bridge loan. CFO Yoshimitsu Goto stated on the earnings call that SoftBank has no reason to wait until close to maturity to act—the refinancing will only happen earlier.
SoftBank is replacing one-year short-term debt with public bonds maturing in three and a half or seven and a half years.
The only goal is to push back the repayment node.
Son Masayoshi is purchasing cutting-edge equity that may take ten years or more to realize, while using rigid debt that must be settled within a year to pay for it.
This is his most effective strategy: using massive capital to suppress probabilities and achieving breakthroughs through sheer force. He once compared his investment style to casting a large net to catch fish, rather than fishing with a rod.
But even the main network can sometimes fail to cover everything.
The Vision Fund’s $11 billion investment in WeWork ultimately ended in liquidation, and SoftBank recorded losses exceeding $32 billion in fiscal year 2023. Even at such a heavy cost, the losses ultimately consumed only its own capital and fund shares—no syndicate of creditors stood at the door.
Fuse attached to valuation
The loan signed on August 5 is what truly welded together the long-term technological vision with immediate cash pressures.
This transaction has been under intense negotiation behind the scenes since early spring. SoftBank initially aimed to secure $10 billion, but investors expressed significant concerns. Valuing equity in a company that is not yet public—and whose business model is still not fully realized—presents an extremely difficult risk assessment. Negotiations reached an impasse, and the loan amount was ultimately reduced to $6 billion.
Until July, Masayoshi Son personally increased his commitment, deciding that SoftBank Group would provide full corporate guarantee, at which point the syndicate finally agreed to restore the loan amount to $10 billion.
The loan was officially finalized on August 5. The borrower is SVF II TSUBAKI (DE) LLC, a wholly owned subsidiary of Vision Fund II, registered in Delaware, with a two-year term expiring in August 2028. The lead syndicate includes an elite group of institutions: Goldman Sachs, JPMorgan Chase, Mizuho Securities, Apollo, and Sumitomo Mitsui Banking Corporation.
The spread quoted during the negotiation phase reached as high as 425 basis points, a full 150 basis points higher than the standard loans used to secure the bridge loan at the same time. This was the risk premium Wall Street imposed on illiquid private equity. The contract specified that the funds were not for direct capital injection, but for general corporate purposes of SoftBank Group and Vision Fund II, to support routine liquidity needs.
When external reports covered this transaction, they commonly summarized it as SoftBank borrowing billions by pledging its OpenAI shares.
However, the financial statements record a completely different structure: the collateral under the agreement has always been merely a cash collateral account in the borrower’s name, and OpenAI’s shares were never actually pledged.
The权益 locked in the drawer serves as a taut cursor within the entire architecture. The contract explicitly stipulates that if the fair value of the OpenAI preferred shares referenced in the agreement experiences a significant decline, it will trigger the cash collateral top-up provision and the mandatory early repayment clause.
The shares were never handed over to the bank, but the price tag on the shares was set up as a target.
Whenever the valuation shrinks, creditors have the right to demand that SoftBank promptly deposit actual cash into the collateral account or directly reclaim the principal and interest of the loan. At that point, SoftBank’s preferred shares locked away in storage still have no legal pathway to liquidity; the shares have not been delivered, yet the valuation has become a trigger that banks can activate at any time.
Extremely dangerous.
In 2000, the dot-com bubble burst, causing SoftBank's stock price to drop by 90%. Afterward, Masayoshi Son himself borrowed money from banks by pledging his personal SoftBank shares, but since these were shares held in his own name, any losses were confined to his personal accounts and did not affect the group.
This time is completely different. The shares themselves weren’t pledged at all, but SoftBank’s stock price became the bank’s benchmark for assessing the safety of this loan. In the past, a drop in stock price only meant a reduction in numbers on the balance sheet; now, if the stock price falls too much, SoftBank is directly forced to cash in to make up the difference.
This becomes even clearer when compared to SoftBank’s other margin loan backed by Arm shares—those loans are non-recourse to the group, meaning if the bank loses money, it can only claim against the Arm shares. In contrast, for this OpenAI loan, SBG is the guarantor; Goto explicitly stated on the earnings call that it is “recourse to SBG,” and when calculating LTV (loan-to-value ratio), this debt is fully counted as SoftBank Group’s liability with no adjustments whatsoever.
In other words, instead of borrowing $10 billion, this time the bank is demanding SoftBank Group’s entire balance sheet as collateral.
Goto said on the call that he believes the safety cushion is "more than sufficient." However, contract terms don't follow feelings—if the stock price falls below the agreed level, SoftBank must either post additional margin or repay the loan early.
Worse still, liquidity would be immediately locked up. On the day margin must be topped up, SoftBank would have to transfer actual funds into the collateral account, where the money would be instantly frozen—unable to be used for new investments or to repay other debts.
It will cost money.
A drop in stock price requires additional margin, but that’s an issue that arises only when something goes wrong. More commonly, the cost is interest, which accumulates quarterly as long as the funds remain on the account.
From April to June 2026, SoftBank Group and its financing subsidiaries incurred interest expenses of ¥281.65 billion, an increase of ¥147.3 billion compared to the same period last year. At an exchange rate of 150, this amounts to approximately $1.88 billion.
A $11.5 billion Arm margin loan borrowed in December 2025, a $20 billion bridge loan drawn in April 2026, rising corporate bond balances, combined with rising global interest rates—these massive new debt tranches stacked together caused interest costs to jump significantly.
SoftBank had originally counted on OpenAI going public soon to liquidate its equity stake and cover this fixed debt.
But Sam Altman told Fortune in September that, due to concerns about AI safety, now is not the right time to ring the bell.
The four words "bad timing" translate to real money on SoftBank's books.
The bridge loan is due in March 2027, and the valuation-linked loan is stuck until August 2028. Simply extending $10 billion in debt by one year, at an 8% interest rate, amounts to $800 million in additional interest.
Harder still, OpenAI itself cannot stop.
A cumulative cash flow gap of $278 billion over five years amounts to more than $55 billion per year. OpenAI’s spending rate requires continuous rounds of funding to sustain operations.
SoftBank is the one putting in the most money; whether it can recoup its investment depends entirely on whether OpenAI can make it to an IPO. If OpenAI’s next funding round falls through, SoftBank’s prior investments could be wiped out entirely.
So you can’t afford to stop—you have to keep paying interest on the money you’ve already borrowed while simultaneously securing the next round of funding to keep your investment alive until IPO.
Hidden card
With so much debt, why is Wall Street still willing to lend him tens of billions of dollars?
The answer is Arm.
This UK-based Cambridge chip architecture company nearly monopolizes 99% of the global smartphone processor architecture. In 2016, Masayoshi Son raised $32 billion in cash by significantly reducing his Alibaba shares to take the company private. By the autumn of 2023, when it relisted on U.S. markets, Arm benefited from intense demand for AI chips’ underlying computational efficiency, causing its stock price to surge approximately 270% within the year and its market capitalization to briefly exceed $300 billion.
SoftBank still holds 90% of Arm's equity, generating a paper unrealized gain of over $220 billion from this asset.
Banks don’t care how sophisticated the large model is—they only recognize the hard currency in SoftBank’s hands.
The SoftBank management team most likes to show the market one metric: LTV (net debt divided by stake value), which measures the proportion of borrowing relative to assets. From March to June this year, this figure dropped from 17% to 13%, suggesting on the surface that leverage has declined.
But this is just a numbers game.
During the same period, SoftBank’s net debt actually increased from ¥8.2 trillion to ¥10.8 trillion. The borrowed money didn’t decrease—it rose by a full ¥2.6 trillion. The so-called reduction in leverage was merely due to Arm’s stock price surge, which inflated the denominator in the calculation.
Asset appreciation does not equate to cash inflow. SoftBank’s cash and cash equivalents on its balance sheet did not increase; instead, they declined from ¥3.5 trillion to ¥2.3 trillion, as large sums were flowing out to make scheduled payments to OpenAI and to repay existing debts.
Relying on this passively expanded denominator, SoftBank portrays a sense of security in its financial reports, as the 13% level is still far from the group’s internal threshold of 25%, prompting S&P to upgrade its rating outlook from “negative” to “stable”.
But this masks the looming repayment peak.
Over the next two years, SoftBank has approximately ¥1.5 trillion in corporate bonds maturing, equivalent to about $9 billion, with one tranche due next year and another the following year. Meanwhile, private equity giant Apollo is also negotiating to increase a NAV loan for Vision Fund II from $5.4 billion to $9 billion.
Wall Street still circles around SoftBank because beneath its balance sheet lies Arm, a card that can still be cashed in.
Back then, regulators in the UK and the EU blocked the merger between Arm and NVIDIA, leaving Masayoshi Son with only a $1.25 billion breakup fee; after all this, the chip company has now become his final pillar for maintaining the entire credit chain.
On September 24, the $10 billion high-risk bond will finalize its interest rate. This figure is the open market’s explicit price tag on Masayoshi Son’s question: “How much longer must we wait for OpenAI?” The higher the price, the more Wall Street views this wait as risky.
Settlement on September 29, funds transferred on October 1. Then, the pointer continues to turn.
He is waiting for OpenAI to go public, while his creditors are waiting for him to repay the principal and interest by March 2027.
Money doesn't wait.
Twitter: https://twitter.com/BitpushNewsCN
BitPush Telegram community: https://t.me/BitPushCommunity
BitPush TG subscription: https://t.me/bitpush