“$65 Billion Bet on OpenAI’s Listing” IPO Delay Turns SoftBank’s Debt Pile Into a ‘Ticking Time Bomb,’ as Rising US Rates Compound Refinancing Pressure

OpenAI IPO delay disrupts SoftBank’s liquidity plan
Blocked share-backed financing compounds investment and debt-repayment burdens
Funding-recovery pressure amid high interest rates raises fears of cascading shocks across AI finance

OpenAI’s decision to postpone its initial public offering (IPO) is disrupting the financing plans of SoftBank, one of its largest investors. With its plan to secure liquidity by borrowing against its stake after the listing now derailed, SoftBank must simultaneously fund large-scale investments and repay existing debt. OpenAI is likewise struggling to enter the public market at its current valuation as revenue growth slows and computing costs reach astronomical levels. Concerns are mounting that a prolonged delay could intensify SoftBank’s funding pressure and transmit shocks to data-center lending and private-credit markets already vulnerable to high interest rates.

SoftBank Extends $6.5 Billion Credit Facility by One Year

According to Bloomberg on September 20 local time, SoftBank extended an existing $6.05 billion credit facility due to expire this month by one year. The move is widely viewed as a response to the suspension of OpenAI’s IPO plans. SoftBank’s financing structure had been predicated on an OpenAI listing. According to company disclosures, SoftBank has committed approximately $65 billion to OpenAI through October. The $64.6 billion already committed corresponds to a stake of roughly 13%. SoftBank had planned to use the official market price established through the listing to pledge the shares as collateral and raise liquidity.

That plan was disrupted when OpenAI Chief Executive Officer (CEO) Sam Altman said on September 12 that there would be no IPO this year. SoftBank must now continue borrowing against an unlisted asset with no observable market price. The company needs to secure $22.5 billion in near-term funding that must be transferred to OpenAI by year-end. Its total committed investment in OpenAI and related AI infrastructure exceeds approximately $60 billion. SoftBank is negotiating with global private-equity firm Apollo Global Management to increase an existing $5.4 billion loan to $9 billion, but no final decision has been made, and it remains unclear whether an expanded credit facility would actually be drawn.

Debt-Funded OpenAI Bet Weighs on SoftBank’s Creditworthiness

SoftBank estimates that it will need $50 billion this year to fund its OpenAI investment and refinance existing borrowings, and has acknowledged that its net loan-to-value ratio (LTV), which it has generally kept below 25%, could temporarily breach that threshold. Even if the expanded Apollo loan materializes, it would provide only an additional $3.6 billion, meaning SoftBank would likely need to combine asset sales with new bond issuance to cover its year-end commitments. Moreover, S&P Global Ratings has revised the outlook on SoftBank’s credit rating to “negative,” estimating that unlisted assets, including OpenAI, could account for more than half of its portfolio. With the IPO delay making it difficult to establish an objective market price for the stake, any markdown in private-market valuations could simultaneously trigger additional collateral calls and higher refinancing rates.

SoftBank is initially responding to its immediate repayment requirements by increasing new borrowing. The company has raised $37 billion through bond issuance and loans this year. Last week, it signed a two-year, $11.87 billion loan agreement with approximately 20 financial institutions, exceeding its $10 billion target. SoftBank plans to use the proceeds to repay the remaining $25.9 billion of a $40 billion bridge loan arranged in March, but fluctuations in collateral value remain the central risk. The company has borrowed $10 billion through a margin loan secured by OpenAI shares, leaving its balance sheet exposed to any decline in OpenAI’s valuation. SoftBank’s credit rating currently stands at BB+, one notch below investment grade, while the cost of insuring its debt has reached its highest level in three years.

IPO Delay Preferred to Valuation Cut

According to The New York Times (NYT), Altman initially sought advice from investment banks (IBs), lawyers and other advisers with a view to listing OpenAI in the third or fourth quarter of this year, but reportedly changed course last month. SpaceX, led by CEO Elon Musk, recently raised a record $85.7 billion through the largest IPO ever, only to disappoint investors with lacklustre post-listing share-price performance. OpenAI’s advisers consequently presented Altman with two options in response to changing market conditions: list next year while preserving the company’s $1 trillion valuation or pursue an earlier IPO at a lower valuation. Altman effectively opted to delay the listing, refusing to accept a valuation adjustment.

OpenAI’s decision to postpone its listing also appears to reflect its belief that it can secure a higher valuation in private markets. After being valued at $852 billion in March, OpenAI is now discussing a new funding round targeting a valuation of more than $1.2 trillion. Private transactions involving a small number of strategic investors can price in long-term growth expectations more broadly, whereas after a listing, the assessments of a much larger investor base are reflected in the share price following every earnings announcement. For OpenAI, it may be more advantageous to use private capital to expand its revenue base and improve cost efficiency before entering the public market.

Table 1. OpenAI Revenue, Cash Flow and Funding Outlook

Slowing User Growth and Weak Monthly Revenue Undermine OpenAI’s Momentum

OpenAI’s decision to postpone its listing was not driven solely by SpaceX’s share-price performance and the desire to protect its private-market valuation. The fundamental issue is OpenAI’s weak financial performance itself. The company continues to target annual revenue three times the $13 billion recorded last year, but its current monthly revenue amounts to only $2 billion. User growth has also moderated, with weekly active users (WAUs) remaining in the 900 million range and still falling short of the 1 billion target set for the end of last year. Newly introduced revenue streams—such as displaying advertisements to users of free and lower-priced plans and adding shopping features through partnerships with e-commerce companies—also remain at an early stage.

According to recent OpenAI documents obtained by the Financial Times (FT), the company’s cumulative free cash flow from 2026 through 2030 is projected at negative $278 billion. Revenue is forecast to increase nearly tenfold, from $36 billion this year to $350 billion in 2030. Projected cumulative revenue from this year through 2030 amounts to $840 billion. Computing resources and infrastructure constitute the largest spending category, with OpenAI expecting related costs to reach $856 billion by the end of 2030. This reflects the enormous capital required to secure the data centers and computing resources needed to train and operate AI models. OpenAI raised $122 billion in March, but under its current spending plan, those funds are also expected to be exhausted by 2028.

The immediate catalyst behind OpenAI’s loss of momentum was Google’s November 18 launch of “Gemini 3.0,” powered by tensor processing units (TPUs), followed by Anthropic’s gains this year. According to The Wall Street Journal (WSJ), OpenAI has fallen behind Anthropic’s Claude in the enterprise market and has repeatedly missed its monthly revenue targets this year. US technology publication The Information reported that OpenAI Chief Financial Officer (CFO) Sarah Friar had already told some colleagues earlier this year that the company was not ready to go public within the year, citing massive spending and procedural issues. Compounding the pressure, Anthropic secured a $965 billion valuation in its Series H funding round last month, overtaking the $852 billion valuation OpenAI recorded at the end of March.

US Treasuries Alone Yield 5%, Intensifying Pressure on AI Capital

Recent increases in US interest rates have also fuelled doubts over whether OpenAI will be able to pursue a listing next year. The US Federal Reserve (Fed) raised its benchmark interest rate by 25 basis points to 3.75–4.00% on September 16, while 16 of its 18 policymakers projected another increase within the year. Long-term US Treasury yields also surged, with the 10-year yield briefly exceeding 5% and the 20-year yield climbing to 5.40%. In an environment where investors can expect returns of more than 5% from US Treasuries alone, the incentive to tolerate losses while locking capital into unlisted AI companies diminishes. Because a higher discount rate also reduces the present value of future earnings, OpenAI will find it even more difficult to secure a valuation exceeding its $1 trillion target.

Higher interest rates are already pressuring the AI investment network built around OpenAI. US Big Tech companies have secured data centers and AI chips through special-purpose vehicles (SPVs), attracting investment by guaranteeing the liabilities incurred by those entities. According to the FT, residual-value guarantees provided by Big Tech companies over the past year total as much as $300 billion. Low interest rates allowed these companies to expand their operations at limited financing cost, but as rates rise, interest burdens increase while the value of collateral assets declines. If OpenAI’s listing delay persists, a reassessment of corporate valuations will become unavoidable, intensifying demands for additional collateral and loan repayments. Data-center and semiconductor projects that fail to secure financing are particularly likely to face delays in their investment schedules.

OpenAI Valuation Decline Risks Cascading Shock Across US and Japanese Finance

These concerns have already begun to be reflected in the pricing of AI-related loans. An $18 billion loan for a New Mexico data center supporting the $300 billion computing agreement between OpenAI and Oracle recently traded at approximately $0.89–$0.91 per $1 of face value. Given that high-quality loans typically trade near par, the discount indicates that investors have begun taking a more conservative view of credit risk and principal-recovery prospects. Arrangers including Santander and Jefferies were forced to retain more of the loan than initially planned because of weak investor demand. With Oracle’s credit rating downgraded in July to BBB-, one notch above speculative grade, and its debt burden continuing to grow, placing the remaining loan exposure is also expected to prove difficult.

AI-related exposure accumulated in the private-credit market is also substantial. According to the Financial Stability Board (FSB), transactions involving AI companies and data centers accounted for more than one-third of all private-credit deals last year. Japanese financial institutions are bound to the same web of interests. Japan’s three largest banks—Mizuho Bank, Sumitomo Mitsui Banking Corporation and MUFG Bank—participated in SoftBank’s large-scale borrowing. A sharp decline in OpenAI’s valuation could result in investment losses and collateral shortfalls at SoftBank, impairing the asset quality of these institutions and constraining subsequent financing for Japan’s AI industry. If loan defaults trigger forced sales of collateral assets and redemptions from private-credit funds, the resulting shock could escalate into a broader credit crunch across US and Japanese financial markets.