SoftBank’s AI Huge Bet Is Facing a Credit Test

SoftBank’s nearly 45% one-month share-price decline is dramatic, but the more consequential signal is coming from the credit market. Its five-year credit-default-swap spread widened to as much as 466 basis points on November 6, the highest level since October 2022. Markit data cited by Reuters put the implied default probability at 30.6%. That is not a forecast that SoftBank will default; CDS pricing embeds recovery assumptions, liquidity and risk premiums. It does show that lenders and hedgers are assigning a much higher price to balance-sheet risk than they were a few months ago.

The timing is uncomfortable. SoftBank has spent 2026 deliberately increasing its exposure to the AI stack: it completed a $30 billion follow-on investment in OpenAI, taking cumulative investment to $64.6 billion and its ownership to about 13%; it completed the acquisition of DigitalBridge; and it has been pursuing ABB’s robotics business as part of Masayoshi Son’s physical-AI strategy. Those investments were made while SoftBank’s own financial policy still targeted loan-to-value below 25% under normal market conditions.

The problem is that SoftBank is an investment holding company whose borrowing capacity is partly supported by the market value of the assets it owns. When Arm, AI infrastructure assets and private-company valuations rise, the balance sheet looks stronger. When those assets fall together, the denominator in the LTV calculation shrinks even if debt has not changed. That is why a broad AI selloff can migrate from an equity problem into a credit problem.

Arm Is the Anchor - and That Creates Concentration Risk

SoftBank’s June 30 NAV disclosure shows just how concentrated the structure has become. Adjusted equity value of holdings was JPY83.11 trillion, of which Arm accounted for JPY49.91 trillion after adjusting for asset-backed financing. That means roughly 60% of SoftBank’s holdings value was tied to Arm alone. SoftBank Vision Fund 2 contributed another JPY19.29 trillion, or about 23%, giving the group unusually high exposure to assets whose valuation is closely linked to AI expectations.

This concentration helped SoftBank enormously during the AI rally. It also works in reverse. A decline in Arm’s share price immediately reduces the market value of SoftBank’s largest holding, while weaker private-market valuations can reduce the value attributed to SVF2 and OpenAI. The company can respond by selling assets, raising equity-like financing or using asset-backed loans, but all three choices can become less attractive when markets are falling.

SoftBank’s own CFO has described this relationship clearly. NAV is market-driven and cannot be fully controlled, while LTV can be controlled by managing debt, monetizing holdings and using asset-backed or hybrid financing. That discipline is useful, but it also exposes the reflexive nature of the model: if asset values fall hard enough, SoftBank may have to reduce debt or monetize assets at precisely the moment it would prefer to keep investing.

Figure 1. SoftBank’s June 30 adjusted holdings composition. Source: DaveFinances.com.

At June 30, SoftBank reported adjusted net debt of JPY10.81 trillion against JPY83.11 trillion of adjusted holdings, producing an LTV of 13.0%. That was comfortably below the company’s normal 25% ceiling and its 35% emergency threshold. On that snapshot, the balance sheet did not look stretched.

But that number is now stale. SoftBank’s next quarterly financial report is due on November 10, and markets have moved sharply since June. An illustrative sensitivity shows why credit investors are watching the asset side so closely. Holding June net debt constant, a 20% decline in portfolio value would lift LTV to about 16.3%; a 40% decline would take it to roughly 21.7%; and a 50% decline would push it just above the 25% normal-policy ceiling.

Those are not forecasts, and the actual portfolio does not move uniformly. SoftBank also has cash, financing flexibility and the ability to monetize assets. The exercise simply shows why a 45% fall in SoftBank’s own share price and a broader selloff in AI-linked assets can matter even before a single operating subsidiary misses a payment. Credit risk rises because the cushion beneath the debt is marked to market.

Figure 2. Illustrative LTV sensitivity using SoftBank’s June 30 asset and net-debt values. Source: DaveFinances.com

OpenAI Has Become a Funding Decision, Not Just an Investment

SoftBank’s OpenAI exposure is the other reason the market is reassessing its credit profile. The group completed its final $10 billion follow-on tranche on October 1, bringing cumulative investment to $64.6 billion. The final payment was funded with proceeds from a large foreign-currency bond issue rather than from the bridge loan initially arranged for the transaction.

The pricing of that debt is revealing. SoftBank issued $10 billion of dollar notes with coupons of 8.625%, 9.25% and 9.75%, plus EUR1 billion of euro notes carrying 7.125% and 8.0% coupons. These are expensive funding costs for an investment whose liquidity event has reportedly moved further into the future. Reuters reported that OpenAI’s expected public offering has been pushed to 2027, delaying the point at which SoftBank might realize part of the investment through public markets.

Using the issued principal and coupons, the September transaction alone implies roughly $940 million of annual coupon payments on the dollar notes, plus about EUR76 million on the euro notes, before considering fees or any hedging. The point is not that SoftBank cannot service that cost. It is that the return hurdle on the assets purchased with that financing is now materially higher.

Figure 3. Coupon rates on SoftBank’s September 2026 foreign-currency senior notes. Source: DaveFinances.com

The reported Chinese prohibition on SoftBank’s $5.375 billion acquisition of ABB’s robotics business adds another complication, but it should be separated from the credit story. ABB confirmed on November 5 that China’s foreign-investment security review mechanism had prohibited the acquisition of ABB Robotics’ business in China. SoftBank issued its own update the following day and said it remained in constructive discussions with ABB.

The transaction was central to Son’s physical-AI strategy because China is a major robotics market and an important manufacturing base for ABB. Losing the Chinese operation would reduce the strategic value of the acquisition and could force the parties to restructure the deal or abandon it. That is a genuine setback to the investment thesis.

Financially, however, a blocked acquisition can also conserve capital. If SoftBank ultimately does not pay the full $5.375 billion enterprise value, the immediate balance-sheet effect would be less negative than completing the purchase at the original terms. The damage is therefore more strategic than mechanical: SoftBank loses a route into physical AI just as investors are already questioning whether Son is stretching the group across too many capital-intensive projects.

Figure 4. Scale of selected SoftBank AI and infrastructure transactions. Source: DaveFinances.com

Alphabet, Amazon, Microsoft and Meta are also spending enormous sums on AI, but they finance much of that spending from operating cash flow generated by large, profitable businesses. SoftBank’s model is different. It owns a portfolio of assets, raises debt against the strength of that portfolio and recycles capital into new investments. That makes asset prices and funding costs central to the investment thesis.

The September bond sale illustrates the difference. SoftBank was able to raise more than $11 billion, proving that it still has substantial market access, but the coupons reached levels normally associated with high-yield credit. If CDS spreads remain near 466 basis points, future unsecured borrowing could become even more expensive. Asset-backed borrowing could also become less attractive if the collateral itself is falling in value.

This is the feedback loop investors are beginning to price: weaker AI sentiment reduces the value of Arm and private AI holdings; lower asset values increase perceived leverage; higher perceived leverage raises funding costs; and higher funding costs make the next AI investment harder to justify. That does not mean the loop must end in distress. It does mean SoftBank’s equity can become far more volatile than the underlying operating performance of Arm or OpenAI.

November 10 Is the Real Test

SoftBank’s November 10 earnings report should matter more than another volatile session in Tokyo. Investors need an updated LTV, cash position and breakdown of holding values after the OpenAI funding, the DigitalBridge acquisition and the recent decline in AI-linked assets. They also need to know how much additional debt Son expects to raise and whether the company still intends to prioritize asset-backed financing over outright asset sales.

The OpenAI carrying value will be especially important. SoftBank’s Vision Fund CEO said the group’s SVF2 stake in OpenAI was valued at $79.6 billion at the March 2026 fiscal year-end, before the final follow-on tranches were completed. Investors will want to see how that value evolves relative to the $64.6 billion cumulative investment and whether any delay to OpenAI’s IPO changes SoftBank’s expected monetization timeline.

The market is not saying SoftBank’s AI investments are worthless. It is saying that the capital structure supporting them deserves a larger risk premium. If updated LTV remains conservative and SoftBank shows ample cash, the CDS widening may prove excessive. If debt has risen faster than portfolio value and Son signals another wave of large investments, the current selloff will look less like panic and more like a repricing of SoftBank’s financing model.

Original Post