Why SoftBank Lost $7 Billion: Masayoshi Son’s AI Bet Explained

SoftBank, the Japanese technology investment group led by Masayoshi Son, has reported a loss of about $7 billion for July-September 2026. Son has invested heavily in artificial intelligence, from OpenAI to the data centres and power equipment needed to run AI systems. The latest results show how expensive that expansion can become when plans change.

The biggest single charge was $3.73 billion linked to power-generation equipment for US AI projects. SoftBank had paid to reserve equipment in advance, but revising its procurement plans meant losing some of those payments and incurring costs to change contracts. Losses on Intel shares and a higher financing bill added to the pressure.

At the same time, falling investment values have reduced SoftBank’s estimated asset value and affected Son’s personal wealth. To understand the damage, investors need to distinguish the money committed to projects, changes in investment values and the losses recorded in its results.

What drove SoftBank’s $7 billion quarterly loss?

The biggest identifiable charge concerned equipment procurement. But the result also reflected a sharp reversal in investment earnings and a much larger financing bill. Rising revenue from operating businesses could not offset those pressures.

Sources: SoftBank Group Q2 FY2026 consolidated financial report, supplemental data sheet and investor accounting presentation, November 10, 2026. 

SoftBank’s earnings include changes in investment values alongside income and costs from its businesses. Consequently, stronger sales do not necessarily produce a stronger bottom line. The size of the investment swing helps explain why treating SoftBank like a conventional telecom company gives an incomplete picture.

How did AI equipment commitments become a $3.73 billion loss?

SoftBank committed money early to secure power-generation equipment, anticipating competition for supplies. When its procurement plans changed, some payments could not be recovered and modifying contracts created additional costs.

The latest reporting identifies Son’s large gas-turbine reservations as central to the setback. These machines generate electricity, making them important to the power supply for AI data centres. Reserving equipment ahead of shortages can protect a construction schedule, but it also puts money at risk before projects are ready.

The equipment figures describe money already committed, losses recognised and possible future discounts. Keeping them separate makes the cost easier to understand.

Sources: SoftBank Group Q2 FY2026 consolidated financial report, equipment disclosure and Note 12; The Wall Street Journal, November 11, 2026, reporting on Son’s turbine reservations.

The distinction matters: the charge recognises the reduced value of earlier payments and additional contractual obligations. It does not mean the entire amount left SoftBank’s bank account during this quarter. Equally, describing everything as a paper loss would understate the money already committed.

The rebates are conditional discounts on future purchases, rather than an immediate refund. They had not been recognised as an asset at September-end. Subtracting them from today’s charge would assume purchases and recoveries that have not yet occurred.

SoftBank has clarified that the charge is unrelated to Ports-Pike, its Ohio project. Its November 12 update sets out when it expects to recover the remaining equipment advances:

The financial test now is whether those remaining advances are recovered as expected. A recognised loss establishes what has already gone wrong; the recovery schedule identifies what still needs to go right.

The procurement loss has also prompted a dispute about default. Elon Musk alleged that SoftBank had defaulted on turbine orders; CFO Yoshimitsu Goto rejected that claim at the November 10 briefing. Changing reservations and recognising contractual costs do not, by themselves, establish a failure to meet payment obligations.

Source: SoftBank Group November 10, 2026 earnings-call Q&A, as transcribed by Investing.com.

Why did Intel and financing costs add to the pressure?

SoftBank faced both a decline in an investment and costs from financing arrangements linked to it. Those are separate exposures.

Its Intel holding lost value during the quarter. Separately, financing contracts backed by Intel shares also generated a loss. Those contracts provide cash upfront, with future settlement linked to the shares.

Sources: SoftBank Group Q2 FY2026 earnings presentation.

The two losses use different starting points. Intel declined between the June and September quarter-ends, while its price increased between the August financing transaction and September-end. That later increase raised the value of SoftBank’s contractual obligation.

These individual losses are already reflected in the reported results; adding them to the headline loss would double-count them. Gains elsewhere also offset part of the group’s derivative losses.

The financing bill creates a more persistent concern. More borrowing means more interest to service even when investment values disappoint. The period’s costs also included accelerated recognition of remaining financing fees after early bridge-loan repayment, so the latest bill should not simply be annualised as a recurring expense.

Has SoftBank lost money on its OpenAI investment?

The September results did not record a new OpenAI valuation loss. SoftBank’s valuation basis was unchanged, which also meant there was no fresh valuation gain to repeat the boost seen in the previous year.

The disclosed figures distinguish the amount invested from the estimated value of that investment:

Sources: SoftBank Group Q2 FY2026 investor accounting presentation and October 1, 2026 OpenAI investment completion announcement. 

The widely cited $64.6 billion therefore describes money invested, rather than a loss or the latest market value of SoftBank’s stake. The earlier gains also do not represent cash already realised through a sale.

There is a newer valuation development. November 6 reporting described an initial $11 billion funding close valuing OpenAI at $1.25 trillion, with SoftBank participating. That is a reported financing event after the September reporting period, rather than a valuation already reflected in these results.

A higher valuation could support a later uplift, but the exact benefit depends on the financing terms, dilution and accounting assessment. Multiplying the new headline valuation by the previously disclosed ownership percentage would give a misleading impression of precision.

How much has SoftBank’s valuation fallen and why does Arm matter?

SoftBank’s net asset value or NAV, estimates the value of its holdings after subtracting adjusted net debt. It is a more useful starting point here than a price-to-earnings ratio built on volatile investment profits.

The company’s November 9 pro forma calculation put NAV approximately 25.3% below September-end in yen terms. SoftBank’s own share price had fallen more sharply across those dates, widening its discount to the asset estimate.

Arm is particularly important because it represents SoftBank’s largest asset exposure. At September-end, its value after asset-backed financing adjustments represented approximately 54% of the adjusted equity value of holdings.

Sources: SoftBank Group NAV disclosure and Q2 FY2026 earnings presentation; Investing.com historical Tokyo prices. Also, NAV is before tax. 

However, changes in Arm’s quoted share price do not flow directly into SoftBank’s consolidated investment profit or loss because Arm is a consolidated subsidiary. They still affect the portfolio’s estimated value and financing flexibility. This explains how a share-price decline can damage NAV without creating an equivalent charge in quarterly earnings.

A wider NAV discount does not establish that the shares are automatically undervalued. Investors must assess how reliable the asset marks are, what can actually be sold and how much value future funding commitments may absorb.

How has Masayoshi Son been affected and what happened to the shares?

Son’s estimated fortune fell $3.9 billion to about $29 billion during the November 6 selloff.

A decline in the market value of his holdings reduces estimated wealth. It does not establish that he sold investments or paid out an equivalent amount in cash.

The immediate earnings reaction also needs context. SoftBank rose 13.88% on November 10 before results were released after the Tokyo close. On November 11, it fell as much as 10.5% intraday, then recovered to close at ¥3,855, down 0.44%.

Calling the earlier rally a response to the results, or presenting the next morning’s steepest fall as the closing decline, would misread the sequence.

What could constrain Son’s next AI investments?

Funding capacity and regulatory approvals are two immediate constraints. The September disclosures show a funding buffer, alongside growing sensitivity to asset prices.

That provides a buffer, but falling asset values can raise leverage ratios even without fresh borrowing. Share-backed financing can also require additional collateral or repayment if specified thresholds are breached.

SoftBank has fully repaid and closed the bridge facility used for its scheduled OpenAI investments. However, this involved refinancing and other funding arrangements. Closing one facility improves the maturity profile without necessarily reducing total indebtedness by the same amount.

Another unresolved issue concerns the planned ABB robotics acquisition. SoftBank confirmed that China prohibited the acquisition of ABB’s robotics business in China, while discussions with ABB continued. The announcement did not declare the entire global transaction terminated.

That is a separate strategic obstacle. November regulatory developments cannot explain losses recorded in a quarter that ended in September.

Source: SoftBank Group, November 6, 2026 ABB robotics acquisition update.

Author’s take: What should investors check next?

Use the next disclosures to test whether SoftBank is recovering committed money and strengthening its funding position. A rebound in AI share prices would help NAV, but it would not by itself demonstrate that infrastructure commitments are earning acceptable returns.

First, track the equipment advances against the recovery timetable. Look for amounts actually recouped, revisions to timing and any additional charges. A forecast recovery becomes more useful evidence when financing closes and money returns.

Second, examine any new OpenAI mark alongside the investment cost, ownership changes and cash committed. A valuation uplift can improve reported wealth while leaving funding requirements substantial.

Third, update the NAV discount and leverage measures together, using compatible dates and definitions. A bigger discount offers less reassurance if the asset estimate is falling or financing flexibility is shrinking. Conversely, recovering advances and maintaining access to funding would provide evidence that the setback is becoming more manageable.

The next useful milestone is a measurable reduction in the money still at risk, rather than another larger headline valuation.