🔍 Read the full analysis: SoftBank’s $7 Billion Loss Puts The Cost Of AI Power Equipment In Focus on ThorstenMeyerAI.com
Get business pricing on monitors, keyboards and dev gear
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
SoftBank Group reported a ¥1,115.3 billion ($7.0 billion) net loss for July–September 2026, with a ¥597.2 billion ($3.73 billion) loss tied to changes in power-equipment procurement for U.S. AI data centers. The episode shows how advance commitments to secure scarce equipment can leave developers with losses when project requirements change; the full supplier and project details remain undisclosed.
SoftBank Group reported a ¥1,115.3 billion ($7.0 billion) net loss for the July–September 2026 quarter, including a ¥597.2 billion ($3.73 billion) loss connected to revised power-generation equipment plans for U.S. AI data centers. The charge shows the financial exposure created when data-center developers pay ahead to reserve equipment and factory capacity, then change their plans.
SoftBank said a subsidiary, Silver Bands 7 (SB7), paid ¥1.5 trillion, or about $9.48 billion, by September 30 for power-generation equipment and reserved manufacturing capacity. The payments were made as advance investments to secure equipment amid anticipated shortages. SB7 is an entity in which SoftBank Group and its second Vision Fund have invested.
After reviewing its projects, SB7 revised its procurement strategy. SoftBank recorded ¥457.0 billion ($2.9 billion) as losses on non-refundable advance payments for equipment no longer included in the revised plans, and another ¥140.2 billion ($0.9 billion) for contract modifications and settlements. SoftBank said cash payments related to the latter amount are expected from the third quarter onward.
SoftBank said ¥6.6 billion in advance payments remains allocated to equipment for the PORTS-Pike Technology Campus in Ohio and other projects. It also said it may receive rebates of up to $2.1 billion on future purchases over 10 years, subject to conditions. The company did not recognize those potential rebates as an asset because the amount could not be determined.
The power-equipment charge was one major factor in the quarterly loss, not the only one. Finance costs rose to ¥461.9 billion from ¥174.7 billion in the same quarter a year earlier. Net sales increased to ¥2,215.4 billion from ¥1,916.5 billion, but the company’s net result swung from a profit of ¥2,502.2 billion a year earlier to a loss. SoftBank reported a ¥347.3 billion profit in the April–June quarter, leaving a ¥768.0 billion net loss for the first half of its fiscal year.
SoftBank’s Loss Puts AI Power Costs in Focus
Revised equipment plans turned advance commitments into major losses. The case shows how securing scarce power equipment can expose data-center developers when project needs change.
How the commitments changed
SB7 revised its procurement strategy as project requirements evolved.
Advance investments secured power-generation equipment and manufacturing capacity amid expected shortages.
SB7 changed procurement to reflect the needs of existing and planned projects.
Non-refundable payments and contract changes led to a ¥597.2 billion loss.
Where the power-equipment loss came from
The recorded charge had two parts.
| Cost | Amount | What it covers |
|---|---|---|
| Non-refundable advances | $2.9B | Equipment removed from revised plans |
| Modifications and settlements | $0.9B | Contract changes; cash payments expected from fiscal Q3 onward |
What remains uncertain
The disclosure leaves important details open for investors and data-center customers.
- ¥6.6 billion in advances remains allocated to PORTS-Pike and other projects.
- SoftBank expects to recover its investment through project financing.
- Suppliers, equipment details, and the split of remaining advances.
- Financing timetable, project effects, and whether conditional rebates will be received.
The Cost of Securing Data-Center Power
AI data centers need more than processors and buildings: they require dependable power infrastructure, including generation equipment that can take time to procure. SoftBank’s reported losses show one way scarcity can create costs before a facility is operating. Advance payments can secure capacity, but they may be difficult to recover if equipment orders no longer match project needs.
The case also illustrates that the exposure can reach beyond the company building a data center. SoftBank’s source material describes a network involving its investment in OpenAI, data-center plans, and equipment commitments. It says OpenAI is expected to be a long-term tenant and that Nvidia guarantees part of the tenant’s lease obligations. These are distinct relationships, and the available figures do not establish that any one of them caused the equipment charge. Still, changes in projects, financing or demand could affect parties across the chain.
For companies seeking AI computing capacity, the report is a reminder that supply constraints may involve power equipment and factory slots as well as chips. The financial consequences for customers will depend on their contracts and providers; SoftBank’s disclosure does not show that prices or service availability have changed.
Popular right now
Shark NV752 Rotator Powered Lift-Away TruePet Upright Vacuum, Bordeaux
Vitamix Touch & Go Advance 32-Ounce Bar Blender, White
Dyson V15 Detect Cordless Vacuum
As an affiliate, we earn on qualifying purchases.
How SoftBank’s Orders Changed
SoftBank’s fiscal year runs from April through March, so the reported second quarter covers July through September 2026. In its November 10 financial report and investor materials, the company said SB7 entered into manufacturing-capacity reservation agreements, equipment purchase contracts and other arrangements for power-generation equipment for U.S. AI data-center projects, including PORTS-Pike in Pike County, Ohio.
SoftBank said it made the advance investments because it anticipated equipment shortages. It reported payments of $3.9 billion in fiscal 2025 and $5.6 billion in the first half of fiscal 2026. It later changed the procurement strategy in response to the evolving requirements of existing and planned projects. The company said it expects to recoup the investment through project financing, but did not provide a timetable or detailed financing plan in the supplied materials.
At a Tokyo briefing, SoftBank Chief Financial Officer Yoshimitsu Goto said competition for energy supply and power equipment had led the company to buy before long-term specifications were final, Reuters reported. Chief Accounting Officer Kazuko Kimiwada said the cancelled orders could not be refunded, according to the source material. SoftBank’s report says its financial statements are an English translation of the Japanese original and had not been reviewed by auditors.
““made bridge investments in advance to secure power generation equipment in anticipation of shortages””
— SoftBank Group, investor presentation
Unreported Suppliers and Project Details
SoftBank has not identified the equipment involved, the suppliers, or the specific purchases that were cancelled or modified. It also has not detailed how the remaining $6.6 billion of advance payments is divided among PORTS-Pike and other projects. The amount and timing of any future rebates remain uncertain because they depend on conditions and were not recorded as an asset.
The company’s expectation that it can recover its investment through project financing is a stated outlook, not a completed recovery. The supplied information does not say which lenders or financing terms may be involved. Nor does it establish how the procurement changes affect construction schedules, data-center capacity, tenant commitments or customers’ access to computing services.
The quarterly net loss also reflects factors beyond power equipment, including increased finance costs and lower investment gains than in the comparable quarter. The disclosed charge should not be treated as a complete measure of the cost of SoftBank’s AI infrastructure plans.
Financing and Equipment Recovery
SoftBank said cash payments for the contract modifications and settlements are expected from the third quarter of fiscal 2026 onward. The company’s next financial updates may clarify whether it secures project financing, how the remaining equipment commitments are used, and whether it can qualify for the potential rebates.
Investors and data-center customers will also be watching for information on project schedules and power-equipment availability. SoftBank has not provided a specific timetable for those updates in the cited material, so the effect on construction and operating capacity remains open.
Key Questions
Why did SoftBank report a $7 billion quarterly loss?
SoftBank reported a ¥1,115.3 billion ($7.0 billion) net loss for July–September 2026. A ¥597.2 billion ($3.73 billion) loss tied to revised power-equipment procurement plans was a major factor. Higher finance costs and other changes also affected the result.
What power-equipment costs did SoftBank record?
SoftBank recorded $2.9 billion in losses on non-refundable advance payments for equipment no longer in its revised plans, plus $0.9 billion for contract modifications and settlements.
Could SoftBank recover some of the advance payments?
SoftBank said it expects to recoup the investment through project financing and may receive rebates of up to $2.1 billion over 10 years on future purchases, subject to conditions. It did not recognize the potential rebates as an asset because their amount could not be determined.
Which projects were linked to the equipment orders?
SoftBank named U.S. AI data-center projects, including the PORTS-Pike Technology Campus in Ohio. It has not disclosed the suppliers, equipment types or detailed allocation of the remaining advance payments.
Does the loss mean AI data-center construction will be delayed?
The reported information does not establish whether construction or service schedules will change. SoftBank has not described the specific effect of the procurement revision on project timelines or data-center capacity.
Source: ThorstenMeyerAI.com
