Editorial illustration for SoftBank Takes USD 7 Billion Hit on US AI Data Centre Investments
SoftBank's $7B Loss on US AI Data Centre Bet
SoftBank Takes USD 7 Billion Hit on US AI Data Centre Investments
SoftBank paid $9.5 billion in advance between 2025 and the second quarter of 2026 to lock down equipment for the power generation facilities behind its US artificial intelligence data centre buildout. That bet on scarcity has now backfired in part. As SoftBank's data centre plans shifted, so did its equipment needs, leaving some of that prepaid gear unused. The company has written off $3.7 billion tied to that stockpile and to changes in its purchase agreements, though it says the losses leave its PORTS-Pike power project untouched.
The equipment write-off isn't the only hit tied to SoftBank's AI ambitions. A separate loss of roughly $1.3 billion traces back to its arrangement with OpenAI, specifically warrants tied to SoftBank's data centre unit, Energy Global, which gave OpenAI the right to buy shares in that business. SoftBank also booked about $900 million in additional employee-related costs during the period. Together, the figures show the financial strain building up behind one of the largest corporate bets on AI infrastructure anywhere, and how fast-moving technical decisions can turn upfront caution into a costly miscalculation.
SoftBank's massive bet on artificial intelligence infrastructure in the US is proving expensive. The Japanese investment giant reported billions of dollars in losses after revising its plans for equipment used to power AI data centres. It has also recorded losses linked to its agreement with OpenAI.
Why this matters
SoftBank's $9.5 billion in advance payments and reserved manufacturing capacity were supposed to be a hedge against shortages, a bet that getting ahead of the queue would pay off. Instead it shows how fast AI infrastructure plans can go stale: equipment ordered for one set of data centre specs doesn't necessarily match where the project ends up, and the gap between the two cost SoftBank billions. For founders and infrastructure teams racing to lock in GPUs, power generation gear, or cloud capacity months ahead of need, this is a concrete warning rather than an abstract risk.
Prepayment deals that look smart when demand curves are steep can turn into write-offs the moment your architecture or scale assumptions shift. The OpenAI-linked losses add another layer: even partnerships with the most hyped names in the field don't insulate capital from being misallocated. Readers building or financing data centre capacity should treat this as a data point on how brittle long-lead-time procurement bets are right now, not a one-off SoftBank stumble.
Watch whether other hyperscalers disclose similar advance-payment losses next earnings season.
Related Reading
Business & Startups Nordic pilot adds Gemini for Education, NotebookLM to boost AI literacy
Business & Startups Kling launches Video O1, all-in-one model with MVL bridge using transformer
Business & Startups DeepSeek Seeks More Capital Weeks After USD 7B Funding Round
From Archives Anthropic Hits USD 17B in Q3 as Enterprise Sales Push OpenAI to USD 10B
From Archives Coforge's Nair: No Proprietary Platform, Engineers Trained to Talk to CEOs
Common Questions Answered
Why did SoftBank write off $3.7 billion related to its US AI data centre investments?
SoftBank wrote off $3.7 billion because it prepaid $9.5 billion for equipment between 2025 and Q2 2026 to secure power generation facilities for its AI data centre buildout, but as its data centre plans shifted, much of that prepaid equipment became unused. The company's changing specifications and revised purchase agreements left it with surplus stockpiled gear that no longer matched its infrastructure needs.
What was SoftBank's original strategy with its $9.5 billion advance equipment payments?
SoftBank's strategy was to hedge against equipment scarcity by prepaying and reserving manufacturing capacity, betting that getting ahead of the queue would secure necessary power generation equipment for its AI data centre expansion. However, this bet backfired as AI infrastructure plans evolved rapidly, rendering some of the reserved equipment obsolete and causing significant financial losses.
How does SoftBank's situation illustrate the risks of AI infrastructure planning?
SoftBank's $7 billion loss demonstrates how quickly AI infrastructure requirements can become outdated, as equipment ordered for specific data centre specifications may not align with where projects ultimately end up. This shows that even massive advance investments cannot protect against the rapid pace of change in AI technology and infrastructure design, making it a cautionary tale for founders and infrastructure teams racing to lock in GPUs and power generation capacity.
What additional losses did SoftBank report beyond the equipment write-off?
Beyond the $3.7 billion write-off for unused equipment and purchase agreement changes, SoftBank also recorded losses linked to its agreement with OpenAI. The article indicates these losses contributed to the overall $7 billion hit the company took on its US AI data centre investments.
Further Reading
- Papers with Code - Latest NLP Research - Papers with Code
- Hugging Face Daily Papers - Hugging Face
- ArXiv CS.CL (Computation and Language) - ArXiv
