SoftBank Group shocked investors with a ¥1.1tn net loss for the July-to-September quarter, its first quarterly loss in nearly two years, after a hefty writedown tied to US data-centre procurement and an Ohio project. The shares still jumped 13.9% on Tuesday as the market focused less on the red ink and more on the Japanese conglomerate’s role as one of the biggest private betters on artificial intelligence. Even so, the latest results show how quickly the cost of chasing the AI build-out can turn from ambition into pain.
The loss, equal to about $7bn, came after SoftBank had posted a ¥2.5tn profit a year earlier and handily missed the profit forecast from analysts surveyed by LSEG. The group said the damage stemmed from a ¥597.2bn writedown on data-centre procurement in the US. SoftBank also said financing costs more than doubled to ¥462bn in the quarter, adding another pressure point to a balance sheet already tied closely to volatile AI investment cycles.
The biggest hit came from a writeoff connected to procurement plans that no longer fit the company’s changing needs. SoftBank said the writedown reflected “the evolving requirements of its existing and planned projects,” a phrase that captures the shifting economics of building out the physical infrastructure behind AI. CFO Yoshimitsu Goto put the point more bluntly, saying: “We recorded a loss related to non-refundable advance payments and modifications to signed contracts, which resulted in $3.7bn in losses.”
That $3.7bn figure is linked to the $7bn quarterly loss and shows how expensive it can be to pivot midstream on large infrastructure orders. Of the $9.5bn in upfront payments, $2.9bn went to equipment that was no longer needed, according to the company. For investors, the message is clear: the AI boom is not just a software story or a chips story. It is also a procurement story, where timing, contract terms and project design can hit earnings quickly.
SoftBank did not frame the writedown as a retreat from AI, but as a reset in how it wants to buy and build. Goto said, “We’ve decided to conduct a thorough reassessment” of procurement strategy. That suggests the company still wants exposure to the same long-term theme, but with less tolerance for mismatched orders and non-refundable commitments. In a market where AI spending remains a central narrative, that kind of caution can matter as much as the size of the investment itself.
The stock reaction looked disconnected from the headline loss, but that is partly the point. SoftBank shares rose 13.9% on results day even after the company reported one of its worst quarterly numbers in years. Investors appeared to focus on the scale of SoftBank’s AI portfolio and the possibility that its latest misstep was a one-time procurement problem rather than a broader collapse in strategy. The share price, however, remains more than 50% below its June record high, underscoring how much volatility has already been baked into the name.
That gap between the quarter’s damage and the stock’s move reflects how heavily SoftBank is tied to sentiment around artificial intelligence. When enthusiasm is high, investors are willing to look past losses in search of upside from future assets. When sentiment cools, the same holdings can be marked down quickly. SoftBank is especially sensitive because it has become a public proxy for AI exposure, with every earnings release offering a fresh read on whether the trade still has momentum.
A major part of that story is OpenAI. SoftBank holds about a 13% stake in the company, and its cumulative investment stands at about $64.6bn, according to the fact pack. That makes OpenAI one of the most important components of SoftBank’s broader AI thesis, but also one whose value is not always immediately reflected in reported results. Goto said revaluation of that stake would not occur until the current funding round is complete, and he said that could show up next quarter.
That timing matters because it leaves the current quarter looking worse than it may prove to be in hindsight if OpenAI’s valuation is updated later. Still, the market can only trade on what is already visible. For now, investors have to sort through a business that is taking losses on infrastructure procurement while sitting on a sizable bet that could be re-marked once a financing round closes. That combination is exactly why SoftBank remains so closely watched by AI bulls and skeptics alike.
More broadly, the OpenAI stake keeps SoftBank embedded in the center of the AI ecosystem even when its own operating numbers wobble. The company’s challenge is that its results can be distorted by accounting, timing and the pace of external funding events. That makes the stock a fast-moving instrument for anyone trying to express a view on AI, but it also means the shares can swing on details that would barely register at a more traditional industrial company.
SoftBank’s problems are not just about one bad quarter. They also show what happens when a giant investor tries to match long-term AI demand with massive physical spending before requirements are fully settled. The company’s description of “existing and planned projects” suggests it had to deal with changes after commitments were already made. In practical terms, that is where the risk lies: the world may still want more data centers, but the wrong contracts can be expensive long before those facilities generate revenue.
The financing line added another layer of caution. Reuters reported that financing costs more than doubled to ¥462bn in the quarter, showing how funding pressure can compound investment mistakes. When rates, capital structure and project costs all move against a company at once, even a sprawling group with deep assets can see its earnings swing sharply. For SoftBank, that is especially important because the company’s investment thesis depends on being able to keep funding large bets across cycles.
Investors now have two likely checkpoints to watch. First is the completion of the OpenAI funding round, which could give SoftBank a cleaner mark on one of its most important holdings. Second is the company’s next read on procurement and infrastructure spending, after Goto’s vow to reassess the strategy. The market will want to know whether this was a one-off cleanup or the start of a more disciplined phase in how SoftBank executes its AI ambitions.
There are also longer-dated catalysts in the background. SoftBank has potential plans for an SB Energy IPO, though that has been delayed, and there are Roze listing plans as well. Those moves could help surface value inside the group, but they do not change the immediate issue: SoftBank just posted a loss large enough to jolt investors, even as the stock climbed. That is the paradox of the company right now — the market is buying the AI story, but it is not ignoring the bill.