What happened
Semafor reported that SoftBank took a $3.7 billion hit to its earnings after cancelling a glut of orders for gas-producing equipment it had ordered to power its AI hardware buildout but could not use. The report said the write-off intensified investor fears that the AI infrastructure boom, which has seen hyperscalers take on large amounts of debt to build data centers, may be losing momentum.
Semafor reported that SoftBank recorded a $3.7 billion hit to its earnings on Tuesday after cancelling a surplus of orders for gas-producing equipment that it had ordered to power its AI hardware buildout but could not use. The outlet characterized the cancellation as evidence that the company had ordered more power-generation capacity than its actual needs justified.
According to Semafor, the news intensified fears that the AI investment frenzy — during which hyperscalers took on substantial debt to finance data center construction — could be losing steam. The report also noted that Nvidia founder Jensen Huang's resignation has made investors nervous that the world's largest chipmaker could back away from its strategy of bankrolling AI hardware under a new leader. Semafor cited a fund manager telling Reuters: 'I think the fear is that the music might stop.'
The same report said the AI buildout is still producing winners. Siemens Energy expects to triple its dividend in 2027 as the US data center boom boosts demand for its gas turbines and grid equipment, with its CEO saying: 'Some of the growth rates for data centers were unrealistic, it's just that nobody wanted to hear it.' Semafor reported that gas turbine customers face multiyear waiting times, developers are increasingly hunting for second-hand equipment, and Chinese gas-turbine manufacturers hope to capitalize on high demand.
Why it matters
A $3.7 billion write-off tied directly to AI data center power equipment is one of the clearest financial signals yet that parts of the AI infrastructure spending cycle may have overshot actual demand. If large investors ordered more energy capacity than they can use, suppliers, lenders, and markets that priced in uninterrupted AI-driven growth may need to reassess. The episode also shows how the AI buildout now moves real-world industrial markets — gas turbines, grid equipment, and second-hand machinery — meaning AI demand expectations now shape capital allocation well beyond software and chips.
The write-off is a concrete, quantified signal that at least one major AI infrastructure backer over-ordered physical equipment relative to actual need, moving the debate over a possible AI spending bubble from survey-based sentiment to a reported balance-sheet loss.
The episode links AI demand directly to industrial energy markets. If AI investors collectively ordered more gas and grid capacity than they can absorb, equipment suppliers, project financiers, and utilities that expanded on the assumption of uninterrupted growth could face knock-on effects.
The contrast within the report is notable: SoftBank cancels surplus orders and takes a multi-billion-dollar loss, while Siemens Energy still expects enough data-center-driven demand to triple its dividend in 2027. That divergence suggests the market is rebalancing rather than uniformly collapsing, though the full picture is not yet clear.
The report's mention of Jensen Huang's resignation at Nvidia adds a second layer of investor uncertainty about whether the largest AI hardware financier will maintain its current strategy under new leadership. Semafor presented this as investor nervousness, not as a confirmed strategic change.
Interactive Mechanism: How It Actually Works
Explore the underlying technology behind this development interactively.
What should a useful AI forecast state?
What to watch next
Whether other AI infrastructure investors disclose similar over-ordering losses in upcoming earnings reports; whether equipment suppliers such as Siemens Energy sustain their demand outlooks as customers absorb excess capacity; and whether financing conditions for AI data center projects tighten if lenders grow more cautious.
Future earnings disclosures from SoftBank and other AI infrastructure investors for similar over-ordering charges or cancelled equipment contracts.
Whether Siemens Energy and other gas turbine and grid equipment suppliers maintain their demand and dividend forecasts, or begin reporting order cancellations from AI-linked customers.
How the second-hand gas turbine market develops, since Semafor reported developers are increasingly seeking used equipment — a possible sign of both scarcity and more cautious spending.
Whether Nvidia's next leadership clarifies its stance on financing AI hardware purchases, which Semafor identified as a source of investor anxiety.
Whether lenders to AI data center projects tighten terms, given the report's framing that hyperscalers took on large debts during the buildout.