SoftBank Takes $3.7 Billion Data Center Write-Down
The charge hits a subsidiary building power infrastructure for U.S. AI data center projects, prompting a revised turbine procurement strategy.
Updated on Nov. 10, 2026 in Data Centers
SoftBank reported a $3.7 billion write-down on its AI data center infrastructure subsidiary in second-quarter earnings. The subsidiary, Silver Bands 7, has revised how it buys gas turbines for U.S. AI data center projects.
Why it matters
The write-down shows the financial strain of securing power capacity for AI data centers, where companies have made large advance and non-refundable payments. It also signals that procurement plans for U.S. AI infrastructure are being reworked.
SoftBank disclosed $9.5 billion in advance payments for power projects, of which $2.85 billion was lost as non-refundable capacity payments. Silver Bands 7 is entitled to rebates of up to $2.1 billion over the next 10 years, though the amount on future equipment purchases remains undetermined.
The players
SoftBank
SoftBank is a Japanese technology investment conglomerate and a major investor in OpenAI, which has pursued large-scale AI infrastructure investments.
GE Vernova
GE Vernova is a power generation equipment maker spun off from General Electric, with a turbine plant in Greenville, South Carolina.
OpenAI
OpenAI is an artificial intelligence research company whose growth has driven demand for large data center and power infrastructure investment.
The details
SoftBank said its subsidiary is not ordering as many turbines as it previously planned, and GE Vernova has been named as turbine supplier for the Pike County, Ohio project. The company said it had made non-refundable advance payments to secure power generation capacity, with the losses recorded in the quarter.
Timeline
Tuesday: SoftBank disclosed the write-down and procurement revision in its second-quarter earnings.
Last week: A Musk social-media post claimed SoftBank defaulted on a turbine order.
Next 10 years: Silver Bands 7 may receive rebates of up to $2.1 billion on future equipment purchases.
The Tech Race
The write-down follows a wave of advance commitments for gas turbines and grid capacity as AI developers race to power U.S. data centers. It suggests that the scramble to lock up power equipment is now colliding with more cautious spending plans.
Electricity costs and grid capacity in areas hosting AI data centers could be affected as developers reshape their power procurement. Residents near projects such as the one in Pike County, Ohio, may see effects on local power infrastructure as those plans change.
The takeaway
Large AI infrastructure commitments carry real financial risk when power equipment orders change. Readers tracking data center growth should watch how companies adjust turbine and grid purchasing as the market matures.
Further reading
For more on the infrastructure behind these projects, see Data Centers coverage.
