AI
SoftBank's $3B turbine bet unravels as AI power strategy shifts away from natural gas
A massive over-order of turbines and the emergence of alternative power strategies has left SoftBank nursing substantial losses on its natural gas ambitions for AI data centers.
TechCrunch — AI
SoftBank's aggressive strategy to dominate the energy market for artificial intelligence data centers has backfired spectacularly. The investment giant took a $7.1 billion quarterly loss this week, with more than half attributed to losses on power-generating equipment—a humbling reversal for a company known for making outsized bets.
The Turbine Over-Order Gambit#
Two years ago, SoftBank made an aggressive move to secure natural gas infrastructure for the AI boom. The company's subsidiary SB Energy announced a $33 billion commitment to build a 9.2-gigawatt natural gas power plant and placed deposits on more than 13 gigawatts of turbines—nearly 50% more capacity than its immediate needs.
The logic was sound at the time. With turbine production queues stretching into the early 2030s, SoftBank reasoned that oversizing its orders would guarantee supply and accelerate its projects. As Peter Gardett, CEO of energy analytics firm Noreva, explained, "SB Energy did not have a choice except to overbook turbines in the future if they wanted some turbines right away."
The strategy was also a calculated market play: if AI developers desperately needed power, they would have to work through SoftBank's infrastructure empire. CEO Masayoshi Son had effectively bet his company's war chest on becoming the natural gas gatekeeper for the AI era.
When the Market Shifted#
That bet has now proven catastrophically mistaken. The over-$3 billion in write-downs on worthless turbine deposits signal that the AI industry's power strategy is diversifying faster than SoftBank anticipated. Data center operators are increasingly reconsidering natural gas as their sole solution, with industry experts now advocating hybrid approaches combining grid connections, large battery storage systems, and smaller gas capacity to manage peak demand and handle utility curtailments.
As operators gain real-world experience running AI data centers, they are discovering that grid connections paired with battery backup may offer better reliability and cost-effectiveness than pure natural gas dependency. Winter supply constraints and demand spikes have also raised questions about natural gas's stability.
Lasting Impact Limited#
The losses represent what Gardett called "price discovery"—the market finally settling on realistic valuations after the electricity mania of recent years. While substantial, SoftBank's turbine write-downs are unlikely to cascade into widespread cancellations of natural gas infrastructure. The electricity demands of AI remain enormous, meaning natural gas will continue to play a meaningful role in the power mix.

