The key facts

  • SoftBank reported a quarterly loss of $7.1 billion, more than half of which was losses on power generation equipment
  • The company reserved more than 13 GW of turbines — almost 50% more than it needed — paying non-refundable deposits
  • Write-downs exceeded $3 billion; turbine lead times stretch into the early 2030s

The biggest bet on gas for AI ends in write-downs

SoftBank reported a quarterly loss of $7.1 billion, with more than half of that amount attributable to losses related to power generation equipment. The company deemed a significant portion of the non-refundable deposits paid to turbine manufacturers to be impaired. The total size of the write-downs on the gas business exceeds $3 billion.

How SoftBank tried to 'buy up' the turbine market

In February, SoftBank announced that its subsidiary SB Energy would spend $33 billion on a 9.2 GW gas power plant. Even earlier, roughly two years before that, the company reportedly placed deposits for more than 13 GW of turbines — almost 50% more than it actually needed. The excess order helped guarantee access to equipment and speed up its own projects: as Noreva CEO Peter Gardett explained, SB Energy 'had no choice but to overbook turbines for the future' if it wanted to secure some of the equipment right away. Turbine queues in the industry stretch into the early 2030s, and Masayoshi Son expected that by deploying the capital of his investment firm he could capture part of the market: if the AI industry needed energy for data centers, it would have to come to him.

Why the bet didn't pay off

The bet turned out to be an expensive way of 'price discovery,' Gardett believes: instead of overpaying and still risking a turbine shortage, SoftBank overpaid and over-ordered. 'This write-down is exactly the price that has finally emerged,' he said. The gas market developed more slowly than the company expected: data center operators, having accumulated operational experience, better understand their real energy needs. Brent Nelson, senior managing director at Ascend Analytics, noted on the Equity podcast that natural gas, despite its reputation as a reliable option, carries risks — for example, winter supply restrictions during peak demand. In his view, for most data centers the best solution remains connecting to the grid, albeit at lower capacity than is usually requested, while shortfalls can be offset with large batteries to shave peak loads.

What this means for the market

SoftBank's write-downs are unlikely to trigger a chain reaction of mass abandonment of gas power plants: demand for electricity from AI is too great. For now, natural gas remains a sought-after source for data centers; however, as TechCrunch concludes, SoftBank may have found the ceiling of this market.