SoftBank’s $3 Billion Turbine Loss Puts a Limit on AI’s Gas Boom

Introduction

AI data centers have turned electricity supply into a new investment race. SoftBank tried to get ahead of that race by reserving a large share of the natural gas turbines needed to generate power on site. The strategy helped address a severe equipment shortage, but it also produced more than $3 billion in losses and exposed the limits of betting too aggressively on one energy source.

Key points

• SoftBank previously announced that its SB Energy subsidiary would spend $33 billion on a 9.2-gigawatt natural gas power plant.
• Beginning about two years ago, the company placed deposits on more than 13 gigawatts of turbines, nearly half again as much capacity as it needed.
• Turbine order queues now reach into the early 2030s, making early procurement attractive but also increasing the cost of a wrong forecast.
• SoftBank reported a quarterly loss of $7.1 billion, with more than half of it related to power-generating equipment. Much of the nonrefundable deposits it paid was subsequently considered worthless.

The price of moving first

The original bet had a clear rationale. AI data centers can wait years for a grid connection, while natural gas plants rely on a familiar generation technology and can offer a degree of control over supply. For developers and investors with large cash reserves, securing turbines early could shorten construction schedules and protect projects from a congested manufacturing pipeline.

The problem was not simply that SoftBank ordered equipment early. It ordered substantially more than its immediate requirement. As data center operators gain experience with actual operating loads, they can refine their expectations about capacity, backup power and grid needs. Equipment acquired under more aggressive assumptions may therefore be difficult to use or resell at the expected value.

The resulting charge can be read as an expensive form of price discovery. SoftBank paid to learn how much the market would charge for speed and scarcity, while also discovering that scarcity does not eliminate demand risk.

Gas remains important, but it is not the only option

SoftBank’s loss is unlikely to trigger mass cancellations of gas plants. AI’s electricity requirements remain large, and natural gas still offers a practical near-term option for projects that cannot wait for a full grid connection.

Yet gas generation carries its own risks. Fuel supplies can be constrained during winter, when demand rises and availability falls. Some data center operators may instead seek a smaller grid connection and use large batteries to reduce peak demand or ride through requests from utilities to curtail consumption.

That combination could slow the natural gas market without ending it. SoftBank’s experience suggests that AI power demand is real, but not every advance purchase will become a valuable asset. Gas may remain part of the AI energy mix; the company’s multibillion-dollar write-down may simply mark the point at which enthusiasm meets a ceiling.

Source: TechCrunch AI