SoftBank’s $7.1 Billion Loss May Slam the Brakes on AI’s Natural Gas Ambitions

Nathan Brooks ·

SoftBank has a habit of arriving at a booming investment trend with enormous sums ready to deploy. In early 2025, it placed a $30 billion bet on OpenAI, then announced in February that its SB Energy subsidiary would spend $33 billion on a 9.2 gigawatt natural gas power plant.

That appetite for scale carried directly into equipment procurement. Starting about two years ago, SoftBank reportedly paid deposits on more than 13 gigawatts of turbines, almost 50% more capacity than it needed.

The oversized orders appear to have helped the company secure scarce turbines and move its projects forward more quickly. Yet the same aggressive purchasing strategy also exposed SoftBank to a costly reversal when its projected equipment requirements changed.

“SB Energy did not have a choice except to overbook turbines in the future if they wanted some turbines right away,” Peter Gardett, CEO of Noreva, told TechCrunch by email.

The energy sector had been caught off guard by the speed and scale of rising electricity demand from artificial intelligence. Turbine order queues now stretch into the early 2030s, complicating efforts by data center developers to secure new generation capacity.

As natural gas emerged as a favored source of electricity for the AI expansion, SoftBank CEO Masayoshi Son placed a sweeping bet on the opportunity. By deploying the investment firm’s substantial resources, he sought to control a portion of the market serving data centers hungry for power.

That strategy now appears to have fizzled. SoftBank announced a $7.1 billion quarterly loss on Tuesday, with more than half of the total stemming from losses connected to power generating equipment.

The investor had made nonrefundable deposits with turbine manufacturers to secure its place in crowded order books. SoftBank subsequently determined that much of that spending was worthless, turning its attempt to lock down capacity into a multibillion dollar setback.

Gardett described the result as an expensive exercise in determining the actual price of securing turbines. Instead of risking shortages after paying elevated prices, SoftBank both overpaid and ordered more equipment than it ultimately required.

“This charge is just that price discovery finally happening,” Gardett said. The misplaced wager exceeded $3 billion, reflecting the intensity of the rush to supply electricity for AI infrastructure.

Years long waits for grid connections have pushed many hyperscalers and developers toward natural gas. The fuel has acted as a form of energy security for companies seeking a familiar and proven source while confronting delays in obtaining grid access.

However, operators are gaining a clearer picture of what AI data centers require as the facilities spend more time in operation. Natural gas carries risks, including winter curtailments when available supplies decline and demand increases.

Brent Nelson, senior managing director of markets and strategy at Ascend Analytics, discussed those risks on the Equity podcast Wednesday. He said the grid may still represent the best option for most data centers, though operators could seek smaller connections than they would ordinarily request.

Large batteries could then cover the remaining shortfall at a cost effective price, according to Nelson. The systems could reduce peak loads or help operators continue functioning when utilities request reductions in electricity consumption.

Data center operators may be reaching a similar conclusion, causing the natural gas market to expand more slowly than SoftBank anticipated. That shift would help explain why the company’s vast turbine reservations became a liability rather than the cornerstone of its power strategy.

SoftBank’s natural gas related losses are not expected to trigger mass cancellations of gas power plants. AI electricity demand still appears too large for such a domino effect, leaving natural gas in a strong position even as SoftBank’s costly retreat suggests the market may have found a ceiling.