SoftBank Lost Over $3 Billion Pre-Ordering Gas Turbines for AI Data Centers

SoftBank lost more than $3 billion while pre-ordering natural gas turbines meant to power AI data centers, according to a TechCrunch report published on November 12. The story frames the loss as something that may have put a cap on the industry's natural gas ambitions, and asks what it implies for power markets.

The central figure is the size of the loss: over $3 billion. That money did not go into a completed data center campus; it was tied up in generation equipment orders that did not convert, which makes this a capital-allocation problem rather than a rise in operating costs.

Power demand from AI data centers has been climbing quickly, and gas-fired generation is often treated as the pragmatic bridge because plants can be built and interconnected faster than most alternatives. SoftBank's pre-orders were an extreme version of that logic: lock in turbine capacity early, before competitors do.

The risk in pre-ordering is structural. Once capacity is reserved, a delayed, scaled-back or cancelled project turns the commitment into a sunk cost that is hard to recover, especially when demand forecasts, financing schedules and grid approvals move out of sync.

A $3 billion-plus hit says something about the scale of SoftBank's wager on compute infrastructure, and about how expensive a single misjudgement can be. For a company whose core skill is managing a portfolio, hardware-side exposure of this kind is not easily hedged.

The episode is also a cautionary data point for the sector. Treating power as a resource that can be stockpiled in advance runs into three sources of uncertainty at once: equipment lead times, permitting, and demand that can shift before a plant is energised. Many hyperscalers prefer long-term power purchase agreements and diversified generation rather than a single-fuel bet.

What to watch next: whether SoftBank changes its energy procurement approach, whether the affected data center timelines slip, and whether power markets reprice the so-called AI-driven gas demand story. Those three signals will decide whether this is an isolated loss or a turning point.