AI
SoftBank posts huge loss from power-generation equipment impairment, pressuring natural-gas AI power expansion
SoftBank reported a $7100000000 quarterly loss, more than $3000000000 of which was related to power-generation equipment. SB Energy had previously planned to spend $33000000000 building 9.2 gigawatts of natural-gas power plants and paid nonrefundable deposits for more than 13 gigawatts of turbines. The report argues that demand for powering AI data centers remains strong, but natural-gas expansion may already be showing its limits.
According to TechCrunch, SoftBank reported a $7100000000 quarterly loss this Tuesday, more than half of which came from losses related to power-generation equipment. SoftBank subsidiary SB Energy announced in February that it planned to invest $33000000000 to build 9.2 gigawatts of natural-gas power plants to serve the rapidly growing electricity demand of AI data centers.
Starting about two years ago, SoftBank reportedly paid deposits for more than 13 gigawatts of turbines, nearly 50% more than it needed. Noreva CEO Peter Gardett told TechCrunch by email that if SB Energy wanted immediate access to some turbines, it had little choice but to overbook future capacity. The current turbine order backlog extends into the early 2030s.
SoftBank had hoped to use its financial strength to lock up part of the natural-gas power-generation market and steer power-hungry AI data centers toward its supply resources. But the company paid nonrefundable deposits to turbine manufacturers and now considers many of those orders to have limited value, resulting in more than $3000000000 in losses from a mistaken bet. Gardett said the expense was costly price discovery: SoftBank not only overpaid but also overordered.
The report cites Brent Nelson, senior managing director of markets and strategy at Ascend Analytics, as saying that although natural gas is viewed by many data-center operators as a familiar and reliable energy safeguard, it still carries risks such as reduced winter supply and rationing when demand spikes; most data centers would be better served by retaining grid connections and using large batteries to shave peak load or coordinating power-use limits. The article argues that the losses are unlikely to trigger mass cancellations of natural-gas power plants and that AI power demand remains enormous, but SoftBank may have reached the limits of expansion in the natural-gas market.