AI Computing Devours Turbine Capacity, SoftBank Takes $3.7B Write-Down, Shipping & Shipbuilding Under Pressure

nashnova research
今天发布阅读约 12 分钟

The AI data-center power grab is squeezing global turbine supply chains. SoftBank booked a $3.7 billion write-down after revising procurement plans, while aviation, shipping, and natural-gas infrastructure face ballooning lead times and costs.

01

Why can a single turbine blade bottleneck an entire industry?

The critical component is the single-crystal casting — a nickel-alloy turbine blade that survives temperatures hot enough to melt steel. Roughly 80% of global capacity sits with two producers: Berkshire Hathaway's Precision Castparts and Howmet Aerospace.
This means → once AI data centers flood the queue for gas turbines, aircraft engines, LNG compressors, and marine auxiliary engines all go short simultaneously — they share the same casting suppliers.
GE announced a $11.8 billion acquisition of Consolidated Precision Products in September, one of the few alternative suppliers. Howmet CEO John Plant said AI demand "is testing us on multiple levels."
02

What does Airbus losing 30 deliveries have to do with AI?

Howmet's turbine-blade supply problems directly caused an engine shortage at Airbus in 2024, forcing the company to cut its full-year delivery target by roughly 30 aircraft.
In plain terms = planemakers and data-center builders are fighting over the same parts from the same factory — and the planes lost.
To bypass the bottleneck, AI developers are scouring retired-aircraft boneyards in the American Southwest for old jet turbines to refit for data-center power. Startup Boom Supersonic — originally planning a Mach 1.7, 80-seat supersonic airliner by 2029 — now appears to be pivoting toward data-center power generation.
03

How did SoftBank lose $3.7 billion on turbines?

SoftBank founder Masayoshi Son tried to lock in supply early at massive scale. Per the *Wall Street Journal*, SoftBank's turbine purchases carried generating capacity equivalent to six Hoover Dams.
The bet failed. SoftBank this month booked a $3.7 billion write-down on prepayments and contract amendments after revising its procurement plan. The stock is down 65% from its June high.
This reflects a hard lesson: in a hyper-tight supply market, front-loading massive commitments locks in sunk costs that turn into book losses the moment demand timing or technology direction shifts.
04

Why are shipping and natural gas caught in the same squeeze?

Turbine manufacturers are prioritizing AI clients, starving container-ship builders of auxiliary engines and delaying vessel deliveries.
The natural-gas industry relies on similar equipment for pipeline compressors and LNG units. Compressor lead times have stretched from roughly 40 weeks to about 200 weeks, forcing operators to lock in capital far earlier and pushing up financing costs.
In plain terms = one turbine production line, AI first in the queue, ships and gas fields pushed back — wait times have quintupled.
05

How much have power-plant costs risen — and who is reselling turbines?

Baseload gas-fired power equipment that already has an order slot now costs three times what it did before. Gas-plant construction in Ohio (U.S.) and Jeju Island (South Korea) has been delayed by years.
A shuttered power plant in the Philippines has resold its previously purchased turbines to supply a U.S. data center. This means → turbines are being globally reallocated, flowing to the highest AI bidder.
06

Who ultimately benefits from this buying frenzy?

There is a structural irony: Silicon Valley is indifferent to AI data-center carbon emissions, yet is objectively raising the entry cost of every new fossil-fuel power project.
This means → the relative competitiveness of renewables rises — not because wind and solar got cheaper, but because gas-fired generation got more expensive.
But the next energy project waiting in line for a turbine still has to absorb this round of price pressure first. In the short term, there are no winners.

市场有风险,内容仅供研究参考,不构成投资建议。