Wood McKenzie has released a report that US spending on power generation gear for data centers alone could hit $65 billion by 2030, more than triple the $20 billion logged last year. Data center capacity is forecast to reach 110 GW by the end of the decade, with Bloomberg also commenting that “total US spending on power-plant equipment may climb to $215 billion."

The increased spending for the heavy electrical equipment market sounds great, but unfortunately, there's no equipment to buy domestically.

Lead times for transformers, switchgear, and related gear stretch from 18-36 months and much of the shortfall is filled by imports from China, exposing the supply chain to the very geopolitical risks Washington claims to be racing against. The heavy reliance on imports for these grid-critical items led to the massive stack of Defense Production Act orders put out by the administration in April. 

Sightline Climate data highlighted earlier shows nearly half of the roughly 16 GW of US data center capacity slated to break ground in 2026 now faces delay or outright cancellation. Only about 5 GW sit under active construction.

The explosion in AI data center energy demand has been ongoing for years now. From PJM’s frantic scramble for 15 gigawatts of new supply to feed hyperscaler loads to the eye watering capacity auction price spikes that data centers helped trigger.

That squeeze is accelerating two parallel trends. 

First, hyperscalers are increasingly turning to behind the meter solutions. These include small nuclear reactors or gas fired generation directly on site. This approach allows them to bypass years-long waits for grid interconnection. Examples include Brookfield’s nuclear tied cloud venture to Nano Nuclear modular reactor studies and Talen Energy’s direct hookups.

Second, the cost pressure on households is drawing Washington’s attention. Grid upgrades required by the AI buildout have become the primary driver behind projected electricity rate increases. The Ratepayer Protection Pledge was signed back in March, which pushed hyperscalers to build, bring, or buy their own power and cover every dollar of the new transmission and distribution infrastructure. 

The White House has also framed rapid AI infrastructure buildout as a national security imperative, leading to a conflict of interests between the demand for new data centers without stressing the grid or consumers. AI has been widely labeled by the White House as necessary for the safety of the country, and is the new modern-day arms race.