1 hour ago
22
While the tech world obsesses over GPU supply chains and chip export controls, a far more mundane piece of equipment is quietly strangling the buildout of AI infrastructure worldwide. Power transformers, devices whose basic design dates back to the 1880s, have become the unexpected chokepoint in the global race to scale artificial intelligence.
Here’s the thing about power transformers: they step voltage up or down so electricity can travel long distances and then be safely delivered to buildings, factories, and yes, data centers. Without them, nothing plugs in. Nothing turns on.
Lead times for power transformers have ballooned to as long as five years, according to Financial Times reporting. Before 2020, you could expect delivery in about 24 to 30 months. Demand for these devices has surged approximately 119% since 2019. The US, which imports around 80% of its power transformers, is projected to face a 30% deficit in distribution transformers by 2025.
Nearly half of planned US data center projects slated for 2026, roughly 140 projects representing about 12 GW of capacity, are at risk of delays or outright cancellations.
Global data center capacity is projected to reach 220 GW by 2030, roughly six times the levels seen in 2020. Electricity consumption for data centers alone could land somewhere between 945 TWh and 1,400 TWh by 2030.
This isn’t just an AI story. It’s a power story, and power constraints affect every energy-intensive digital industry, crypto mining included. Bitcoin mining operations, which have increasingly competed with data centers for the same grid capacity and the same power infrastructure, face the same transformer bottleneck. Any large-scale mining facility needs step-down transformers to connect to the grid. If those transformers take five years to arrive, that’s five years before a new facility can flip the switch.
Companies that already have power infrastructure in place, whether they’re data center operators, mining firms, or hybrid operations, hold an increasingly scarce asset. This is partly why we’ve seen a wave of deals where AI companies are acquiring or partnering with existing power-connected facilities, including former and current crypto mining sites.
If 30% to 50% of planned US data center capacity faces delays, the compute supply that underpins everything from large language models to on-chain AI agents could tighten further. Investors watching the AI and digital infrastructure space should pay attention to transformer order books, utility interconnection queues, and which companies have already secured their power infrastructure.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.
English (US) ·