The century-old device quietly choking the world’s AI ambitions

In an unexpected twist, the expansion of artificial intelligence infrastructure is being hampered not by the much-discussed shortages of semiconductors, but by a far more conventional piece of technology: power transformers. These devices, which have been in use since the 1880s, are proving to be the critical bottleneck in the global race to scale AI capabilities. As the tech industry focuses on GPU supply chains and chip export regulations, it is the humble power transformer that is quietly limiting the ability to power the next generation of AI systems.

Power transformers serve a crucial role in the electrical grid by adjusting voltage levels, allowing electricity to travel over long distances and ensuring it can be safely delivered to various end users, including data centers that host AI applications. Without these transformers, the entire electrical system falters—nothing can be powered, and consequently, no AI operations can take place.

Recent reports from the Financial Times indicate that lead times for power transformers have surged dramatically, now stretching to as long as five years. Prior to 2020, customers could expect delivery within a more manageable timeframe of 24 to 30 months. This significant increase in lead time coincides with a staggering 119% rise in demand for power transformers since 2019. The situation is particularly dire in the United States, which imports approximately 80% of its power transformers. Projections suggest that by 2025, the country could face a 30% shortfall in distribution transformers, further complicating the landscape for AI infrastructure expansion.

The implications of this transformer shortage are profound, particularly for the data center industry. Nearly half of the planned data center projects in the U.S. for 2026—around 140 projects with an estimated capacity of 12 gigawatts—are now at risk of delays or outright cancellations. This potential disruption comes at a time when global data center capacity is expected to reach 220 gigawatts by 2030, a sixfold increase from levels observed in 2020. As a result, electricity consumption for data centers alone could range between 945 terawatt-hours (TWh) and 1,400 TWh by the end of the decade.

However, the transformer bottleneck is not limited to AI. It also poses challenges for other energy-intensive digital sectors, including cryptocurrency mining. Bitcoin mining operations, which compete for the same electrical grid capacity as data centers, are similarly affected by the transformer shortage. Large-scale mining facilities require step-down transformers to connect to the grid, and with delivery times extending to five years, these operations face significant delays in becoming operational.

The scarcity of power infrastructure has led to a strategic shift among companies in the AI and crypto sectors. Organizations that already possess established power connections—whether they are data center operators, mining firms, or hybrid entities—are now holding a valuable asset. This scarcity has fueled a wave of mergers and acquisitions, as AI companies seek to acquire or partner with existing power-connected facilities, including those previously used for crypto mining.

If the projections hold true, with 30% to 50% of planned U.S. data center capacity facing potential delays, the supply of computational resources essential for everything from sophisticated language models to on-chain AI agents could tighten significantly. Investors and stakeholders in the AI and digital infrastructure sectors would be wise to monitor transformer order books, utility interconnection queues, and the status of companies that have already secured their power infrastructure. The future of AI infrastructure expansion may very well hinge on the availability of these century-old devices, underscoring the interconnectedness of energy and technology in the modern digital landscape.

Why it matters

The challenges posed by the transformer shortage highlight a critical intersection between energy supply and technological advancement. For creators and technologists, this situation serves as a reminder that infrastructure—often overlooked in discussions about AI development—plays a pivotal role in the realization of ambitious projects. As the demand for AI capabilities continues to grow, the ability to secure reliable power sources becomes increasingly vital.

Moreover, the transformer bottleneck may lead to a reevaluation of strategies within the tech industry. Companies may need to prioritize investments in existing power infrastructure or seek innovative solutions to mitigate delays. This could spur new collaborations between energy providers and tech firms, fostering a more integrated approach to infrastructure development.

Ultimately, the transformer crisis underscores the importance of a holistic view of technological progress, one that recognizes the essential role of energy systems in supporting the digital landscape. As AI continues to evolve, ensuring that the necessary power infrastructure is in place will be crucial for sustaining growth and innovation in this field.