US power companies scramble to secure equipment as surging data center demand strains supplies

AI-driven data center buildouts are intensifying U.S. transformer and related grid-equipment shortages, extending lead times for some high-voltage transformers to multiple years (vs ~1 year in 2020-21). Equipment price inflation is also rising, with transformer costs projected to increase ~4% to as much as 10% over the next year, contributing to grid-connection delays and higher project costs. Utilities and developers are responding by buying 3–5 years ahead and shifting sourcing offshore (e.g., ~3/4 of bids coming from China and South Korea for one utility), while regulators push new protocols to speed data center connectivity.

Analysis

This is more a bottleneck story than a straight AI-demand story. The scarce asset is now grid interconnect hardware and installation capacity, which tends to move pricing power upstream to vendors with existing backlog and proven delivery, while pushing working-capital strain and schedule risk onto late-ordering developers and smaller utilities that cannot pre-buy. In that setup, the first derivative winners are electrical equipment suppliers and EPC contractors, not the end-users of the equipment.

A more interesting second-order effect is within data centers: scarcity of new powered sites should widen the moat for incumbent operators with existing utility access and spare capacity. That argues for relative strength in established colocators such as EQIX over speculative new-build platforms, because the market may re-rate “available megawatts” as the real constraint, not just land or compute demand. On the utility side, regulated names with large capex pipelines can pass through costs over time, but smaller co-ops face the most timing pressure and the highest risk of project deferral.

The key risk to the thesis is policy relief or supply normalization. If regulators speed interconnection and offshore suppliers keep taking share, lead times could flatten faster than expected, capping the pricing tailwind for U.S. manufacturers. Longer term, a powerful contrarian is that extreme grid friction may itself slow AI capacity additions and shift demand toward more efficient models, on-site generation, or regions with surplus power, which would dilute the “everything AI” trade and favor only the best-capitalized incumbents.

Key Decisions for Investors

• Long ETN / GEV / POWL on any pullback over the next 1-3 months; thesis is backlog durability and pricing power as lead times extend, but trim if company commentary shows orders being pushed out rather than repriced.
• Relative long EQIX versus a basket of new-build data-center developers over the next 6-12 months; scarcity of energized sites should favor incumbents with existing power access and leasing inventory.
• Avoid chasing small utility-dependent infrastructure names that need near-term transformer deliveries; the catalyst path is likely 2-6 quarters of delay risk before any revenue shows up, which can compress valuation in the meantime.
• Set an alert on transformer/circuit-breaker lead times and any regulatory move to fast-track interconnections; a meaningful reduction there would be the cleanest falsifier for the equipment-scarcity trade.