AI Data Center Boom Rewires US Power Supply Chain

extract

AI Data Center Boom Rewires US Power Supply Chain Rising demand for electrical equipment drives supply growth, but hyperscale construction accelerates and infrastructure challenges persist. The US data center electrical equipment market is projected to grow to $65 billion by 2030, from approximately $20 billion in 2026, driven by rising demand for transformers, switchgear, and power distribution systems as hyperscale construction accelerates, according to a new Wood Mackenzie report. The data center sector, once a marginal slice of electrical equipment demand, is becoming a dominant buyer – large enough to reshape production capacity, pricing, and lead times across the supply chain. Ben Boucher, research director at Wood Mackenzie, told Data Center Knowledge that both capacity growth and pricing pressure are driving the expansion. US data center capacity is expected to scale from roughly 24 GW to 100 GW between 2026 and 2030, increasing competition with utilities for the same pool of electrical equipment. The report comes as hyperscalers, including Microsoft, Amazon, Google, and Meta, report surging AI demand, driving record capital expenditures and cloud growth. Recent earnings highlight the widening gap between infrastructure supply and demand, with power, cooling, and permitting emerging as key constraints. Electrical Equipment Emerges as Build Limiter The limit on infrastructure capacity is creeping into the physical layer, where constraints on power generation, distribution, and delivery are becoming increasingly apparent. Wood Mackenzie estimates data centers will account for roughly 68% of US load growth through 2030. According to the report’s authors, meeting that demand requires not just more generation, but also equipment that steps down the voltage, conditions the power, and delivers it to the rack. Transformer demand is expected to rise from about 1,500 units annually to more than 9,000 by the end of the decade. Similar growth is expected across switchgear, transfer systems, and power distribution units (PDUs). Boucher said substation transformer lead times have stretched from roughly 140 weeks in 2023 to more than 160 weeks in 2026. Switchgear timelines have held closer to one year but remain elevated relative to historical norms. That surge is colliding with long manufacturing cycles, where projects can be permitted, financed, and partially constructed, but they still stall while awaiting electrical gear. Rising rack densities toward 200 kW are forcing changes in electrical distribution and cooling design. (Image: Alamy) Hyperscale Demand Distorts Supply Wood Mackenzie projects that data centers could account for as much as 40% of total US electrical equipment demand by 2030, up from a historically low single-digit share. That shift concentrates purchasing power in a small number of hyperscale buyers. Utilities, industrial operators, and renewable projects now compete for the same pool of transformers and switchgear. Large data center developers with capital and long-term procurement strategies are securing supply earlier and in greater volume, while smaller buyers and even utilities face longer queues and less certainty. New US data center capacity now hinges on time – not just how much power is planned, but how long it takes to deliver it. For example, in PJM Interconnection territory, projects reaching commercial operation in 2025 spent an average of about eight years in the interconnection queue, based on PJM data and filings. The transmission organization has since overhauled its process and shifted to a cluster-based system intended to move projects through faster, but even under that framework, interconnection studies can take one to two years, depending on system impact. Demand in those queues is still rising: PJM has reported hundreds of gigawatts of generation interconnection requests since 2023, with much of that capacity still awaiting study. The pattern is not limited to PJM. In the Electric Reliability Council of Texas (ERCOT) territory, large-load requests tied to data centers now reach into the hundreds of gigawatts, according to planning disclosures. Utilities across regions are evaluating individual data center projects in the 100 MW to 500 MW range, with multi-phase campuses pushing into gigawatt-scale demand. Even after approvals, timelines stretch further. A Phased Fix Developers are responding by phasing projects (bringing sites online in stages) and by waiting on transmission upgrades and substation buildouts. Rob Gramlich, analyst and president of Grid Strategies, said multiple factors are converging to slow delivery. “The larger data centers require a massive amount of pieces to come together and difficult coordination of all the pieces on-site and the grid connection,” he told Data Center Knowledge. “Lack of grid capacity is a major underlying source of interconnection process delay. When capacity exists, the interconnection tends to be quick and cheap.” He added that the limited availability of electrical equipment and local siting are emerging as additional pressure points. “Large load tariffs, if they are well-designed, help to ration the system’s ability to process the applications and focus the processing and the infrastructure on the projects most ready to proceed,” he said. “We have been doing that in generation for the last five years, and it needs to happen with the load now.” Boucher said the constraint is not singular. “It’s a combination of power limitations and equipment constraints that are impacting the data center buildout,” he said. Even where generation can be deployed behind the meter to avoid interconnection queues, long lead times for turbines and electrical equipment still limit how quickly projects can move. “It’s a combination of power limitations and equipment constraints that are impacting the data center buildout." – Ben Boucher, Wood Mackenzie Design and Procurement Adjust to the Bottlenecks Operators are changing how they design and procure infrastructure. Steve Carlini, Schneider Electric’s vice president of innovation and data center, said demand is shifting toward prefabricated electrical systems due to labor shortages and rising density requirements. “We are seeing a major shift to prefabrication due to skilled labor shortages in electrical and HVAC, as well as the difficulty in designing and building at these very high densities,” he told Data Center Knowledge. Densities are climbing quickly as new accelerated compute systems are pushing rack power toward 200 kW and beyond, forcing changes in both electrical and cooling design. Carlini said operators are also moving earlier in the procurement cycle and locking in supply years ahead of deployment. “Supply capacity agreements or actual orders are placed many years in advance, mainly for critical equipment like UPS [uninterruptible power supply] and switchgear,” he said, adding that this approach helps ensure lead times and price stability. Yet, even as equipment lead times stretch, Carlini pointed to a larger limiting factor: access to power. “By far the number one constraint for data center deployments is access to prime power,” he said. In a 2025 Schneider Electric survey of 149 industry respondents, utility capacity and transmission constraints ranked as the top barrier, cited by 92% of respondents, followed by permitting, fiber access, and chip availability. Equipment shortages slow projects, but grid access determines outcomes. Manufacturing Ramps to Meet Demand Carlini pushed back on the idea that current equipment lead times will remain fixed. “The biggest misconception is that the lead times of today will not change,” he said, noting that large power systems still require custom engineering and specialized assembly. Schneider Electric is expanding domestic manufacturing and standardizing designs to improve delivery timelines, including AI-enabled manufacturing efficiencies. The company has committed more than $700 million in US investment through 2027 to scale power system production capacity. Even so, supply chains are chasing a moving target as manufacturing expands and AI-driven demand continues to climb.