The Infrastructure Gap Slowing Data Center Growth
TL;DR: The immediate limit on data center expansion isn’t capital or customers, it’s physical infrastructure. Long lead times for transformers, slow utility upgrades, and bespoke cooling equipment can add 2-4 years to energization schedules. Prioritize power availability, pre-order long-lead gear, and build projects in phases while you bridge shortfalls with temporary generation and storage.
When “ready to flip the switch” means waiting years
Teams can buy servers in days. Getting them power and cooling that meet reliability and regulatory requirements can take years. That mismatch, which the industry calls the “infrastructure gap, ” now shapes where and how fast cloud capacity appears on the map.
“Global electricity demand from data centres is set to more than double over the next five years, consuming as much electricity by 2030 as the whole of Japan does today.”, Fatih Birol, executive director of the IEA (IEA news release, April 2026).
That projection sits alongside sharper short-term spikes in electricity use. The IEA reported a jump in computing electricity demand in recent years (IEA, April 2026). Those increases create concentrated electrical loads that need substations, transformers, switchgear and cooling systems, each with its own long timeline.
Three choke points that stall projects
- Grid and utility upgrades. Interconnection studies, permitting, right-of-way and substation construction involve many stakeholders and can take many months to years. Developers increasingly require a firm utility in-service date before committing to land or customer contracts (Tom Harper, Gallagher, Construction Dive interview).
- Long-lead electrical gear. Transformers and generator step-up transformers are central bottlenecks. Industry reporting (IndustrialSage, May 6, 2026, citing Wood Mackenzie) shows average lead times of about 128 weeks for standard power transformers and 144 weeks for generator step-up transformers, roughly two to nearly four years depending on order and customization.
- Cooling and heat rejection. Higher rack densities for AI and large-scale training clusters raise heat loads and change cooling architecture. Hyperscale operators optimize cooling aggressively. Smaller enterprise sites can spend a much larger share of electricity on cooling. Operators are evaluating liquid cooling, but that shifts the supply-chain profile toward pumps, custom manifolds and heat exchangers, components with their own lead times.
Hard numbers that change planning
The transformer figures are not theoretical. IndustrialSage (May 6, 2026), reporting Wood Mackenzie and NERC findings, documents that lead-time pressure has been driven by surging demand and constrained material supply, notably grain-oriented electrical steel. IndustrialSage also reports transformer prices up roughly 77% since 2019 and that generator step-up transformer demand has risen dramatically since 2019.
Supply concentration amplifies risk. IndustrialSage notes roughly 80% of large transformers used in the U.S. are imported, which makes projects vulnerable to global supply disruptions. Manufacturers and investors have announced nearly $2 billion of North American transformer-capacity expansions expected online around 2028. That helps long-term, but it does not fix projects that must energize in 2026-2027.
“The data centres of 2026 and beyond are going to be two construction projects in one: Power generation and data halls.”, Ted Way, area executive vice president at Gallagher Construction Services.
How developers and operators are changing their playbook
The squeeze forces practical changes in site selection, contracting and phasing. Common adaptations include:
- Filtering prospective sites first by confirmed utility delivery timelines and resiliency options (on-site generation, substations already sited).
- Placing preliminary orders for long-lead equipment early, at permit filing or before financial close, to secure a manufacturing queue slot. That carries balance-sheet and inventory risk.
- Phasing builds so lower-density capacity can operate on interim power while full grid upgrades arrive.
- Budgeting for temporary on-site generation and battery energy storage systems (ESS) as bridge capacity, and negotiating interconnection and cost allocation with utilities early.
Dave Klusas at AWS put the practical aim plainly:
“We call it cooling, but our goal isn’t a comfortable, 68‑degree data hall. Our goal is to move just enough air through our servers to keep them from overheating.”, Dave Klusas, senior manager of facility cooling systems at Amazon Web Services (AWS cooling report).
Policy and local politics are complicating the math
States are rethinking incentives and rate treatment as they weigh grid impacts and emissions. MultiState’s 2026 legislative tracking shows multiple U.S. states debating bills that tie tax incentives or approvals to emissions, community benefits or new rate classes. Examples include Virginia, Georgia, Indiana, Arizona, Maryland, Washington and New York. That regulatory churn can change the business case for a region quickly, even when hardware is available.
Liquid cooling changes the supplier mix, and the risks
Liquid cooling lowers the electrical penalty of air cooling at very high rack densities, which makes it attractive for dense AI clusters. But it substitutes one set of constraints for another. Pumps, manifolds, leak-tolerant piping and custom heat exchangers require different vendors and lead times. A mid-project pivot to liquid cooling can create new procurement locks and schedule exposure.
Prioritized executive checklist (who owns what, and when)
- 1) Site filter and signoff (Real Estate / Head of Infrastructure), Stage: before land purchase.
Require a firm utility in-service date, or an interconnection agreement with milestones, before closing on land or announcing customer commitments.
KPI: confirmed in-service date within X months of target energization.
- 2) Pre-procure long-lead items (Procurement / Facilities), Stage: permit filing / early design.
Place preliminary orders or secure queue positions for transformers, switchgear and custom heat-rejection components. Accept short-term balance-sheet exposure rather than a 2-4 year schedule slip.
KPI: transformer lead-time in weeks secured; percentage of critical long-lead items ordered.
- 3) Bridge capacity and phase (Finance / Operations), Stage: construction start.
Budget for interim generation/ESS to reach first kW while grid upgrades complete. Design the site for phased activation so early revenue can reduce holding costs.
KPI: months of bridge power contracted; expected time to first kW on-grid vs. target.
A short, anonymized example
An operator planned a 50 MW cluster and secured tenants and finance, but the local utility could not commit to full substation delivery for 18-30 months. By pre-ordering transformers and accepting a manufacturing queue position, and by installing temporary battery and generator capacity, the project enabled a phased handover of 15-20 MW while the permanent substation was built. That added upfront cost, but it cut schedule risk compared with waiting for a spot on the OEM queue.
Key questions you’re likely asking, with short answers
-
How long are transformer lead times today?
Industry reporting (IndustrialSage, May 6, 2026, citing Wood Mackenzie) puts average lead times at roughly 128 weeks for standard power transformers and 144 weeks for generator step-up transformers, about two to nearly four years depending on customization and queue position.
-
How much have transformer costs changed?
IndustrialSage reports transformer prices up roughly 77% since 2019, driven by material shortages (notably grain-oriented electrical steel) and higher global demand (IndustrialSage, May 2026).
-
Is cooling a big slice of a data center’s power bill?
Cooling’s share depends on scale and design. Hyperscale sites that optimize cooling report much lower relative overheads than smaller enterprise facilities; rising rack density increases cooling requirements unless operators change cooling architecture (operator white papers and industry studies).
-
Will new manufacturing capacity fix the problem soon?
Nearly $2 billion of North American transformer-capacity expansion has been announced, with new capacity expected by around 2028 (IndustrialSage). That helps medium-term, but projects targeting 2026-2027 still face constrained supply.
-
How are policy changes affecting site choice?
Several states proposed or passed 2026 legislation tying incentives, permitting or rate treatment to emissions and community conditions. That shifts the economics of preferred markets and can change where operators decide to invest (MultiState, 2026 legislative tracking).
The blunt risk, and a practical opportunity
Compute demand will keep climbing, but the physical plumbing that turns racks into usable capacity moves on a different clock. That mismatch makes infrastructure availability a strategic asset. Companies that treat power, cooling and long-lead procurement as front-end strategic decisions will launch faster, avoid costly idle commitments, and protect margins.
Utilities and policymakers face a choice: clarify and accelerate upgrade lanes and capture local investment, or let constrained supply push data-center growth to regions that act faster. For executives, the action is simple: make the transformer conversation a boardroom topic today, put procurement earlier in your project plan, and accept that the next wave of competitive advantage will come from reliably getting power and cooling equipment on time.