Insights into the forces shaping our industry.

The fastest way for switchgear and transformer manufacturers to ride the data-center boom is to accept a simple premise: power, not floor space or fiber, is the new gating factor. Hyperscale and AI facilities are running into grid constraints at the exact moment their rack densities are exploding, and the companies that win will be those that can deliver megawatt-scale capacity quickly, safely, and with architectures tuned for liquid-cooled, GPU-heavy loads. Recent market signals have been unambiguous. The International Energy Agency now expects global data-center electricity use to more than double by 2030, rising from roughly 415 TWh in 2024 to about 945 TWh, with the United States and China accounting for the bulk of that growth. Demand growth is being driven disproportionately by AI clusters, even as average PUE improvements stall and power availability becomes the operator’s top worry. For manufacturers, that means the bottleneck is upstream power equipment:  substations, MV switchgear, and transformers; and the opportunity is to remove those bottlenecks with product and delivery models purpose-built for AI deployments.

Densities at the IT layer set the tone for the entire electrical design. Where the median rack in legacy enterprise facilities still hovers below 8 kW, AI racks in production are already an order of magnitude higher. A single liquid-cooled, GB200-class rack can demand on the order of 100–130 kW of conditioned power with in-rack CDUs, and early field deployments confirm that these are no longer lab curiosities but real, shipping systems designed to scale. In parallel, major operators and vendors describe 25–100 kW per rack as increasingly common for AI training and high-end inference. Manufacturers that harden their switchgear, busway, and transformer portfolios for these load envelopes:  thermal performance, protection coordination, fault current management, and harmonics will be aligned with the physics of the next decade.

Cooling has quietly become a power-distribution problem. ASHRAE’s 2024 liquid-cooling guidance, along with a flurry of operator announcements around cold-plate and immersion systems, reflects a shift where heat rejection and electrical topologies are tightly coupled. Pumps, CDUs, rear-door heat exchangers, and facility water loops introduce step-changes in auxiliary loads and transient behavior that affect protective device settings, transformer thermal cycles, and generator sizing. Some cloud providers are developing proprietary liquid-cooling modules precisely to retrofit higher-density gear without re-laying entire plants. Switchgear and transformer suppliers who co-engineer with cooling vendors, matching fault-ride-through to pump UPS, validating short-circuit withstand for CDU motor starts, and publishing pre-approved coordination tables, will shorten design cycles and win specifications.

At the medium-voltage layer, architecture is shifting upward. AI campuses that once took dual 115 kV feeds are now planning on-site 230–500 kV interconnections with dedicated utility substations because distribution-level capacity is tapped out in many metros. Concrete site plans for new hyperscale campuses in the U.S. show 26-acre substations stepping from 500/230 kV into on-campus MV rings; the gear closest to the fence must be compact, arc-resistant, serviceable, and fast to commission. Gas-insulated MV switchgear (5–38 kV) is gaining because it squeezes more fault interruption into smaller footprints with lower maintenance, while modern arc-resistant AIS remains essential for cost-sensitive rooms and e-houses. Building product families that cover both, with consistent relaying, IEC/ANSI variants, Type 2B/2C arc-resistance options, and plug-and-play tie-ins for differential protection, gives specifiers freedom at the campus scale.

Transformers are the new long-lead critical path. In the United States, average lead times for large units have stretched from under a year in 2021 to well over two years, with some producers quoting five years for the heaviest classes. Investment is finally flowing. Major OEMs have announced multi-hundred-million-dollar expansions for three-phase and data-center-class units, but the supply imbalance persists. Manufacturers that add modular capacity, qualify multiple core materials (grain-oriented electrical steel and amorphous) to navigate policy and commodity shocks, and pre-engineer data-center-specific designs will capture outsized share. On the policy front, U.S. DOE has finalized efficiency standards for distribution transformers and Congress has taken up further measures; design teams that pre-bake compliance into catalogs, rather than treat it as a custom option, will speed procurement.

Inside the white space, the move to higher utilization voltages and fewer conversions continues. North American data centers that historically ran 480/277 V down to 208/120 V are adopting 415/240 V three-phase distribution at the row to reduce conversion losses and increase PDU capacity, while OCP’s ORv3 standardizes 48 Vdc power shelves and busbars at the rack. For LV switchgear and dry-type transformer manufacturers, this means offering proven 415 V architectures (UL-listed where relevant), harmonic-mitigating transformer options, and busway/PDU systems with high-current, low-impedance paths and monitoring designed for mixed AC/DC topologies. It also means publishing application notes that translate legacy 480 V selectivity and arc-flash practices to denser 415 V layouts with DC components.

Harmonics engineering is back on the critical-path checklist. Where legacy UPS front-ends and server PSUs once made multi-pulse rectification and K-rated transformers table stakes, the rise of PFC front-ends and IGBT rectifiers trimmed the problem, until vast, synchronized GPU loads and liquid-cooling auxiliaries reintroduced non-trivial harmonic content and inrush behavior at scale. Forward-leaning catalogs should include H-mitigating transformer (HMT) options, 12- to 24-pulse UPS transformer sets for brownfield retrofits, and clear IEEE-519 guidance backed by measured data. Publishing short-circuit and THD case studies for 30–130 kW racks—tied to real-world UPS and server power-supply behaviors—will ease approvals from AHJs and utilities.

Safety margins are narrowing as fault duties climb with denser busways and shorter cable runs. That argues for deeper investment in arc-flash engineering: Type 2B/2C arc-resistant metal-clad designs with tested pressure-relief paths; fast bus differential and zone-selective interlocking that coordinate with solid-state trip units; and enclosure systems validated for the blast dynamics of modern rooms, including e-house configurations. Vendors that deliver fully coordinated studies and breaker settings as standard deliverables, rather than post-award change orders, will save months for hyperscalers.

Speed to power is the differentiator, which is why prefabrication has shifted from “interesting” to “inevitable.” Factory-assembled, pre-tested electrical rooms, e-houses, skids, and modular substations, compress schedules, de-risk weather, and enable serial production of complex MV/LV lineups. For manufacturers, the winning move is vertical integration: bring switchgear, transformers, protection, monitoring, HVAC, and fire systems into a single prefabricated package with standardized footprints, lifting points, and interconnects. Offer variants sized for 15 kV, 25/27.6 kV, and 34.5 kV rings; ship with fiber and controls stub-outs for microgrid controllers; and certify to the same arc-resistant performance as stick-built rooms. E-house lines tailored for data centers are already positioned that way, leaning into that with AI-specific reference designs will accelerate adoption.

Microgrids and on-site generation are moving from marketing slides to procurement queues. With interconnection timelines stretching, operators are contracting gas-turbine “power foundries,” reciprocating-engine farms, and large battery systems to firm their loads and hedge price risk.  Public announcements point to multi-gigawatt buildouts using modern 7HA-class turbines as well as rapidly deployable LM-series units, and specialist developers are pitching turnkey natural-gas microgrids that parallel the utility. For switchgear and transformer suppliers, this unlocks a parallel product track: generator-paralleling gear with fast synchronizing, black-start compatible MV switchrooms, step-up transformers with generator-friendly impedances, and controls that coordinate with UPS ride-through to avoid nuisance trips. Packaging those as modular blocks that slot beside utility lineups gives campuses optionality as grid timelines shift.

Supply-chain strategy is product strategy. Copper, electrical steel, aluminum, and SF6 alternatives are all under cost and regulatory pressure. SF6-free MV switchgear based on clean air and vacuum interrupters has moved from pilot to production, with large colocation providers partnering to deploy at scale; publishing full type-test dossiers and lifecycle CO₂ models will help owners satisfy sustainability reporting that now spans embodied carbon as well as operations. On the transformer side, dual-winding designs that can swap between aluminum and copper without footprint changes, plus options for amorphous cores, give buyers flexibility as policy and commodity belts tighten. Investment announcements to expand three-phase transformer capacity in North America show where the demand is:  data-center-grade units are the growth SKU.

Digital is no longer a bolt-on. Every MV lineup and transformer shipped into an AI campus should leave the factory with condition and performance telemetry: dissolved-gas analysis for oil units or hot-spot modeling for dry-type; bushing monitors; partial-discharge sensing on MV bus; breaker health counters; and IEC 61850 gateways that drop into the site’s data lake. The Uptime Institute has tracked a stubbornly flat curve on outages and PUE even as complexity rises, and operators increasingly accept AI for narrow operational analytics. If manufacturers expose clean, standardized data models and pre-built dashboard templates for breaker operations, transformer thermal margins, relay event logs, and harmonic alarms, they will become reliability partners rather than commodity vendors.

Finally, think like a campus integrator, not just a component supplier. Publish end-to-end reference designs that begin at the utility bay (230/115 kV), step down through 34.5 kV ring bus GIS, land at 13.8 kV/LV e-houses, and terminate at 415 V row PDUs and ORv3 racks, with options for N, N+1, and 2N topologies. Include short-circuit studies for 30–130 kW racks, arc-flash labels, relay coordination, harmonic compliance roadmaps, and liquid-cooling auxiliary power graphs. Create a bill of materials that is 80% fixed and 20% selectable (cooling topology, UPS class, microgrid intertie) so they can stamp out copy-exact blocks. When hyperscalers publish plans for campuses anchored by purpose-built substations, the suppliers who can match that speed and scale, down to pre-negotiated lead-time bands on transformers and tested MV lineups, will be on the short list every time.

If there’s a single lesson in the current build cycle, it’s that data-center power is becoming a product in its own right. The sector’s most aggressive buyers are even negotiating future contracts for clean supply that borders on the speculative, from fusion pilots to next-generation nuclear and long-duration storage. Moves that underline just how valuable assured megawatts have become. In that context, switchgear and transformer manufacturers who pair credible capacity expansions with architectures tuned for liquid-cooled AI loads, SF6-free sustainability, harmonics-aware LV distribution, and prefab speed will not merely sell into the boom; they’ll define it.

At Egret Consulting, we understand that the data center boom represents not just a surge in demand, but a fundamental transformation in how power distribution must be designed, engineered, and delivered. Egret has proven expertise working with the manufacturers and innovators who build the switchgear, transformers, and power systems that are now mission-critical to AI and hyperscale operations.

As operators race to meet unprecedented growth, having the right talent in place: engineers, product managers, technical sales, and leadership, will define who captures this market. Egret Consulting specializes in building these high-performance teams, ensuring you have the expertise and capacity to innovate at the pace the data center industry demands.

If you’re ready to strengthen your organization to meet the immense opportunities ahead, partner with Egret Consulting today and let us help you power the future of data centers.

Rob Wieska – Executive Recruiter / EVP
Power Distribution | Automation & Renewables Technologies