The AGI buildout is the largest infrastructure supercycle since electrification. Hundreds of billions of dollars are flowing into data centers, power generation, and the electrical grid — and every one of those projects requires physical equipment: transformers, switchgear, HVAC, copper wire, structural steel, concrete, heavy machinery, and the engineering firms to design and build them.
This report maps 18 narrow subsectors across Industrial, Construction & Engineering, identifies 62 US-listed companies, and rates each sector's exposure to AGI-driven demand. The highest-conviction plays are in electrical equipment (transformers, switchgear), HVAC/thermal management, copper wire & cable, and engineering & construction firms — all of which face a demand surge that far outstrips current supply capacity.
Power transformers step voltage up or down for efficient transmission and distribution of electricity. Large power transformers (LPTs) are custom-engineered, weigh hundreds of tons, and take 12–24 months to manufacture. They are the single biggest bottleneck in connecting new data centers and power plants to the grid.
Every GW-scale data center campus needs multiple large power transformers. Hyperscaler buildout plans imply 2–3x the historical annual transformer demand through 2029. Grid interconnection queues are dominated by AI-related loads.
Severely constrained. Lead times have stretched from 12 months to 3–4 years. Only a handful of factories globally can produce LPTs. Grain-oriented electrical steel (GOES) is itself supply-limited. New factory capacity takes 3+ years to build.
Switchgear controls, protects, and isolates electrical circuits. Medium- and low-voltage switchgear, busway, and panelboards route power from the utility feed to individual server racks inside data centers. Every megawatt of IT load requires corresponding distribution equipment.
Data centers are the most switchgear-dense buildings on the planet. AI clusters push power density higher (60–100+ kW/rack), requiring more distribution gear per square foot. Demand growth is directly proportional to data center MW under construction.
Lead times for medium-voltage switchgear have extended to 40–70 weeks. Eaton, Schneider, and ABB have announced major capacity expansions but these won't fully come online until 2027–2028.
Data centers generate immense heat. Cooling systems — including precision air handlers, chilled water plants, computer room air conditioning (CRAC), and increasingly liquid cooling — remove that heat. AI GPU clusters run far hotter than traditional servers, making advanced cooling mission-critical.
AI training clusters at 60–100+ kW per rack cannot be air-cooled effectively. Direct-to-chip liquid cooling and rear-door heat exchangers are becoming standard. Cooling is now ~40% of data center capex for AI facilities, up from ~25% for traditional ones.
Moderate constraints. Liquid cooling is a relatively new segment with few scaled suppliers. Vertiv, Schneider, and niche players are ramping capacity. Traditional HVAC is less constrained but data-center-specific configurations have longer lead times.
Copper wire and cable carry electrical power from substations through buildings to end loads. Data centers use enormous quantities of power cable (medium-voltage feeders, low-voltage branch circuits) plus structured cabling for networking. A single 100 MW data center can use 5–10 million feet of copper cable.
Data center construction is the fastest-growing end market for wire and cable. Grid upgrades, new power plants, and renewables all require vast amounts of copper conductor. AI buildout is additive to already-strong electrification trends.
Copper prices at multi-year highs. Wire and cable manufacturers are capacity-constrained with extended lead times. Copper itself faces a structural supply deficit as mine output lags demand growth.
E&C firms provide design, engineering, procurement, construction management, and commissioning services for large infrastructure projects. They are the general contractors and program managers for data centers, power plants, substations, and industrial facilities. Revenue scales directly with capital spending.
Hyperscalers and colocation providers are spending $100B+ annually on data center construction. Each GW campus is a multi-year, multi-billion-dollar construction project. Power plant construction (gas, nuclear, renewables) adds further demand. Backlogs at major E&C firms are at all-time highs.
Severe skilled labor shortage. Electricians, ironworkers, and project managers are in critically short supply. Firms cannot hire fast enough to match backlog growth. This is a multi-year constraint with no quick fix.
Industrial gas companies produce and distribute atmospheric gases (oxygen, nitrogen, argon) and process gases (hydrogen, helium, specialty gases). These are essential for semiconductor fabrication, cooling systems, fire suppression, and power generation. Contracts are long-term, take-or-pay, and often on-site.
Semiconductor fabs (for AI chips) consume vast quantities of ultra-high-purity nitrogen, argon, and specialty gases. Data centers use nitrogen for fire suppression. Hydrogen is growing for power generation. Every new fab and data center campus is a new long-term gas supply contract.
Moderate. Air separation units take 2–3 years to build. The big three (Linde, APD, APD) have strong pricing power and are disciplined on capacity. Helium remains structurally scarce. Specialty gases for advanced chip nodes are tight.
Conduit and raceways are the tubes, trays, and channels that protect and route electrical wiring through buildings and underground. Data centers are among the most conduit-intensive structures ever built, with miles of cable tray, rigid conduit, and underfloor raceways per facility.
Directly proportional to data center and power infrastructure construction volume. Higher power density means more circuits and more conduit per building. Grid infrastructure upgrades also drive demand.
PVC conduit and steel conduit producers are running near capacity. Lead times modestly extended. Less constrained than transformers or switchgear but still benefiting from pricing power in a tight market.
Industrial automation companies provide PLCs, DCS, SCADA systems, variable-frequency drives (VFDs), sensors, and building management systems that control and monitor industrial processes. In data centers, these systems manage power distribution, cooling, and building operations.
Data center building management systems (BMS) and power monitoring are a growing market. VFDs are essential for energy-efficient cooling. Factory automation demand grows as AI optimizes manufacturing. However, data center automation is a smaller share of revenue for these diversified companies.
Supply chain has normalized post-COVID. Lead times are reasonable. Not a bottleneck sector. Growth is more demand-pull than supply-constrained pricing.
Specialty fasteners and electrical/electronic connectors are the nuts, bolts, pins, and sockets that hold infrastructure together and create reliable electrical connections. They are low-cost, high-volume, mission-critical components in every electrical and structural assembly.
Data centers, power infrastructure, and server racks all consume fasteners and connectors at scale. High-power connectors for GPU racks are a growing niche. Volume tracks overall construction and manufacturing activity, which AI buildout is boosting.
Generally not constrained. High-power specialty connectors for AI server applications can be tight, but the broader fastener market is well-supplied. Pricing power is limited.
Test and measurement companies build oscilloscopes, power analyzers, network analyzers, and environmental monitoring instruments used to design, validate, and commission electronic and electrical systems. They are essential during both R&D and deployment phases of any infrastructure project.
AI chip development and validation, high-speed networking, and power integrity testing all drive demand for advanced instruments. Data center commissioning requires power quality analysis. However, these are capital goods with cyclical ordering patterns.
Not meaningfully constrained. These are high-margin, moderate-volume products. Growth is driven by technology transitions (higher data rates, new chip architectures) rather than supply tightness.
Pumps move water, coolant, and other fluids through industrial and commercial systems. In data centers, chilled water pumps circulate coolant through precision cooling systems. In power plants, pumps handle feedwater, condensate, and cooling tower circulation. They are mechanical workhorses of any large facility.
Liquid cooling adoption is driving demand for pumps in data centers — both facility-level chilled water and rack-level coolant distribution. Power plant construction (gas turbines, nuclear) also needs large-scale pump systems. Growth is real but pumps are a small share of total project cost.
Moderate. Lead times have extended for large industrial pumps. Specialty pump makers for mission-critical applications have pricing power. The broader market is not severely constrained.
Compressors are used in refrigeration cycles (chillers), compressed air systems, and gas processing. In data centers, centrifugal compressors are the heart of large chiller plants that provide cooling. In natural gas power plants, compressors handle fuel gas and process applications.
Every data center with a chilled water cooling plant needs large compressor-driven chillers. Natural gas infrastructure buildout (pipelines, LNG, gas turbine plants) drives compressor demand. The connection to AI is real but indirect through cooling and power generation.
Moderate. Large centrifugal compressors are specialty products with limited manufacturers. Lead times have extended but are manageable. Aftermarket and service revenue provides a recurring base.
Heavy equipment — excavators, bulldozers, loaders, dump trucks, and cranes — performs the earthmoving, grading, and material handling required for any large construction project. Data center campuses are massive sites (100+ acres) that require extensive site preparation, foundation work, and ongoing construction logistics.
AI data center construction is additive to an already-busy heavy construction market. However, equipment is rented or leased for the construction phase only — it's a transient demand driver per project, not a recurring revenue stream. The secular trend is real but diffuse across many equipment types.
Fleet utilization is high. Rental rates have firmed. New equipment production is near capacity at major OEMs. Used equipment values are elevated. However, this is a cyclical market with well-established supply chains.
Structural steel (beams, columns, rebar, decking) forms the skeleton of data centers, power plants, and substations. Steel is also used in transmission towers, equipment enclosures, and server racks. It is a commodity with price driven by supply/demand and trade policy.
Data center construction is a growing end market for structural steel. However, AI-related demand is a small percentage of total US steel consumption (~100M tons/year). The incremental demand is real but unlikely to move the needle on steel pricing alone.
US steel is moderately constrained, supported by tariffs that limit imports. EAF (electric arc furnace) steelmakers have been disciplined on capacity additions. But this is fundamentally a commodity market with limited pricing power beyond trade policy effects.
Concrete is the most-used construction material on Earth. Data centers require massive concrete foundations and slabs to support heavy server equipment. Power plants, substations, and transmission infrastructure all require concrete and aggregate. Cement is the binding agent; aggregates (sand, gravel, crushed stone) provide bulk.
Similar to steel: data center construction is a growing but still-small share of total US concrete/cement demand. A single large data center can use 50,000+ cubic yards of concrete. At scale, hundreds of campus builds will be material demand. But it's hard to isolate AI-specific pricing power.
Cement capacity in the US is tight, with imports filling gaps. Aggregates are inherently local (heavy, cheap, uneconomic to ship far). Permitting for new quarries is difficult. Pricing has been strong, driven by infrastructure bill spending and housing.
Uninterruptible power supplies (UPS) and backup diesel/gas generators provide power continuity during grid outages. Every data center has UPS systems (battery-backed) for instant switchover and generators for extended outages. These are essential but represent a relatively small portion of total data center cost.
Demand grows in proportion to new data center builds. However, AI-specific GPU clusters are actually less dependent on traditional UPS architectures as some hyperscalers move toward novel power topologies. The market is real but not a high-growth differentiator.
Not significantly constrained. Multiple global manufacturers compete. Lead times have normalized. Batteries (lithium-ion BESS) are plentiful and declining in cost.
Insulation materials (fiberglass, mineral wool, spray foam) and fireproofing coatings are required in all commercial and industrial construction for thermal efficiency and fire safety. Data centers require insulation for HVAC ductwork, chilled water piping, and fire-rated wall assemblies.
Data center construction drives incremental demand but insulation is a tiny fraction of total project cost. The AI-specific demand signal is weak — insulation demand tracks overall non-residential construction more broadly.
Not meaningfully constrained for data center applications. Commercial insulation is well-supplied. Pricing tracks input costs (energy, raw materials) rather than AI-specific demand.
Tower cranes, mobile cranes, and overhead cranes lift and position heavy components during construction. Data center builds, power plant construction, and transmission tower erection all require crane services. These are typically rented from specialty crane operators for the duration of a project.
Construction volume drives crane utilization and rental rates. AI data center buildout adds to demand but is not transformative for the crane market — it's one of many construction end markets. Transformer installation requires specialized heavy-haul and rigging, which is a niche.
Crane fleet utilization is healthy but not severely constrained. Operators can mobilize equipment across regions. New crane production is adequate. This is a cyclical market following construction activity broadly.