AI Infrastructure Boom Drives Industrial Leasing Beyond Data Centers
Cushman & Wakefield analyzed 388 million square feet of leases and found an overlooked winner in the AI infrastructure race. The Google News headline is not simply about bigger data centers. It concerns the factories, warehouses, contractors, and suppliers gathering around them.
Data center-related businesses represented 10.4% of new industrial leasing across six major United States markets from 2022 through 2025. Their share reached 14.4% in 2025, according to the firm. Related leasing rose 44% that year to 13.5 million square feet.
Those figures challenge the usual argument over whether a server facility creates enough permanent jobs to justify its enormous resource demands. Cushman & Wakefield shifts attention toward a surrounding industrial network that includes electrical equipment makers, cooling suppliers, engineers, and specialized contractors.
The tension is important. Industrial spillovers strengthen the economic case for AI infrastructure, but they do not settle debates over electricity, water, land, or public subsidies. The central question is whether measurable supplier growth outweighs the costs absorbed by host communities.
What Cushman & Wakefield Actually Measured
The report measures a wider physical supply chain, not employment inside server halls alone.
Cushman & Wakefield published its industrial demand research on July 20, 2026. The study covers industrial leasing between 2022 and 2025 in six leading United States data center markets.
Researchers reviewed more than 388 million square feet of leasing activity. They classified tenants according to their connection with data center construction, equipment, maintenance, and operations.
Those tenants leased about 40 million square feet during the four-year period. That total represented 10.4% of all new industrial leasing captured by the analysis.
The annual pattern matters more than the cumulative figure. Data center-related leasing reached 13.5 million square feet in 2025, an increase of 44% from the previous year.
Its share of new leasing also climbed to 14.4%. In other words, almost one of every seven leased square feet involved a business supporting the data center ecosystem.
That category extends far beyond cloud providers. It includes manufacturers of electrical equipment, cooling systems, generators, power infrastructure, and other essential components.
Engineering firms, telecommunications providers, wholesalers, logistics operators, and specialized construction companies also fall within the network. Their work happens outside the data center, but the facility creates the underlying order flow.
This distinction changes how readers should interpret the Google News story. It does not claim that server buildings suddenly employ warehouse-sized workforces after opening.
Instead, the study follows demand into separate industrial properties. A switchgear maker, cooling supplier, or electrical contractor might lease space miles from the computing campus it serves.
The report also estimates wider employment and economic effects. Cushman & Wakefield associates the 40 million leased square feet with 81,000 to 124,000 industrial and downstream jobs.
Its model estimates that each 100 megawatts of new data center development supports nearly 1,300 local jobs. It also associates that capacity with roughly $110 million in annual wages.
The same model produces estimates of $344 million in gross output and $187 million in gross regional product. These are modeled effects, not direct headcounts recorded at individual facilities.
That qualification matters because multiplier models include indirect and induced activity. Supplier employment counts as an indirect effect, while household spending supported by new wages creates induced effects.
Readers should therefore separate three layers. The report contains observed lease transactions, classified tenant relationships, and modeled economic outcomes.
The leasing figures provide the strongest evidence because they describe completed market activity. The job and output estimates depend more heavily on assumptions about supply chains and local spending.
Cushman & Wakefield’s central finding remains significant even with that distinction. AI infrastructure demand is appearing in ordinary industrial buildings that rarely feature in announcements about computing capacity.
Why AI Demand Is Escaping the Data Center
Modern AI facilities behave more like industrial projects than isolated technology buildings.
A large data center requires more than servers and a connection to the grid. It needs electrical distribution systems, cooling equipment, backup generation, networking hardware, security, and ongoing technical services.
Generative AI increases several of those requirements. Dense computing clusters consume substantial electricity and produce heat that specialized systems must remove continuously.
That creates demand for equipment before a facility opens. It also produces maintenance, replacement, logistics, and technical-service requirements throughout the building’s operating life.
The resulting supplier network needs physical space. Manufacturers require production floors, contractors store equipment, and distributors position inventory near active construction markets.
Proximity can reduce delays when a component fails or a project schedule changes. It also helps contractors coordinate workers and materials across several campuses within one region.
This mechanism explains why data center clusters matter. A single project can source equipment nationally, but a sustained regional pipeline gives suppliers more reason to establish permanent local operations.
The effect is especially visible in facilities between 30,000 and 400,000 square feet. Cushman & Wakefield identified data center support companies as a meaningful tenant group in that size range during early 2026.
The trend intersects with a broader industrial recovery. The firm reported that national industrial vacancy fell to 6.9% during the second quarter of 2026.
Net absorption, which measures occupied space gained after move-outs, reached 62.1 million square feet during that quarter. First-half net absorption totaled 113.6 million square feet.
New leasing reached 193.4 million square feet in the second quarter. That was its highest quarterly level since the middle of 2022.
Industrial buildings completed since 2020 captured 137 million square feet of first-half net absorption. Tenants favored properties with higher ceilings, operational efficiency, and stronger electrical capacity.
These national totals do not come solely from AI infrastructure. E-commerce, manufacturing, onshoring, and supply-chain redesign remain major sources of demand.
That is an important limit on the claim. A strong national warehouse market can make the data center contribution appear more consequential, even when several forces are operating together.
However, the supplier mechanism is direct enough to observe. Data centers require transformers, switchgear, cooling assemblies, fiber systems, and construction services regardless of consumer shopping patterns.
Supply limitations also amplify the effect. Manufacturers and contractors often need suitable power, loading access, zoning, and skilled labor within the same location.
An available warehouse without sufficient electricity might not meet those needs. A building with better power capacity can therefore attract a premium from automation users and data center suppliers.
Cushman & Wakefield separately modeled AI’s longer-term effect across commercial real estate. Its AI property scenarios assign the largest potential increase to industrial space.
The baseline scenario projects 298.5 million additional square feet of industrial absorption through 2035. That forecast includes automation, economic growth, and supply-chain changes beyond data center construction.
Unlike the leasing study, the 2035 figure is a scenario rather than an observed result. It assumes gradual AI adoption and wider productivity gains.
Together, the studies describe two different channels. Data center suppliers create immediate industrial demand, while broader AI adoption might reshape warehouse operations and economic activity over time.
The Google News Headline Hides the Real Contest
The central contest is broader economic spillover against concentrated local costs.
A conventional data center debate focuses on permanent employment inside the finished facility. Server campuses can involve huge investment while needing relatively few on-site workers after construction ends.
Cushman & Wakefield answers that criticism by expanding the frame. Its research counts the industrial businesses that build, equip, supply, and maintain the facilities.
That approach captures activity excluded by direct-employment comparisons. A cooling-system factory or electrical contractor might owe part of its workload to several data center projects.
The reframing has practical consequences for economic-development officials. A project’s value might depend less on its internal headcount and more on whether suppliers establish nearby operations.
However, a wider frame also creates an attribution problem. A manufacturer serving utilities, factories, and data centers cannot assign every new lease or job to AI infrastructure.
Cushman & Wakefield says researchers identified businesses tied to the data center ecosystem. Public summaries do not provide every tenant-level classification or counterfactual assumption.
A counterfactual asks what would have happened without the data center expansion. It determines whether a lease represents genuinely additional activity or demand relocated from another market.
The report establishes correlation through tenant relationships and leasing patterns. It does not prove that every classified lease would disappear without AI investment.
Timing presents another complication. Construction and equipment demand can surge while campuses are being built, then settle when the regional pipeline slows.
A supplier lease might outlast a construction cycle. Whether the business remains productive depends on repeat orders, maintenance work, and access to other customers.
The same uncertainty applies to induced employment. Workers spend wages at local stores and service businesses, but those effects depend on where workers live and purchase goods.
Still, dismissing every indirect effect would be equally misleading. Data centers cannot operate without extensive electrical, mechanical, and communications supply chains.
The evidence therefore weakens one simplistic argument. Counting only employees who enter the server building understates the industrial labor supporting the facility.
It does not validate the opposite extreme. A multiplier estimate cannot automatically justify any tax incentive, grid investment, or land-use decision attached to a project.
The strongest interpretation lies between those positions. AI infrastructure produces a broader economic footprint, but the distribution of benefits determines its local value.
Some communities might capture manufacturing, contracting, wages, and property taxes. Others might host the power-intensive campus while suppliers and workers remain elsewhere.
Distance changes the result. A facility can buy equipment from another state, employ traveling contractors, and direct profits toward companies headquartered outside the region.
Local officials therefore need project-specific evidence. They should ask which suppliers will lease locally, how many jobs will persist, and what portion of wages stays nearby.
The Google News framing also risks confusing the messenger with the event. Google News aggregated the citybiz headline, while Cushman & Wakefield produced the underlying research.
That matters for source evaluation. Search visibility does not turn an aggregated headline into independent confirmation of the report’s conclusions.
The report should be read as research from a commercial real estate services firm with direct interests in leasing and investment markets. Its transaction access can improve data quality.
Its industry position also makes methodological transparency important. Independent local data should test whether the reported spillovers appear in tax receipts, payrolls, and business formation.
Jobs Multipliers Meet Power and Subsidy Constraints
Industrial spillovers improve the jobs argument, but they do not erase the infrastructure bill.
Electricity is the most immediate constraint. Data centers require large, continuous loads, while AI computing has accelerated demand projections across several regions.
The United States Department of Energy reported that data centers consumed 176 terawatt-hours during 2023. That was roughly three times their 2014 consumption.
The department’s energy usage report projected consumption between 325 and 580 terawatt-hours by 2028. Actual demand will depend on deployments, utilization, and efficiency.
Newer estimates extend that uncertainty. The Electric Power Research Institute’s power demand scenarios place 2030 peak data center load between 45 and 94 gigawatts.
The width of that range shows why utilities face difficult investment decisions. They must plan generation and transmission before knowing which proposed projects will reach operation.
Industrial suppliers add another layer of electricity demand. Manufacturers of power and cooling equipment might need substantial capacity within the same constrained markets.
Labor can become constrained as well. Data center construction competes for electricians, engineers, welders, and specialized trades needed by factories and other infrastructure projects.
That competition can raise wages and attract training investment. It can also delay unrelated developments or increase their construction costs.
Land creates similar tradeoffs. Data centers, substations, warehouses, and manufacturing buildings all require suitably zoned sites with transport and utility access.
A supplier cluster can strengthen the local tax base. It can also place more pressure on industrial land, housing, roads, and public services.
The subsidy question is therefore unavoidable. State and local governments frequently offer sales-tax exemptions, property-tax reductions, or infrastructure support to attract data centers.
Cushman & Wakefield’s findings provide officials with a stronger possible benefit than direct facility employment alone. Yet modeled spillovers should not substitute for enforceable local commitments.
Independent research has found sharp variation among projects and facility types. A 2026 employment analysis from Brookings argued that economic outcomes depend on what a facility attracts around it.
The analysis distinguished basic colocation facilities from projects that help form larger technology clusters. It also noted growing community resistance and reconsideration of incentive programs.
That distinction aligns with the leasing report in one respect. The surrounding ecosystem matters more than a generic count of server buildings.
However, it also raises the burden of proof. Officials must show that their project will create a local ecosystem rather than merely consume local resources.
The report’s 81,000 to 124,000 job range should be understood in that context. It covers industrial and downstream effects associated with leasing across six markets and four years.
It is not a promise that each new campus creates nearly 1,300 permanent positions within its property boundaries. The model describes a regional economic chain.
Public officials should separate temporary construction work, recurring supplier employment, direct operations, and induced service jobs. Each category has a different duration and policy value.
They should also compare benefits against public exposure. Relevant costs include grid upgrades, incentive foregone revenue, road improvements, water infrastructure, and environmental mitigation.
Project agreements can reduce uncertainty by tying benefits to measurable outcomes. Governments can require reporting on payroll, local purchasing, electricity demand, and supplier locations.
Clawback provisions can recover incentives when a developer misses contractual commitments. Transparent reporting lets residents judge whether an announced supplier ecosystem actually forms.
None of these safeguards negates the leasing evidence. They convert a broad industry finding into a testable local proposition.
Who Faces Pressure as Industrial Demand Expands
The boom pressures utilities, developers, manufacturers, and local governments to coordinate decisions that previously occurred in separate markets.
Utilities face the clearest deadline. A data center cannot operate without sufficient power, and its suppliers might also seek capacity in the surrounding industrial market.
Utilities must decide which projects deserve connection commitments. They also need protections against stranded investments if speculative campuses never open.
Data center developers face pressure to demonstrate more than capital expenditure. Communities increasingly want evidence about permanent employment, electricity costs, water use, and tax revenue.
A credible supplier strategy can help. Developers can identify local purchasing opportunities, support apprenticeship programs, and coordinate space for contractors near their campuses.
Industrial landlords also face a changing tenant mix. Some data center suppliers need stronger electrical service, heavier floor capacity, outdoor storage, or specialized ventilation.
A generic warehouse might not satisfy those requirements. Owners must decide whether expensive upgrades will attract durable tenants or chase a temporary construction wave.
Manufacturers confront their own capacity problem. Demand for transformers, switchgear, cooling systems, generators, and electrical components can exceed existing production schedules.
Expanding production requires capital, workers, permits, and supplier confidence. Companies need reasonable certainty that AI infrastructure orders will survive changing financing conditions.
Specialized contractors occupy a favorable but exposed position. A deep regional pipeline can support years of work, yet cancellations can quickly reduce demand.
Communities carry the widest set of responsibilities. They control zoning and often participate in incentive negotiations, but utilities may manage power planning through separate processes.
That fragmentation can produce mismatched decisions. A municipality might approve industrial development before transmission upgrades, housing capacity, or road improvements are secured.
The leasing data suggests that planning should encompass more than the campus parcel. Supplier facilities and contractor yards can alter traffic, land use, and workforce needs across a region.
Workforce agencies also face pressure to respond. The relevant occupations extend beyond software engineering into electrical trades, equipment maintenance, construction management, and advanced manufacturing.
Training programs take time. A region that approves several projects simultaneously might import labor unless it builds local capacity before peak construction.
For technology companies, the result is a less predictable deployment environment. Computing plans now depend on permitting, utility schedules, equipment availability, and community acceptance.
AI developers often discuss infrastructure through accelerator availability or model-training capacity. The real bottleneck can appear in transformers, substations, cooling equipment, or skilled trades.
That expands the competitive field. Hyperscalers do not only compete with one another for chips and electricity.
They also compete with factories, housing projects, and other industries for construction labor and grid capacity. Their suppliers compete for suitable buildings near development corridors.
Investors should therefore avoid treating every industrial property near a proposed campus as an automatic beneficiary. Location, specifications, and project certainty determine actual demand.
A supplier needs access to customers, labor, transport, and power. Proximity without those elements provides little advantage.
Industrial demand can also move before a campus opens. Contractors and distributors often establish space during construction, creating early signals about project credibility.
Lease duration offers another clue. Short commitments might reflect temporary staging, while longer terms can indicate an enduring regional operation.
Tenant identity matters as well. A manufacturer establishing production capacity represents a different economic contribution from a contractor storing materials for one project.
The Cushman & Wakefield findings encourage a more detailed map of the AI economy. Server capacity is only the center node.
The surrounding network includes physical businesses with different labor needs, property requirements, and exposure to project cycles. Their decisions reveal where AI spending reaches the wider economy.
Three Signals Will Test the AI Leasing Story
The next phase must show that record supplier leasing becomes durable local activity rather than a temporary construction surge.
The first signal is data center-related leasing after 2025. Cushman & Wakefield recorded a 44% annual increase and a 14.4% market share during that year.
Another strong year would support the view that supplier demand is structural. A sharp decline would suggest that construction timing inflated the record.
Readers should watch both total square footage and the composition of tenants. Manufacturers, recurring service firms, and long-term logistics operations indicate greater durability than temporary staging users.
Lease terms deserve attention too. Longer agreements and facility improvements show greater commitment than short leases designed around one construction schedule.
The second signal is actual project delivery. Announced campuses create expectations, but energized capacity determines whether suppliers receive sustained orders.
Utility interconnection progress, equipment deliveries, and completed buildings provide better evidence than announced investment totals. Delays would weaken near-term industrial demand.
Power availability remains the decisive constraint. Regions that connect capacity without shifting excessive costs toward other customers will have a stronger foundation.
The third signal is transparent local economic reporting. Governments should publish direct employment, supplier payrolls, tax collections, incentives, and infrastructure obligations.
Those figures can test Cushman & Wakefield’s multiplier estimates at the market level. They can also reveal whether benefits stay near the host community.
The industrial spillover thesis gains strength if supplier hiring and business formation rise alongside computing capacity. It weakens if power use grows while local payroll remains flat.
The next few months should also reveal how landlords respond. More investment in power-ready industrial buildings would show confidence in sustained tenant requirements.
Excessive speculative construction would introduce another risk. Developers could create vacancies if proposed data centers are canceled, delayed, or moved toward better-connected regions.
Readers arriving through Google News should therefore resist the simplest interpretation. The report neither settles the economic debate nor reduces AI infrastructure to a server-building story.
It supplies a better measurement target. Track the businesses that manufacture equipment, lease nearby facilities, employ trades, and maintain the computing network.
Then compare those gains with the electricity, land, incentives, and infrastructure committed by each community. That comparison determines whether the wider boom creates durable value.
For developers and local officials, the practical task is straightforward. Publish supplier commitments, separate direct jobs from modeled effects, and report results after facilities open.
For investors, watch signed leases and energized projects instead of announcement volume. For residents, ask who receives the wages and who carries the long-term costs.
The AI infrastructure story becomes more credible when those answers appear in public data. Until then, Cushman & Wakefield has identified an important effect, not delivered a universal verdict.
