Where the Data Center Order Was Decided

Every commercial forecast in the lighting and electrical channel now carries a data center line. The line is usually large, usually round, and usually undefended. Ask who selected the product and who wrote the purchase order, and the answer is typically a project name rather than a mechanism.

The money is not in dispute. Wesco reported second-quarter 2026 data center sales of roughly $1.5 billion, up about 45 percent from a year earlier, now above $5 billion on a trailing twelve-month basis and more than 20 percent of company sales. Graybar and Rexel are running programs of their own, and Eaton and Schneider are posting record results on the same demand. The channel’s largest members are participating heavily.

This article is about how a data center actually buys, and what that process leaves for a rep agency, a mid-market manufacturer, or a regional distributor to sell.

A forecast line that cannot name who decided the order is not a pipeline entry. It is a zip code.

How a Data Center Buys

Three procurement models now govern the electrical scope of a data center build, and each one determines who the seller is before a local sales call ever happens.

• Operator-led. The owner engineers its own reference design, selects the equipment, purchases it directly as owner-furnished gear, and hands it to a contractor or integrator purely for installation. In a July 2026 review of modular construction, SemiAnalysis describes AWS engineering its own prefabricated data hall skids under a program called Project Houdini and procuring the equipment directly, with Cupertino Electric as its design partner. QTS holds roughly 7 million square feet of warehouse in Kansas so that UPS systems, switchgear, and cooling equipment can sit in inventory before a specific building needs them. Compass Datacenters runs modular medium-voltage switchgear under a Siemens agreement covering up to 1,500 units over five years. None of these is a specification in the channel’s sense of the word — they are portfolio-level supply agreements between an owner and a factory.
• EPC-led or integrator-led. The owner sets performance requirements, and a contractor or modular integrator sources the equipment, assembles it, tests it, and delivers it as finished construction scope. This is where the large distributors earn much of their volume, supplying kitted material and jobsite inventory on compressed schedules. The product decision, however, was typically made upstream, in a basis of design the contractor executes rather than shapes.
• OEM-led. A vendor packages its own power, cooling, and white-space infrastructure into a repeatable module or full platform and sells the stack as a single product. SemiAnalysis notes that the large OEMs are now allocating factory capacity selectively, favoring bigger projects and setting capacity minimums. Here the manufacturer is not looking for a rep to find the customer — the customer is standing at the factory door.

None of those three models describes the lighting package, and a lighting principal is entitled to ask what any of it has to do with a business that never expected the switchgear order. The answer is the specification. Large operators repeat a validated design across sites; SemiAnalysis cites EdgeConneX estimating that a common design can reach a 30 to 60 percent permit set before site-specific work begins. It follows, and this is reasoning rather than a measured finding, that the lighting and controls schedule in that design was written once, to the owner’s standards, by an engineer of record who may sit in a different region from the campus, and then carried into each project’s basis of design with limited local change, mostly for energy code and site conditions. The local agency is not losing the switchgear. It is defending a lighting spec it did not write, against a buyout that runs through a national distributor agreement, in a territory the spec never visited.

Why This Is a National Account

MAG’s earlier piece, “Channel Conflict: When You Sell Direct and Through Reps,” proposed a five-part test for whether a deal is a real national account: an identified customer, a documented lighting or controls schedule, a centralized purchase, a standardized price, and a defined build schedule across locations on a known or estimated timeline. The part that does the most work is the centralized purchase, and the test is indifferent to whether the paper runs through a distributor. It measures where the decision sits.

The first four parts are easy to check against a hyperscale program: the customer is named, the schedule is the reference design, the purchase is centralized with the operator or its EPC, and the price is set at the portfolio level. The fifth part is the one worth pausing on, because it is the part that separates a national account from a large project. Its archetype is the multi-location rollout, the distribution center program or the retail chain, where the same package lands in territory after territory on a cadence. A single campus does not pass it. A hyperscaler repeating one validated design across dozens of campuses passes it more completely than any warehouse program ever has, and a wholesale colo running a standard kit across its portfolio passes it too. A one-off enterprise facility or a build-to-suit for a single tenant does not, and that is not a flaw in the test — it is the reason those jobs sit in a different tier below. The program buyer is a national account by the channel’s own definition, and the channel is discovering what that means for commission treatment after the fact.

A hyperscaler repeating one design across dozens of campuses passes the national account test more completely than any warehouse program ever has. The commission conversation should have happened before the first campus, not after the fortieth.

The mechanism is already in the trade record. At the 2026 NEMRA annual conference, rep agencies pressed for better point-of-sale reporting because product increasingly ships across rep territories as distributor consolidation extends the reach of regional and central distribution centers, opening gaps between where influence is applied and where commission is tracked. Data center campuses, supplied under a national agreement and delivered by a national distributor, are that gap at scale.

The Channel’s Actual Share

Two figures circulate on how much of this spend the channel sees, and they do not agree. Wood Mackenzie, in an April 2026 supply-chain analysis, projects the U.S. data center electrical equipment market growing from roughly $20 billion in 2025 to $65 billion by 2030, measured at the OEM level, a figure DISC Corp.’s research has since carried into the electrical channel via Electrical Trends. A Brightlio estimate, cited by the sister publication HVACR Trends, puts 55 to 60 percent of U.S. data center market revenue as flowing through distribution. The two are not measuring the same thing; the likely explanation is different denominators: one is the electrical equipment package and the other is facility-wide spend, and neither source publishes its underlying method. Read against the three procurement models above, the more defensible inference is that the equipment bought under the operator-led and OEM-led models bypasses distribution by construction, and that the share which does flow through distribution is concentrated in the EPC-led model and in a small number of distributors with the balance sheet to carry it. For lighting and controls specifically, typically contractor-furnished and a small fraction of the electrical package, no credible share figure exists, and none appears here.

What can be said with confidence is structural. The large distributors are winning on scale, credit capacity, and logistics, and that is a capability, not a consolation prize. Carrying packages of that scale on terms, staging them in a jobsite yard, and delivering against jobsite schedules is work few distributors can execute, which is why it concentrates. Regional distributors earn the discretionary business around it: temporary power, conduit and fittings, consumables, the material the contractor did not forecast. Both are fulfillment. Fulfillment is legitimate revenue; price selling time accordingly and report it to ownership as what it is.

The Rep That Is Winning

Independent reps are not absent from this market; some are building a place in it. Nexus Power, ABB’s exclusive representative across eleven western and midwestern states, took a strategic investment from Madison River Capital in July 2026. The company describes two segments: a commissions business under exclusive territory agreements and a solutions business coordinating gear, engineering partners, and delivery logistics, and reports more than 20 experienced engineers on staff, with electrical design folded into the sales process.

That agency is a distinction, not a contradiction. It sells into the design and procurement layer of the operator or EPC, on switchgear, UPS, busway, and power infrastructure, while the reference design is still being written. A lighting or controls agency calling on the contractor’s project manager after the basis of design is issued is in a different business at a different stage. The variables are which reps, which products, which buyer, and when.

Reps are winning in data centers. They are winning at the design table, on power infrastructure, before the spec freezes. That is a different business from the one most agencies run.

A Framework: Where the Order Was Decided

For every data center line on a forecast, ask one question: who selected the product, and who wrote the purchase order. The answer lands in one of three tiers.

The tiers track owner scale. Hyperscale campuses and large wholesale colocation sit in tiers one and two almost entirely; SemiAnalysis notes that a wholesale colo chooses its equipment early and without waiting on a tenant, which is operator-led buying by another name. Below the roughly 40 megawatt floor that Data Center Dynamics describes as the consensus threshold for hyperscale, the picture changes. Enterprise facilities, edge deployments, and build-to-suit tenants are the operators SemiAnalysis describes as increasingly working through system integrators because the large OEMs’ capacity minimums exclude them. For a lighting or controls agency, that segment is where tier-three work exists in quantity and where a local specifier relationship typically still moves the schedule. The framework is therefore also a targeting tool: pursue tier three with the agency’s existing model, pursue tier two with open eyes, and pursue tier one only as an agency willing to become structurally different, with engineering on staff and a place at the design table, by deliberate choice rather than as a hope attached to a forecast.

For a distributor, the same three tiers separate what “When Demand Creation Becomes Order Fulfillment” warned against from what it did not. Fulfillment at scale should not be reported to ownership as demand creation.

For a mid-market manufacturer, the framework says something the sales organization may not want to hear. Tier-one and tier-two volume is won at the owner’s standards group and with the engineer of record, before any campus is sited, and it is won by getting onto a reference design that will repeat. That is not a territory sale, not a job the rep contract was written to cover, and typically not what the regional sales manager was hired to do. A manufacturer that wants that volume needs a national account function with engineering behind it, and a decision made before the pursuit begins about what the agencies in the landing territories will be paid, and for what.

The diagnostic is mechanical. A forecast line that cannot be assigned a tier is a line no one in the organization understands. The question worth bringing to the next pipeline review: of the data center revenue claimed last year, how much did the organization actually cause? The number will be smaller than the forecast line. It will also be the only number on the page that means the same thing in every room.