Primary vs. Secondary Data Center Market Pricing Dynamics

Atlanta, a market long treated as secondary to Northern Virginia, overtook it in H1 2026 as the top U.S. market for total construction, with under-construction capacity up 52.3% year over year, CBRE reports. That single fact breaks the old pricing model. For years, buyers and developers assumed primary markets like Northern Virginia, Dallas, Chicago, and Phoenix would always command the higher rent, thanks to proximity to customers, dense carrier networks, and ecosystems built up over two decades. Secondary markets were supposed to trade at a discount, because they had thinner fiber routes, smaller labor pools, and less demand. That framework assumed the gap between the two tiers would stay roughly fixed, but it hasn't. Markets with available power are pricing up no matter what tier they sit in, and markets without it are stalling no matter how prestigious their address. The tier label still describes where a market has been. It no longer predicts where its pricing is going.

Power scarcity as the operative pricing variable in established primary markets

Primary markets haven't lost their pull with tenants. They've run into a physical limit that turns scarcity into rent, and that limit gets measured in megawatts, not acres. Northern Virginia has effectively run out of available utility power near Ashburn, and Dominion Energy's interconnection queue for large commercial load additions now stretches past 36 months. A developer without an existing utility relationship in that market isn't moving quickly, no matter how much capital sits behind the deal. The same strain appears in Dallas, where ERCOT capacity is stretched thin; in Chicago, where ComEd's large-load queue has a real backlog; and in Phoenix, where SRP and APS are under pressure from both residential growth and new semiconductor load.

CBRE's H1 2026 North America Data Center Trends report lays out the numbers behind this squeeze. Vacancy across primary markets dropped to a record low in H1 2026, and that decline started the year before, but it kept going even as supply additions hit record volume. Preleasing kept pace with construction: commitments now cover more than four-fifths of all capacity currently under construction, up sharply from the prior year. If you add up future capacity across all primary markets, it covers roughly six months of demand at the current absorption rate. Northern Virginia, still the largest market by total inventory, saw its vacancy rate fall to 0.2%, a number that makes "available capacity" close to a theoretical concept rather than something a tenant can actually lease.

Rent follows directly from that supply floor. Tenants competed for what little capacity remained, so every deployment size in CBRE's H1 2026 report gained pricing ground, with 3-to-10-MW deployments climbing fastest, then mid-size deals, then hyperscale commitments. Capital markets have priced in the same scarcity. If you underwrite a data center asset, contracted power matters most, because utility capacity is locked in under a signed service agreement with a real delivery timeline. That's a different category from stranded power, where an allocation exists on paper with no signed agreement and no construction milestone behind it. Assets backed by contracted power trade at tight cap rates. Assets resting on an aspirational allocation carry a real premium, because investors are pricing the risk that the power never actually arrives.

Secondary markets are not simply cheaper versions of primary markets

Secondary markets are not a cut-rate version of Northern Virginia or Dallas. They run on a different cost structure entirely, shaped by their own power, land, and permitting conditions, and that structure makes some of them genuinely competitive while leaving others still well behind. Brightlio's September 2026 data center statistics compilation found that land costs more in primary markets than in secondary markets, and more in secondary markets than in tertiary markets. Hard costs for shell-and-core construction span a wide per-MW range, because power density, cooling approach, and local market conditions all drive it. Labor costs explain part of that spread, but subcontractor capacity matters just as much. Power-driven construction backlogs in primary markets tie up the same mechanical and electrical crews that a secondary-market project would need.

Capital is following power, not cheap dirt. Build's Q1 2026 analysis, citing PitchBook and Cushman & Wakefield data, points to three markets drawing real institutional interest as capital follows power. Columbus, Ohio has AEP Ohio building out grid infrastructure alongside Intel's semiconductor investment, and land costs sit well below primary-market levels, but state tax exemptions for new data center applicants paused as of May 2026. Indianapolis has available capacity through AES Indiana, its fiber infrastructure is growing, and labor costs run lower than coastal markets. Reno, Nevada has shown real deal activity as of Q1 2026, and PitchBook infrastructure deal flow and Cushman & Wakefield pipeline data back that up.

None of that makes secondary markets a blanket replacement for primary ones. Dark fiber gets thin outside the major corridors, and AI inference workloads have strict latency limits that rule out sites too far from population centers. Community opposition and permitting delays aren't a primary-market problem alone. CBRE's H1 2026 report names local opposition as a serious obstacle stalling projects across North America, and securing local approval now matters as much to site selection as power access. Mechanical and electrical subcontractor capacity is tight everywhere: lead times for medium-voltage switchgear run 52 to 78 weeks per Build's analysis, regardless of which tier the market sits in.

The emerging secondary markets succeeding right now, Columbus, Salt Lake City, San Antonio, Indianapolis, Reno, share one thing: developers there can get to power and move fast. Moving fast depends on tight coordination between site feasibility, design, documentation, and permitting, since a delay in any one of those steps erases the time advantage power access bought. Archil Labs connects layouts, routing, RFIs, and operations-ready handoff in a single workflow, which shortens the path from site selection through permitting and construction coordination. So that kind of coordination amplifies whatever economic edge a secondary market's power position already gives it. A secondary market with real, contracted power is not the same asset as a secondary market with power still on paper, and pricing is starting to reflect that distinction directly.

AI workload density and power access

Power access used to mean one thing: enough megawatts at the property line. That definition no longer holds. A site can have plenty of available power and still fall short if its infrastructure was built for CPU-era rack densities, because that infrastructure isn't actually usable for the workloads driving most of today's demand. That shift moves pricing pressure to a new place: not just which markets have power, but which sites have power built for the right density.

The problem is physical. AI accelerators need to exchange data constantly during training, which changes how a data hall has to be laid out. Cable and fiber routing designed for older east-west CPU traffic patterns doesn't hold up for GPU cluster fabrics. Layout and routing decisions now have to get made alongside power and cooling planning, not after it, the way they used to.

That adds real cost. Brightlio's data center statistics compilation found that AI infrastructure fit-outs can add substantially to per-MW costs on top of basic shell-and-core construction. A site's available-power position and its AI-readiness are two separate variables, and both now factor into pricing. Primary markets already short on power face a double disadvantage here: not enough megawatts to begin with, plus existing infrastructure that was never designed for AI-level density. Some secondary markets, building from scratch, can design for that density from day one, which is a real structural advantage over retrofitting an older primary-market facility.

Data Center Knowledge's 2026 trends analysis describes how engineering is moving toward productization, so you can deploy power systems, cooling plants, and white space as configurable industrial products, the same way across regions. Parametric MEP systems are becoming a baseline expectation for institutional capital, not a differentiator. Power access is a more layered variable than a megawatt count at the fence line, and that added layer affects pricing in both primary and secondary markets alike.

Permitting certainty and execution speed as the third pricing variable

If a site has power secured but permitting still uncertain, it doesn't price like a site that's genuinely shovel-ready. Execution risk is its own pricing variable alongside power access, priced before power access itself is resolved.

You see permitting pressure even in primary markets already squeezed on power. CBRE's H1 2026 report names community resistance and zoning delays as a serious obstacle stalling projects across North America, with local approval now mattering as much to site selection as power access. Northern Virginia's construction pace slowed relative to Atlanta specifically because of land and permitting limits, not because of power. Atlanta passed Northern Virginia as the top U.S. market for total construction in H1 2026, with under-construction capacity up 52.3% year over year, a shift that tracks permitting and land availability as much as it tracks power.

Construction execution adds another layer of risk on top of permitting. Developers who locked in guaranteed maximum price contracts back in 2024 can deliver projects below current replacement cost, but new projects get underwritten against much higher hard cost assumptions. Stranded power is an allocation that exists on paper without a signed agreement or construction milestone, so it deserves a real cap rate premium to compensate for execution risk. The same logic carries over to permitting: a project facing community opposition doesn't price like one that already has its approvals cleared.

Power, permitting, and execution function as three conditions that all have to hold at once, not three boxes checked off one after another. A site with abundant power but contested permitting isn't priced like a site with both resolved. Getting a project to the pricing premium its power position deserves depends on delivering the project correctly, and that's where execution strategy takes over.

Project delivery and the pricing premium

A site's fundamentals, power, permitting status, land cost, only set the ceiling on what it can command. Whether a project actually reaches that ceiling depends on how well the delivery gets executed, from design through construction to commissioning.

Understanding which sites have contracted power, not merely allocated power, requires designers and operators to work from real-time equipment specs, grid interconnection timelines, and load forecasts, all tied together rather than tracked separately. Platforms like Archil Labs automate equipment parameters and power requirements across a connected workflow, so you can model power availability and execution risk directly into site selection and design economics, instead of finding out about a power constraint halfway through the project.

That matters most when conditions shift mid-project, and that happens often in markets running 36-month interconnection queues. A utility allocation changes, or an equipment selection gets swapped, and the change cascades across layouts, power routing, cable paths, and documentation. Design automation platforms that keep those elements connected can cut out the manual rework that normally stretches out schedules and drives up cost when power parameters move. So a multi-day coordination problem turns into a single automated update instead.

Secondary markets bring their own version of this complexity. Permitting timelines, interconnection requirements, and land-constraint trade-offs all demand the same kind of structured traceability that primary-market projects need. If you're evaluating a site in Columbus or Reno, you get the same connected workflow, one that surfaces equipment specs, power requirements, and facility constraints early enough to shape site economics and execution planning before capital gets committed.

The pricing gap between primary and secondary markets was never really about geography. It tracks power availability, permitting certainty, and execution speed, and these three variables interact rather than stack in sequence. A market's tier label describes its history. Its pricing today reflects whether it can deliver power, approvals, and construction fast enough to meet demand that isn't waiting for any market to catch up.