Primary vs Secondary
AI Data Center Markets.
The map of US data center capacity is being rewritten. Northern Virginia is full. Power, water, and policy are pushing AI campuses into a second tier of markets. Where the next gigawatt lands has stopped being a real estate question and become a capital allocation question.
The short answer
Primary markets buy latency and comps at a premium. Secondary markets buy power and time at a discount. The right market is whichever one your workload, your tenant, and your financing window actually need.
The fourteen market scan
Where the deals are actually landing.
Six primary corridors carry most of the announced capacity. Four secondary corridors are absorbing the spill. Four emerging corridors are the next places to watch.
Northern Virginia
4 to 5 GW deployed · multi year wait on new interconnect
- ·Dominion Energy load forecast doubled in 24 months
- ·Loudoun County moratorium pressure on greenfield builds
- ·Anchor for AWS, Microsoft, Google, Meta hyperscale regions
Saturation. New AI campuses can be permitted but cannot get power inside a financing window. Demand spilling to PJM adjacencies.
Dallas · Fort Worth
ERCOT grid · large block additions possible with on site gas
- ·Behind the meter gas plants approved at multi GW scale
- ·Friendly permitting and competitive land basis
- ·Anchor tenants include Meta, Google, ORACLE, neoclouds
Hottest non Virginia market. Constraints shifting from land to skilled labor and water rights along the Trinity basin.
Phoenix · Goodyear · Mesa
APS and SRP service territory · large block constrained but expanding
- ·Major hyperscale campuses live and under construction
- ·Water consumption under public scrutiny
- ·TSMC, Intel adjacency anchors broader tech corridor
Water is the binding constraint for liquid cooled AI loads. Sponsors solving it with closed loop and dry cooler builds.
Columbus · Central Ohio
AEP service territory · multi GW load forecast inside five years
- ·Intel Ohio One semiconductor fab anchoring the corridor
- ·Google, Meta, AWS all expanding
- ·New Albany and Hilliard absorbing pipeline
Transitioning from secondary to primary fast. Land basis still attractive versus coastal markets.
Atlanta · Douglas County
Georgia Power · large block additions tied to integrated resource plan
- ·Microsoft and Meta anchoring greenfield campuses
- ·Strong fiber backbone, competitive cost structure
- ·Vogtle nuclear adds long dated firm power supply
Pulls deals away from Virginia for buyers who can take a slightly higher latency to Tier 1 peering.
Chicago · Northern Illinois
ComEd and PJM West · large block additions possible with grid upgrades
- ·Microsoft, Meta, AWS established presence
- ·Strong network interconnect ecosystem
- ·Aurora and Elk Grove Village absorbing AI builds
Mature primary. Most attractive for inference at scale and enterprise hybrid landing zones.
Reno · Northern Nevada
NV Energy · meaningful new capacity but transformer lead times bind
- ·Apple, Google, Switch anchor presence
- ·Land basis a fraction of Phoenix or coastal markets
- ·Strong PPA optionality with regional renewable additions
Climbing fast. Sponsors using Nevada to derisk water and land while staying inside western interconnect.
Salt Lake City · Wasatch Front
Rocky Mountain Power · growth corridor along the Wasatch front
- ·Adobe, NSA Bumblehive, smaller hyperscale and neocloud anchors
- ·Skilled workforce from local universities
- ·Strong fiber connectivity to Denver and west coast
Wins on talent and cost. Loses on absolute block size. Best fit for mid sized AI campuses and edge inference.
Quincy · Central Washington
BPA hydro · firm renewable power with tight allocation
- ·Microsoft, Yahoo, Sabey established
- ·Hydroelectric base load remains a differentiator
- ·Cooler climate reduces mechanical load on liquid systems
Power allocation is the gate. Sponsors who lock in hydro PPAs get a structural cost advantage.
Kansas City · Topeka
Evergy · grid upgrades underway · land and water favorable
- ·Meta and Google announcing greenfield builds
- ·Central geography for latency to both coasts
- ·Friendly state level incentives and workforce programs
Emerging as a real second tier hub. Watching transformer lead times and substation upgrades closely.
Indianapolis · Indiana
AES Indiana · meaningful capacity available with willing regulator
- ·New hyperscale campuses announced in 2025
- ·Land basis attractive · workforce drawing from Purdue
- ·Indiana Economic Development Corp aggressively courting AI
Will move from emerging to secondary inside two years if substation upgrades stay on schedule.
Cheyenne · Wyoming
Black Hills Energy · wind dominant generation · grid upgrades planned
- ·Microsoft long history with Cheyenne presence
- ·Cool climate, low land basis
- ·Strong wind PPA pipeline · grid is the binding constraint
Energy abundant. Buildable parcels abundant. Transmission and substation capacity is the choke point.
Memphis · Mississippi
TVA · meaningful blocks possible with multi year planning
- ·xAI Colossus drew capital, attention, and a buildable template
- ·Strong fiber along the I 40 corridor
- ·Land and water cheaper than peer markets
TVA decisions and water rights will decide pace. Could become a major secondary inside three years.
Lincoln · Omaha
OPPD and LES · grid upgrades planned · favorable land basis
- ·Google and Meta footprints established
- ·Strong fiber connectivity east west
- ·Climate moderate for mechanical efficiency
Quiet contender. Watching substation upgrades and state workforce policy as the pace decider.
The constraint stack
Eight axes that decide pace.
Underwriting AI capacity is no longer a real estate exercise. It is a constraint stack problem where the binding factor shifts market to market.
Power availability
Multi year waits on large blocks. Interconnect queues measured in years.
Mid sized blocks available inside a financing window. Smaller margin for upside expansion.
Transformer lead times
24 to 36 months for large units, with constant slippage. Reserve early or lose the deal.
Still long, slightly easier to slot inside the queue. Sponsors hold inventory more aggressively.
Land basis
Per acre pricing has tripled in five years in the top corridors. Buildable parcels scarce.
Land remains a fraction of primary basis. Buildable parcels still plentiful.
Workforce
Deep operator, electrician, and mechanical contractor pools. Wage pressure increasing.
Thinner pools. Sponsors stand up training partnerships with local community colleges and trade unions.
Latency to peering
Sub millisecond to major hyperscale regions. Optimal for inference at scale.
Few milliseconds penalty. Acceptable for training workloads and most inference patterns.
Regulatory posture
Increasing scrutiny on grid impact, water use, and community impact. Permitting is slower.
Active courting by state and county. Faster permits and stronger incentive packages.
Water availability
Constrained in Virginia, Phoenix, and Texas. Closed loop and dry cooling becoming standard.
Less constrained in Plains and Mountain West. Still requires diligence.
Capital efficiency
Higher absolute cost per MW. Stronger comps for exit and refinancing.
Lower cost per MW. Thinner comps. Higher reliance on anchor tenant credit.
Four sponsor playbooks
Where you land depends on what you are building.
Hyperscale primary build
When latency to existing regions and proximity to peering is mission critical · large inference fleets, enterprise hybrid landing zones, model serving for consumer products. Pay the premium. Northern Virginia, Phoenix, Atlanta, Dallas.
Training factory secondary build
When the workload is sustained dense training, latency is secondary, and the math favors land plus power abundance over peering. Reno, Salt Lake, Quincy, Cheyenne, Kansas City.
Sovereign or industrial off taker build
When the sponsor is a national program or an industrial off taker that wants dedicated capacity and a long dated PPA. Lean to secondary or emerging markets with friendly regulators and abundant generation.
Neocloud capacity expansion
When tenant credit is improving and the sponsor needs to ladder capacity inside financing windows. Mix primary anchors with secondary expansion campuses to balance cost and time to power.
The framing line
The next gigawatt is not going where the cloud went. It is going where the electrons, the water, and the willing regulators are. Markets that solve all three first will compound the next decade of AI capacity.
Where Bttr. operates
Geography is the substrate. Product is the surface.
Sponsors picking a market still need a product on top of it. Bttr. designs the buyer surfaces, capacity contract experiences, operator dashboards, and financing flows that turn a campus into something a tenant can actually evaluate and underwrite.
Frequently asked
The questions sponsors actually ask.
What is a primary AI data center market?
A market with multi GW deployed capacity, established hyperscale anchors, deep operator labor pools, sub millisecond latency to major peering, and the worst supply constraints. Northern Virginia, Dallas, Phoenix, Columbus, Atlanta, Chicago are the canonical six.
What is a secondary AI data center market?
A market with meaningful new capacity in development, lower land and operating basis, smaller absolute block sizes, and a clearer path to power inside a financing window. Reno, Salt Lake City, Quincy, Kansas City sit here today.
Why is Northern Virginia full?
Loudoun County and Prince William County together host the densest concentration of data center load on earth. Dominion Energy load forecasts have doubled in 24 months. Substation, transmission, and substation queue capacity all bind. New AI campuses can be permitted, but cannot get power inside a financing window.
Why is Texas attractive for AI?
ERCOT allows behind the meter generation at scale, permitting is faster than in PJM, land basis is competitive, and the state has actively courted hyperscale and neocloud sponsors. Constraints shift toward water rights and skilled labor as the next binding factors.
How do sponsors decide between primary and secondary markets?
Three questions. One · is latency to existing regions mission critical for the workload? Two · is the financing window short enough that power timing matters more than cost? Three · is the anchor tenant credit strong enough to underwrite a secondary market with thinner comps? The answers map deals to corridors.
Which emerging markets are most likely to graduate to secondary?
Indianapolis, Cheyenne, Memphis, and Lincoln are the most likely to move up the ladder inside two to three years if substation and transmission upgrades stay on schedule. Each has either an anchor tenant signal already in place or strong regulator support.
Where does Bttr. operate inside this map?
Bttr. designs the commercialization layer on top of these campuses · buyer portals, capacity contract experiences, dashboards, financing flows, and the operating systems that turn a campus into a product. Geography is the substrate. The product surfaces are where we work.
The Frontier Series
Get the next Frontier brief in your inbox.
Field notes on AI infrastructure commercialization · power, capital, contracts, and the markets where the next gigawatt is landing. One brief, no filler.