Key takeaways
- Latency-sensitive and interconnection-heavy workloads favor metro locations; AI training and batch workloads are relatively latency-insensitive and can go where power is.35
- Light in standard fiber covers a kilometer in about 4.9 microseconds, so every 100 km of route adds roughly 1 ms of round-trip delay before equipment delay is counted.6
- Established metro markets are crowded on the grid: Dominion Energy reports about 70 GW of large-load requests in its Virginia queue, nearly three times its all-time peak.7
- Remote sites often have power and land sooner but need long-haul fiber, construction labor and local capacity built up; one rural Louisiana campus expected a peak construction workforce of 5,000.8
- Local opposition is widespread: tracked opposition groups roughly doubled in early 2026 and were active in 49 states.9
- The outer ring of a metro, on a high-voltage corridor, is often the practical compromise.
01Why this trade-off exists
Data center demand concentrates where users, businesses and network exchanges are. Grid capacity does not follow the same map. Metro areas have dense fiber and a long history of data center development, but their transmission systems are also serving homes, offices and industry, and spare capacity can be scarce. Rural areas near large generation or strong transmission may have megawatts available but little else a data center needs.
As loads have grown into the hundreds of megawatts, power has become the binding constraint more often. CBRE reported that power constraints were extending completion timelines in the primary North American markets in the first half of 2026, even as vacancy fell to a record low.1 Secondary coverage of JLL’s 2026 outlook put the national average wait for a utility connection at about four years, with longer waits in some primary hubs.2
That has pushed developers to look further out. CBRE’s 2026 global report highlights emerging locations such as Tennessee and West Texas, where land and power availability have drawn new campuses.10 The question for any one project is how far to go, and what the move costs. See what makes land “powered land” for the power side of that search, and emerging data center markets for where activity is moving.
Fig. 1Power scarcity in the core markets, 2026
02Near-metro and remote sites compared
| Factor | Near the metro | Remote, near the power |
|---|---|---|
| Latency to users | Low | Higher; depends on distance and fiber route |
| Fiber | Many carriers, diverse paths, exchange points nearby | Often one or two long-haul routes; laterals may need building |
| Power availability | Often constrained; multi-year timelines for large loads in major hubs2 | Can be strong near large generation or high-voltage lines |
| Land | Scarce, expensive, smaller tracts | Large tracts, lower cost per acre |
| Labor | Deep construction and operations workforce | Smaller local pool; crews may travel |
| Entitlement | Familiar process, but tightening in some markets11 | Less precedent; process may be new to the jurisdiction |
| Community | Established expectations; sensitivity to further growth | Large change for a small community |
| Water and sewer | Municipal systems, possibly constrained | May need wells, new mains or on-site treatment |
These are tendencies, not rules. Some metros still have strong grid capacity on their edges. Some remote sites have excellent fiber because a long-haul route runs past them.
03The workload decides much of the answer
Start with what will run in the building. Different workloads tolerate distance very differently, and our guide to latency requirements by workload goes into the numbers.
- Interconnection and colocation. Facilities whose value comes from customers and carriers meeting each other need to be near internet exchange points and carrier hotels and dense fiber. They tend to stay close to metros.
- Cloud regions. Cloud providers place capacity within network reach of their users and of their other sites in the region. They can move outward, but within limits set by their network design.
- AI inference. Inference has to return answers in real time or near real time, which pulls it toward users through regional facilities; industry analysts describe smaller, latency-optimized inference sites near population centers.35
- AI training and batch processing. Training needs a great deal of power over long runs and is relatively latency-insensitive, so it can be housed in remote, centralized campuses where land, water and energy are available.34 See AI training vs. inference site needs.
The physics sets a floor. Light in standard single-mode fiber travels about 4.9 microseconds per kilometer, so 100 km of route adds roughly 0.5 ms one way and about 1 ms round trip, before routers and other equipment add their own delay.6 Real routes are longer than straight lines, so measure latency on the actual fiber path, not on a map.
Fig. 2Minimum round-trip fiber delay by route length
- 50 km~0.5 ms
- 100 km~1 ms
- 300 km~3 ms
- 800 km~8 ms
- 1,600 km~16 ms
milliseconds
Fig. 3Where workloads tend to land
Follow the power
AI training and batch campuses on remote, power-rich sites.
Outer ring
Large cloud and inference capacity near a metro, on a strong corridor.
Flexible
Backup, archive and enterprise loads that can sit in many places.
Stay in the metro
Interconnection, colocation and edge sites near exchanges.
Tolerant ← Latency sensitivity → Sensitive
04What a remote site has to solve
A remote site with strong power is not finished when the utility says yes. Several things a metro site takes for granted have to be planned for.
- Fiber. Distance to a long-haul route, the number of carriers on it, and whether a second, diverse path is possible. See long-haul and subsea fiber routes and fiber route diversity and latency.
- Construction labor. Large campuses need many skilled trades for years. Meta reported that its Richland Parish, Louisiana, campus had supported 3,700 construction workers by late 2025 and expected a peak of 5,000 by June 2026.8 In rural areas, crews of that size need housing and travel, which affects cost and schedule.
- Operations staff. Permanent jobs are fewer than construction jobs, but they need people who can live nearby.
- Roads and access. Rural roads may not be built for heavy equipment deliveries, and upgrades may fall to the developer.
- Water and wastewater. Municipal systems may be small or absent. Wells, new mains or on-site treatment may be needed. See municipal water and sewer capacity.
- Local capacity. A small county may not have staff experienced with large industrial projects, which affects how long review takes.
Power-first siting also tends to change what “a good availability zone” means. One analysis of AI location strategy argues that the best zones are increasingly defined by energy access rather than fiber density, which puts the burden on the developer to bring the fiber.4
05What a near-metro site has to solve
Metro-adjacent sites come with their own problems, and they are getting harder in established markets.
- Grid congestion. Dominion Energy, which serves Northern Virginia, reports about 70 GW of large-load delivery points in its queue: about 25 GW with projected connection dates through 2031 and about 45 GW still under study.7 The Data Center Coalition told Virginia regulators in 2026 that the proposed queue could take up to 15 years to work through.12
- Land cost and size. Tracts large enough for a campus are scarce, and prices reflect competition from other uses.
- Local restrictions. In March 2025, Loudoun County, Virginia, voted to make data centers a special exception use in districts where they had been allowed by right, which adds public hearings before the planning commission and board.11 See data center moratoriums and local restrictions.
- Neighbors. Suburban sites are more likely to border homes, which raises noise, visual and traffic concerns.
Community pushback is not limited to the metro, though. Data Center Watch counted at least 75 projects worth about $130 billion disrupted by local opposition in the first quarter of 2026, and active opposition groups grew from 396 at the end of 2025 to 833 by March 2026, across 49 states.9 A remote site with no zoning framework for data centers can move quickly, or it can become the county’s first contested hearing.
06The middle ground: the outer ring
Many developers look first at the outer ring of a metro area, where a high-voltage transmission corridor, a long-haul fiber route and a highway meet. These locations can be close enough for many cloud and inference workloads, with more land and sometimes more grid room than the core market. Our comparison of rural vs. suburban data center sites looks at the same choice from the land side.
The outer ring is also where competition moves next, so the window can be short. Screening a corridor early, before other developers have filed large-load requests on the same substation, matters more than picking the perfect parcel.
Look for a few signs that an outer-ring corridor can work: a substation fed by more than one high-voltage line, a long-haul fiber route with more than one carrier, a highway interchange within a short distance, and a jurisdiction with industrial zoning or a track record of approving large projects. A site with all four is uncommon. A site with three, and a clear plan for the fourth, is usually worth a closer look.
07How to decide: a short checklist
- 01Define the workload and its real latency tolerance, in network terms, not miles.
- 02Map where power can be delivered at the needed load and date, starting from transmission and substations. See power timelines and interconnection queues.
- 03For each candidate area, check fiber: carriers, routes and the cost of reaching a second diverse path.
- 04Check labor and construction logistics for the full build period, including housing for peak crews.
- 05Screen local zoning, restrictions and community history in each jurisdiction.
- 06Confirm a water and wastewater plan that fits the cooling design.
- 07Compare total time to revenue, not land price or power price alone.
When we run a national search, we usually start from the power map and the fiber map together, then let the workload’s distance tolerance set how far from the metro we look. If you have a candidate tract, you can get a site reviewed.
Common questions
How far from a city can a data center be?
It depends on the workload. Interconnection-focused colocation generally needs to be in or near the metro, and inference is pulled toward users, while AI training and batch work are relatively latency-insensitive and can sit far from them if fiber is adequate.3 What matters is latency over the actual fiber route: about 1 ms of round-trip delay per 100 km of fiber before equipment delay, on a route that is usually longer than the straight-line distance.6
Why are data centers being built in rural areas?
Mainly because power and large tracts of land are easier to find there. In established markets, vacancy is at a record low1 and reported grid waits average about four years.2 Rural sites near high-voltage transmission or large generation can sometimes be served sooner, and CBRE points to emerging markets where land and power are more available.10 The trade-offs are thinner fiber, smaller labor pools and less local experience with data center projects.
Do remote data center sites have enough fiber?
Some do and some do not. Sites along a long-haul fiber route can reach several carriers with a short lateral. Sites far from any route may need many miles of new construction, and a second diverse path can be harder still. Fiber should be checked early, including the number of carriers, route diversity and the cost of laterals, before the land is committed.
Is land cheaper for remote data center sites?
Usually, per acre, yes. Rural land typically costs less than land near a metro, and larger tracts are easier to find. The total cost picture is less simple. Remote sites may need fiber laterals, road upgrades, water and wastewater infrastructure and a traveling construction workforce that can number in the thousands at peak.8 Comparing sites on time to revenue and total development cost gives a truer picture than comparing land prices alone.
Is community opposition lower at remote sites?
Not reliably. Opposition groups tracked by Data Center Watch were active in 49 states by early 2026.9 A small community can feel a large campus more than a suburb does, so local outreach and a clear account of water, noise and tax effects matter in both settings.
Notes
- 1.Construction Dive, “Data center vacancy hits all-time low amid record construction,” 2026. constructiondive.com
- 2.Electrical Contractor Magazine, “How Long Does It Take to Power a Data Center?,” 2026. ecmag.com
- 3.Analysys Mason, “AI data centre,” 2025. analysysmason.com
- 4.STL Partners, “The power-first data centre: how AI is reshaping location strategy,” n.d. stlpartners.com
- 5.Data Center Knowledge, “AI and Latency: Why Milliseconds Decide Winners and Losers in the Data Center Race,” n.d. datacenterknowledge.com
- 6.Telecom Engineering, “Tip of the Week, 04 Feb 2011,” 2011. telecomengineering.com
- 7.Data Center Dynamics, “Dominion files large load connection queue plan with state regulators,” 2026. datacenterdynamics.com
- 8.Meta, “Meta’s Richland Parish Data Center Supports Louisiana Economy, $875 Million in Contracts,” 2025. about.fb.com
- 9.Data Center Watch, “Q1 2026: Data Center Watch Report,” 2026. datacenterwatch.org
- 10.CBRE, “Global Data Center Trends 2026,” 2026. cbre.com
- 11.McGuireWoods, “Loudoun County Eliminates By-Right Use for Data Centers, Increasing Development Hurdles,” 2025. mcguirewoods.com
- 12.Virginia Mercury, “New state law mandates review of Dominion’s load forecasting, as data centers raise concerns,” 2026. virginiamercury.com
Have a site in mind?
Get a straight answer on your land.
Send a parcel number, an address, a map pin or a target load. We’ll tell you what it can support and what it would take.
This guide is general information about data center site selection. It is not engineering, legal, tax or investment advice. Requirements vary by state, utility and county, so confirm the specifics for any site with the relevant authorities and advisors.
Terms in this guide
Related guides
More in Decisions and trade-offs
- Buy, Option or Lease: How Developers Secure Data Center Land
- Phased vs. Full Build-Out: How to Stage a Data Center Campus
- New Site or Expand an Existing Campus? How to Decide
- Grid Power vs. On-Site Generation: How to Decide
- One Large Data Center Campus vs. Several Smaller Sites
- Build, Lease or Colocate? How to Source Data Center Capacity
- Regulated Utility vs. Competitive Market Sites for Data Centers
- Tax Incentives vs. Power Costs: Which Matters More for a Data Center Site?
- Utility-Owned vs. Customer-Owned Substations for Data Centers
- Converting a Warehouse or Factory to a Data Center vs. Building New
- How to Choose a Data Center Site Selection Firm
- Site Selection Consultant vs. Broker vs. Engineer: Who Does What
Guide 6 of 14 in Decisions and trade-offs · all 242 guides · glossary · calculators
