BlackForge Data Centers
Menu

Water, gas & fiber

Fiber Connectivity for Data Center Sites

A data center site needs at least two physically diverse fiber routes, reachable at a reasonable construction cost and timeline, and ideally served by more than one provider. How much more it needs depends on the workload: latency-sensitive colocation and cloud sites want to be close to network hubs, while large AI training campuses mainly need high-capacity, diverse long-haul paths. Fiber is rarely the deciding factor on its own, but a site with no diversity or a long, difficult lateral can stall.

Last reviewed · 6 min read · BlackForge Data Centers

Key takeaways

  • Diversity means physically separate paths, not two providers sharing the same conduit or bridge.
  • Latency depends on route distance, not straight-line distance, and matters more for some workloads than others.
  • Lateral construction from the nearest route to the site is often the real cost and schedule question.
  • Get route maps from providers, usually under a nondisclosure agreement, rather than relying on public maps.
  • Plan at least two separate fiber entry points into the campus in the site layout.

01What fiber does a data center site need?

Every data center needs reliable, high-capacity connections to the networks it serves. For site selection, that breaks into four questions:

  • Diversity: can the site be reached by at least two physically separate routes, so a single cut does not isolate it?
  • Capacity and providers: are there long-haul and regional providers with routes near the site, and more than one of them?
  • Latency: how far, by fiber route, is the site from the network hubs, cloud regions or users it needs to reach?
  • Laterals: how far is it from those routes to the property, and what does building that connection involve?

The weighting depends on the data center type. Our guide to site needs by data center type covers how requirements shift from edge to hyperscale.

02Types of fiber routes

Fiber route types and their role in site selection
TypeWhat it isWhy it matters for a site
Long-haulHigh-count routes between cities and regions, often along railroads, highways or pipeline corridorsGives large campuses direct paths to distant markets and diverse routes out of the region
Regional and metroRings and routes within a metro area connecting data centers, carrier hotels and network hubsProvides low-latency access to interconnection points and other facilities
Lateral or last mileThe connection from a route to the buildingUsually new construction for a greenfield site; drives cost and timeline
Dark fiberUnlit fiber strands leased or bought, with the customer supplying the opticsGives large users control and capacity; common for hyperscale and AI campuses
Lit servicesBandwidth sold by a provider over its own equipmentSimpler to buy; common for smaller and enterprise facilities

Long-haul routes often follow linear corridors, so a site near a rail line or interstate may be closer to fiber than it appears. That is a starting point, not proof. Access points along long-haul routes are limited, and a route passing nearby may not have a place to connect.

03Route diversity: what actually counts

Two fiber providers do not automatically mean two diverse paths. Providers often share conduits, bridges, rail corridors and river crossings, so a single construction accident or flood can cut both. True diversity means the routes stay physically separate all the way from the building to different network hubs.

Check diversity at every stage:

  • Separate entry points into the property and the building, ideally on opposite sides of the site.
  • Separate lateral routes that do not share a road, bridge or easement.
  • Separate long-haul or metro routes that do not converge at a shared crossing.
  • Separate destinations, so the paths do not end at the same hub.

Verifying this takes route maps from each provider, typically shared under a nondisclosure agreement, overlaid on the site and its surroundings.

04Latency and distance

Light in optical fiber travels at roughly two-thirds of its speed in a vacuum, about 5 microseconds per kilometer one way, or roughly 1 millisecond of round-trip time for every 100 kilometers of fiber route. Fiber routes rarely run straight, so route distance is longer than the distance on a map, and network equipment adds delay.

Whether that matters depends on the workload. Interconnection-heavy colocation, financial services and cloud regions with multiple linked facilities are sensitive to latency and tend to cluster. Large AI training campuses mostly exchange data internally and can sit much farther from metro hubs, as long as they have high-capacity, diverse long-haul paths. Inference workloads that serve users directly fall in between. Our guide to AI data center site requirements covers how that shift opens up more remote sites.

05Building laterals: what drives cost and time

For most greenfield sites, the fiber question becomes a construction question: how to get from the nearest usable route to the property. The main drivers are:

  • Distance from the nearest access point on each route.
  • Aerial versus underground construction. Aerial lines need pole attachment rights and make-ready work; underground needs boring or trenching.
  • Crossings of railroads, highways, rivers and wetlands, each with its own permit or agreement.
  • Right-of-way: public road right-of-way permits and, where the route leaves public roads, easements across private land.
  • Provider interest: a provider may build a lateral at its own cost for a large customer, or require the customer to fund it.

06Fiber on a multi-building campus

On a large campus, fiber planning is part of the site plan. The layout needs at least two separate points of entry, conduit or duct banks from the property line to each building, and a path between buildings that does not create a single point of failure. Reserve these corridors early so they are not squeezed out by roads, substations and stormwater ponds. Site shape and frontage affect where entries can go, as our guide to site access, shape and frontage explains.

07How to evaluate fiber at a site

When we screen a site, fiber review follows a simple sequence:

  1. 01Map known long-haul and metro routes near the site from public and commercial sources.
  2. 02Identify the providers on each route and request route maps and access points.
  3. 03Sketch at least two lateral routes and list the crossings and landowners along each.
  4. 04Check route diversity from the site to separate network hubs.
  5. 05Estimate route distance and latency to the hubs or regions the user cares about.
  6. 06Ask providers about build interest, timing and whether they would fund the lateral.

Fiber is weighed against power, land and the other items in the site selection criteria. It is usually solvable, but how quickly and at what cost varies a great deal from site to site.

Common questions

How close does fiber need to be to a data center site?

There is no fixed distance. A site needs at least two diverse routes that can reach it through laterals at a reasonable cost and schedule. A short lateral along a public road can be straightforward, while a longer one with rail or river crossings takes more time. Latency-sensitive sites also need to be close, by route distance, to network hubs.

What is fiber route diversity?

Route diversity means a data center is connected by at least two fiber paths that are physically separate, so one cut, fire or flood cannot disconnect it. The paths should enter the property at different points, avoid shared conduits, bridges and crossings, and lead to different network hubs. Two providers sharing one conduit offer provider diversity but not true route diversity.

Does latency matter for AI data centers?

It depends on the workload. AI training mostly moves data within the campus, so training sites can be far from metro hubs as long as they have high-capacity, diverse long-haul fiber. Inference that serves users in real time is more sensitive to latency and benefits from being closer to population centers and network hubs, though usually less strictly than interconnection-heavy colocation.

What is dark fiber?

Dark fiber is optical fiber that is leased or purchased without a provider’s equipment attached. The customer installs its own optics to light the strands and controls how much capacity it runs over them. Large data center operators often prefer dark fiber because it offers high capacity, control and predictable long-term costs. Lit services, where the provider runs the equipment, suit smaller needs.

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.

Start a conversation →

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.

Related guides

More in Water, gas & fiber