Digital MidstreamEDGE AI INFRASTRUCTURE

EDGE AI DATA CENTER DEVELOPMENT

Inference.
At the edge.

We identify and develop highly interconnected sites for AI inference. Existing telecom infrastructure. Strategic new locations. Closer to the networks that serve users.

Build at the edge
DM / EDGE INFRASTRUCTUREINTERCONNECTION + INFERENCE
Conceptual network of interconnected edge inference locations, networks and applications
01 / The interconnected edgeConcept illustration

OUR FOCUS

Developing the
interconnected edge.

Two paths to edge AI infrastructure: bring existing facilities into a new era, or develop in the right connected location.

PURPOSE-BUILT FOR THE INFERENCE ERA

The site is the starting point.
The network is the advantage.

Our specialty is finding and developing locations where AI inference infrastructure can connect to the networks and applications it serves.

Repurpose existing sites

Identify disused edge data centers and telecom facilities, evaluating their existing infrastructure and potential for a new role in AI inference.

Develop strategic locations

Identify and develop new edge data center sites with a focus on strong interconnection, fiber access and diverse network paths.

Discuss an edge site ↗︎

THE EXISTING INFRASTRUCTURE OPPORTUNITY

The next edge may
already be standing.

Former and underused telecom facilities can bring together established power service, fiber and proximity to users.

~25,000

U.S. central offices

Schneider Electric’s 2019 estimate of the national footprint—a historical indication of scale, not a current inventory of available sites.

Schneider Electric · 2019 ↗︎
139

Former Sprint sites converted

Cogent reported 52 data centers and 87 edge data centers converted from former Sprint facilities by year-end 2025.

Cogent · 2025 annual report ↗︎
74

Underused central offices

AT&T’s January 2025 sale-leaseback covered more than 13 million square feet while preserving space needed for network operations.

AT&T · January 2025 ↗︎

Existing power.
A head start worth investigating.

Megawatt-scale spare power is documented at some telecom sites. In a 2025 interview, Ziply Fiber described four or five sites with more than 1 MW available, alongside many smaller facilities.

Our focus is finding the sites where existing power and interconnection can support a credible development path. For a 1–5 MW opportunity, usable capacity, cooling, electrical conversion and network access must be verified at the individual site.

Ziply Fiber executive interview · 2025 ↗︎

Industry examples, not Digital Midstream inventory. These sources do not establish how many U.S. sites have 1–5 MW of spare, deployable capacity today.

WHY LOCAL INFERENCE

A shorter path
to a useful answer.

Locating inference near users—and interconnecting with the providers that serve them—can reduce the network portion of response time.

How interconnection reduces network latency ↗︎

Connect to local networks

Rich interconnection with local ISPs and access providers creates opportunities for direct paths, avoiding unnecessary trips through distant network hubs.

Build around the user experience

Shorter network round trips can help conversational AI, voice interfaces and other interactive applications respond more promptly.

Measure the whole response

Geographic proximity is only part of the story. Peering, routing and congestion matter, alongside model execution, available compute and queueing time.

The development perspective

Inference runs on compute. Its reach depends on the infrastructure around it.

THE PHYSICAL FOUNDATION

Connected. Powered. Buildable.

We evaluate each location as part of a connected system.

01 / LAND

Room to develop.

Location, access, planning constraints and room for expansion shape a site’s potential long before construction begins.

02 / POWER

A path to capacity.

Available capacity, connection requirements and delivery timelines are central to evaluating a development opportunity.

03 / CONNECTIVITY

Connected from the start.

Fiber access, route diversity and proximity to network infrastructure matter alongside the physical site.

INTEGRATED SITE PREPARATION

From site potential
to work on the ground.

Our development team works with an established excavation and site engineering business through a family-business relationship.

This relationship brings site preparation expertise into early planning, connecting site selection with the work required on the ground.

Discuss development and site preparation ↗︎

Our pipeline

Macon, Georgia

1 MW site

Under contract

Feasibility Study

Evaluating conversion and reuse opportunities across Georgia, Florida, and Tennessee.

Priority review

Rogersville, Tennessee

Industrial conversion

5–6 MW preliminary IT target

Priority review

Winder, Georgia

Industrial conversion

5–6 MW preliminary IT target

Priority review

Hartwell, Georgia

Phased conversion

0.5–1.5 MW preliminary IT target

Priority review

Orlando, Florida

Former telecom data center reuse

1–2 MW preliminary IT target

Reserve candidate

Savannah, Georgia

Regional inference site

1–2 MW preliminary IT target

Reserve candidate

Eatonton, Georgia

Industrial conversion

0.75–1.25 MW preliminary IT target

Initial screening · September 2026. These sites are feasibility candidates, not controlled or committed projects. IT capacity ranges are preliminary engineering estimates, subject to utility confirmation, site condition, availability, and commercial diligence.

OUR TEAM & PARTNERS

Operator-led.
Privately backed.

Digital Midstream brings together a development team, private capital partners and site engineering relationships. Led by our managing partner, the team identifies and evaluates opportunities, coordinates diligence and advances sites toward development.

Our managing partner brings ten years of experience in CDN, global connectivity and traffic engineering. In recent years, that work has expanded to the network infrastructure and traffic steering behind distributed AI inference. It informs how we assess interconnection, compute placement and the needs of prospective operators.

Our industry relationships can open conversations with prospective edge inference tenants. We explore introductions where an operator’s requirements align with a site’s potential, with any tenancy subject to independent evaluation and agreement.

Explore Tucker’s professional background ↗︎

Start a conversation

Let’s build
the next edge.

Have a former telecom facility, a disused data center or a strategically connected location? Let’s explore its potential for AI inference.