Cogent Communications has entered into a definitive agreement to sell 10 data centers in the United States for $225 million in cash to a new entity backed by I Squared Capital. The transaction, conducted through Cogent Fiber, involves facilities located in Phoenix, Anaheim, Burbank, Stockton, Atlanta, Chicago, Elkridge, Kansas City, Nashville, and Houston. The closing is scheduled for June 12, 2026, or when the applicable review period under U.S. antitrust law Hart-Scott-Rodino expires.
The deal goes beyond mere real estate asset rotation. I Squared Capital aims to use these data centers as the foundation for a new US platform focused on colocation, high-density deployments, and infrastructure for artificial intelligence inference. The firm has also announced a commitment of up to $1 billion to develop the platform through asset investments, customer-linked expansions, and new acquisitions.
This move reflects changing trends in the data center market. In recent years, much of the conversation has centered on large AI training campuses—enormous facilities with hundreds of megawatts or even gigawatt plans. But the next phase of AI also requires a different infrastructure approach: more distributed, interconnected centers closer to users and businesses, prepared to perform low-latency inference.
Ten Assets for an Inference and Edge Platform
The package acquired by I Squared totals approximately 53 MW of installed capacity and around 259,000 square feet of available colocation space, spread across nine US markets: Chicago, Atlanta, Phoenix, Los Angeles, Kansas City, Baltimore, Houston, Nashville, and Stockton. According to I Squared, these ten facilities are purpose-built data centers, with property owned and room for future expansions.
The buyer highlights three key features: location, power, and connectivity. This is no coincidence. Currently, securing available energy in connected locations has become one of the main bottlenecks for new data center projects. AI has increased demand, but timelines for land acquisition, permits, electrical interconnection, cooling, and network capacity remain lengthy.
I Squared also emphasizes that these assets are ready for higher-density deployments, including liquid cooling configurations. This is especially relevant for AI, where racks with GPUs or accelerators can easily surpass traditional colocation densities. While inference does not always demand the same power as large-scale training models, it requires stable capacity, low latency, robust network, and scalability.
| Key Data Point | Details |
|---|---|
| Purchase Price | $225 million in cash |
| Number of Data Centers | 10 |
| Markets Covered | 9 |
| Installed Power | Approximately 53 MW |
| Available Colocation Space | Approximately 259,000 sq ft |
| Additional Commitment by I Squared | Up to $1 billion |
| Focus of New Platform | Colocation, high density, and AI inference |
The list of cities is also noteworthy. It’s not just about major traditional hubs like Ashburn, Dallas, or Silicon Valley. I Squared is acquiring assets in regional markets with enterprise demand, good connectivity, and proximity to populations. The firm notes these nine markets serve over 63 million people collectively, making them attractive for low-latency workloads, content, retail and wholesale colocation, and inference services.
Cogent Monetizes Legacy Assets from Sprint
For Cogent, this sale aligns with a strategy of unlocking physical assets following the acquisition of Sprint’s wireline business from T-Mobile. Data Center Dynamics reported in 2025 that Cogent was converting former Sprint technical buildings and switching locations into colocation and edge centers. By Q2 2025, the company added seven more edge centers to its portfolio, reaching a total of 187 facilities, with over 2.1 million square feet, 26,100 cabinets, and 214 MW of owned power.
Selling these 10 data centers allows Cogent to generate liquidity from part of this portfolio. The company remains mainly a provider of high-speed internet access and private network services for bandwidth-intensive enterprise customers, with an optical IP network serving 306 global markets, according to its statement.
Strategically, this move separates business profiles: Cogent can focus on its core network and connectivity services, while I Squared aims to transform these assets into a dedicated data center platform. For infrastructure investors, regional centers with available energy, dense fiber, and owned facilities may be more valuable as part of a specialized platform than staying within a company primarily focused on IP connectivity.
AI Moving Toward Distributed Inference
I Squared’s interest in these assets is better understood within the context of AI’s evolution. Model training demands massive compute, energy, and cooling resources. Running models daily for users, enterprise applications, agents, assistants, search, voice, video, or automation requires a different architecture. Inference needs to be closer to consumption, reducing latency and managing continuous traffic.
This explains the appeal of a network of regional centers. Not all AI use cases will run in remote mega campuses. Real-time voice assistants, industrial applications, security systems, local recommendation engines, enterprise agents, or vision loads can benefit from infrastructure nearer to users or data sources.
I Squared describes these centers as being close to local internet exchanges, with multi-carrier connectivity, capable of hosting loads such as retail colocation, wholesale colocation, content, and latency-sensitive inference. The term “AI inference” is increasingly appearing in infrastructure operations as investors seek to anticipate demand that won’t mimic large training campuses.
| Type of Data Center | Typical Use | Critical Factor |
|---|---|---|
| AI Training Campus | Training large models | Massive energy, GPUs, cooling, and scale |
| Regional Colocation | Enterprises, content, networks, and digital services | Connectivity, proximity to clients, reliability |
| Edge / AI Inference | Real-time response and low latency | Location, network, available power, and density |
| Traditional Corporate Data Center | Internal workloads and continuity | Control, security, and predictable costs |
I Squared Strengthens Its Digital Infrastructure Focus
This acquisition also aligns with I Squared Capital’s broader digital infrastructure strategy. In April, the firm announced the acquisition of Elea Data Centers, one of Brazil’s largest carrier-neutral platforms, comprising nine interconnected campuses, over 300 MW of energized infrastructure, and more than 1 GW in development, requiring over $10 billion of capital for deployment.
This operation in Brazil and the US assets from Cogent share a common logic: building platforms where energy, connectivity, land, and AI demand converge. In Brazil, the appeal is based on renewable energy availability and the growth of cloud and AI in Latin America. In the US, the opportunity lies in a network of existing assets with available power and regional sites suitable for developing colocation and inference platforms.
The key difference from buying a large under-construction campus is that these assets already exist. This reduces some development risk but doesn’t eliminate the need for further investment. Upgrading facilities for high density, liquid cooling, new clients, and AI standards may require additional capital—hence the commitment of up to $1 billion.
Potential Market Implications
This deal confirms that AI infrastructure will not be limited to big announcements of mega campuses. The inference phase could spark a second wave of investment in medium-sized, regional, connected centers capable of upgrades. For owners of aging assets, technical buildings, or underutilized data centers, the market may offer an opportunity to sell or reposition those assets.
It may also increase competition in colocation. If new funds invest capital to adapt regional assets for high-density workloads, clients will have more options outside saturated major hubs. However, not all centers can be converted. Actual energy availability, network quality, building structure, cooling space, and proximity to customers will be key factors.
For Cogent, the success of this deal will depend on regulatory approval and how they utilize the proceeds. For I Squared, the challenge will be transforming these scattered facilities into a coherent platform with unified operations, clear commercial offerings, and the ability to attract AI, content, regional cloud, and enterprise clients.
Fundamentally, the message is clear: AI is increasing asset value, even for properties once considered secondary. A regional data center with power, fiber, and room for expansion can become an invaluable asset if inference is more distributed closer to the user. The data center war is no longer only about giant hyperscale hubs—it’s also about mid-sized cities, connected buildings, and nodes bringing AI closer to where it’s used.
Frequently Asked Questions
What has Cogent Communications sold?
Cogent has signed an agreement to sell 10 data centers in the US for $225 million in cash to a new entity backed by I Squared Capital.
Where are the sold data centers located?
Facilities are in Phoenix, Anaheim, Burbank, Stockton, Atlanta, Chicago, Elkridge, Kansas City, Nashville, and Houston.
What does I Squared plan to do with these assets?
I Squared aims to create a new US platform focused on colocation, high-density deployments, and AI inference infrastructure, with an additional commitment of up to $1 billion.
Why is AI inference important for regional data centers?
Because many AI applications require low latency and proximity to users or data sources. This can enhance the value of regional centers with good connectivity and available power.
via: prnewswire