1 hour ago
24
I Squared Capital just wrote a $225 million check for 10 data centers, and that’s apparently just the appetizer. The Miami-based infrastructure investment firm announced an agreement to acquire the portfolio from Cogent Fiber, LLC, with plans to funnel up to $1 billion in total capital into building a standalone US data center platform focused on AI inference workloads.
The deal, announced on May 26, covers approximately 53 megawatts of installed power capacity and 259,000 square feet of colocation space spread across nine major US markets. Cities on the list include Chicago, Atlanta, Phoenix, and Los Angeles. Closing is expected in Q3 2026 following regulatory approvals.
What I Squared is actually building
The 10 facilities are purpose-built to support liquid cooling and high-density configurations, both essential for the kind of GPU-packed racks that inference workloads demand. I Squared isn’t just buying buildings. The firm has committed up to $1 billion in total capital to grow the platform through infrastructure upgrades, facility expansions, and additional acquisitions. That means the $225 million purchase price represents roughly a quarter of the total planned investment.
Why Cogent is selling
On the other side of the transaction, Cogent Fiber is making a calculated retreat. The company plans to use proceeds from the sale to pay down existing debt while refocusing on its core operations. Cogent’s stock reacted positively to the announcement, climbing roughly 10% in pre-market trading.
What this means for investors
At $225 million for 53 MW of installed power capacity, the deal prices out to roughly $4.2 million per megawatt. That figure will matter as a benchmark for future transactions in the space.
Liquid cooling capability is becoming a key differentiator. Traditional air-cooled facilities struggle to handle the thermal output of modern GPU clusters. Facilities already equipped for liquid cooling, like the ones in this portfolio, command premium pricing because retrofitting older data centers is expensive and time-consuming.
The risk to watch is execution. Committing capital is one thing. Actually upgrading facilities, securing power contracts, attracting tenants, and managing the operational complexity of high-density AI deployments is another. Investors tracking this space should watch for utilization rates and power draw figures once the platform begins operating independently, likely sometime in late 2026 or early 2027.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.
English (US) ·