I Squared Capital is betting $1 billion that the next phase of artificial intelligence will run not in massive training clusters but in edge data centers close to where people live and work.
I Squared Capital agreed to buy 10 data centers from Cogent Communications for $225 million in cash, seeding a new US platform targeting AI inference workloads across nine metro markets with a total commitment of as much as $1 billion.
"Location, power, and connectivity are the three variables that determine a data center's long-term value, and these facilities have all three in markets where new supply is severely constrained," Gautam Bhandari, co-founder, managing partner and global chief investment officer at I Squared Capital, said.
The seed portfolio comprises 53 megawatts of installed power capacity and 259,000 square feet of colocation space across Chicago, Atlanta, Phoenix, Los Angeles, Kansas City, Baltimore, Houston, Nashville and Stockton — markets that together serve more than 63 million people. All 10 facilities are owned fee simple, have room for expansion and support liquid-cooled, high-density configurations. The transaction is expected to close in the third quarter of 2026, subject to regulatory approvals including the expiration of the Hart-Scott-Rodino waiting period.
The bet reflects a structural shift as AI moves from the training phase — where models are built in hyperscale clusters dominated by companies such as Equinix and Digital Realty — to the inference phase, where models are deployed in real time for consumers and businesses. Inference workloads favor distributed, low-latency facilities over centralized mega-campuses, creating an opening for infrastructure investors to acquire assets in secondary markets where new supply is constrained by power availability and zoning restrictions.
The Inference Opportunity
I Squared, which manages $60 billion in assets across more than 100 companies in 70 countries, is establishing a standalone operating platform with a management team in place from day one. The facilities offer multi-carrier interconnection and proximity to local internet exchanges, positioning them for retail colocation, wholesale colocation, content delivery and latency-sensitive AI inference workloads.
The $225 million acquisition price values the portfolio at roughly $4.25 million per megawatt of installed capacity, a discount to the $8 million to $12 million per MW that hyperscale data centers command in primary markets such as Northern Virginia, according to industry data. The lower cost base gives I Squared flexibility to compete on pricing for colocation customers while investing in upgrades for higher-density deployments.
Cogent Communications, the seller, is monetizing a non-core asset as it focuses on its fiber and internet connectivity business. The Washington-based company, which trades on Nasdaq under the ticker CCOI, did not disclose the use of proceeds from the sale.
TVG Consulting Group served as commercial and market advisor on the transaction. Kirkland & Ellis provided legal counsel, and FTI Consulting served as carve-out and stand-up advisor.
Investor Takeaway
For investors, the deal signals that institutional capital sees AI inference infrastructure as a multiyear growth theme distinct from the hyperscale buildout. I Squared's $1 billion commitment — roughly 1.7 percent of its $60 billion in assets under management — is a measured bet on a thesis that will take years to play out. The platform's success hinges on whether enterprise demand for edge inference materializes at the scale that cloud providers have achieved for training workloads. If it does, the portfolio's 53 MW of existing capacity could expand significantly through the fee-simple land holdings at each facility.
This article is for informational purposes only and does not constitute investment advice.