I Squared Capital buys $225M data center portfolio from Cogent Fiber

This article is for informational purposes only. Always verify information independently before making any decisions.

We’re watching I Squared Capital acquire a $225 million data center portfolio from Cogent Fiber to serve as the backbone for its new AI inference platform. The transaction launches a planned $1 billion investment program targeting US-based artificial intelligence infrastructure, according to Stocktitan and Channelnewsasia. It covers 10 strategically positioned data centers equipped for high-intensity AI workloads, positioning I Squared for aggressive expansion into the AI infrastructure sector.

Cogent Fiber exits these non-core assets as investor demand for AI-ready digital infrastructure accelerates across the US.

Channelnewsasia reports that I Squared established selection criteria focused on sites with robust power access and high-volume network connectivity to support GPU-based inference at an industrial scale. Their asset sweep prioritized edge locations in Atlanta, Dallas, Chicago, and Phoenix—four markets with consistently strong enterprise AI spending and fiber density.

retrofits for AI payloads drive distinct capital allocation compared to legacy server hosting. I Squared’s core plan is to retrofit all acquired sites specifically for AI inference workloads—a path that requires targeted investment in advanced GPUs and thermal management, according to Realassets.

City	Site Count	Power Capacity (MW)	Connectivity
Atlanta	3	~30 MW	Multi-carrier fiber
Chicago	2	~20 MW	Edge connectivity
Dallas	3	~25 MW	Enterprise interconnect
Phoenix	2	~15 MW	Cloud peering

---

Investment Approach

I Squared’s core plan is to retrofit all acquired sites specifically for AI inference workloads—a path that requires targeted investment in advanced GPUs and thermal management, according to Realassets.

So I Squared’s $1 billion capital commitment is structured for staggered deployment, targeting enterprise AI customers who need reliable, low-latency responses at scale. The company aims to bring the initial batch of AI-optimized pods online by early 2027, backing applications in healthcare, financial analytics, and logistics automation that rely on rapid inference cycles.

• High-density retrofits:Prioritize increased rack densities and modular scaling in select US metro zones.
• AI model hosting:Keys to success include attracting cloud SaaS firms, autonomous systems developers, and quantitative research teams.
• Energy sourcing:At least 4 sites are being integrated with local renewables or utility-backed microgrids.
• Deployment window:Phase 1 launches are scheduled for Q1–Q3 2027, setting a clear benchmark for market watchers.

---

Secondary Focus

Realassets notes that beyond hardware retrofits, I Squared aims to tightly integrate the new sites with fiber backbone networks and direct cloud on-ramps—an essential step as major cloud providers look to outsource more inference workloads. According to Stocktitan, Cogent Fiber’s previous tenants ranged from regional ISPs to streaming providers, offering a base for expansion. Now I Squared intends to provide direct managed services for AI model deployment and scaling, enabling customers to shift their trained models seamlessly into production-ready environments.

Channelnewsasia reports that new partnerships with network specialists will ensure that customer queries can be processed between regional clusters in under 20 milliseconds—crucial for real-time applications.

Integration Area	Purpose	Target Tenants
Direct fiber connects	Low-latency model serving	Cloud platforms, AI SaaS
Colocation services	Customizable hardware	Enterprises, fintech, health AI
Deployment management	Seamless model transition	Startups, integrators
Sustainability retrofits	ESG compliance, renewable offsets	Fortune 100 clients

---

Week in Review

Stocktitan states that the $225 million transaction constitutes the largest US edge data center deal in 2026 to date, outpacing other private deals this quarter. Realassets underscores the sharp escalation in private capital chasing digital infrastructure. Broader forecasts suggesting global deployments will exceed current annualized investment rates by a wide margin as early as next year. Cogent Fiber’s asset sale marks a fundamental pivot as they exit non-core US holdings and return focus to international fiber builds, freeing up capital that can be redeployed outside the capital-intensive US retrofit cycle. At least $1 billion is now earmarked for follow-on upgrades in the next two years as part of I Squared’s AI infrastructure program.

Big Tech firms are set to spend over $600 billion on data centers this year, driven by AI investment, as pressure mounts to show profitability pic.twitter.com/C2js1m2pX8

— Reuters (@Reuters) April 20, 2026

Detail	Information
$225 million transaction	Largest 2026 US edge data center deal so far, per Stocktitan.
Growing investment forecast	Private capital in digital infrastructure is rising worldwide, according to Realassets.
Cogent Fiber exit	Pivots out of non-core US data centers, focusing on overseas fiber operations.
I Squared follow-on	$1 billion capital allocated for 2026–2027 AI infrastructure, driving sector expectations.

---

Sign up for our newsletters

Channelnewsasia reports that I Squared plans quarterly digital briefings and newsletters to update stakeholders on site retrofits and AI deployment milestones.

• Quarterly updates:Newsletters will track every pod deployment, energy upgrade, and major migration by metro area.
• Site dashboards:Stakeholders get real-time metrics on power, cooling, and network performance at each upgraded center.
• Outage alerts:Automated notifications detail downtime incidents and resilience checks, now presented as standard practice.
• ESG benchmarking:Regular disclosures will summarize renewable energy integration and power use for investors.

---

Get the CNA app

The Channelnewsasia app adds real-time coverage for infrastructure deals like I Squared’s rollout, per Channelnewsasia.

• Instant deal alerts:Users receive up-to-minute notifications as I Squared’s buildout moves through stages.
• Interactive infrastructure maps:Visual overlays show which metros and nodes are powering the AI wave.
• Research insights:Subscribers get concise executive summaries of new peer infrastructure deals and regulatory shifts.

---

Data Center Investment Timeline

The sequence of I Squared’s portfolio buildout reflects shifting priorities and accelerated timelines for AI infrastructure, per Realassets. In February 2026, the company started scouting secondary-market sites suitable for GPU retrofits across the US. By April, Cogent Fiber agreed to sell ten qualifying data centers for $225 million, clearing the path for a May closing and launch of upgrade planning. Stocktitan notes that by Q3 2026, at least one major site will be announced for immediate AI retrofit, with phase one deployments across multiple cities mapped for Q1–Q3 2027. By the end of 2027, the intention is to allocate the full $1 billion—turning every acquired site into a production hub for AI inference workloads.

1. February 2026:I Squared evaluates US secondary-market data centers for AI upgrades.
2. April 2026:Deal is struck with Cogent Fiber to buy 10 sites for $225 million, according to Realassets.
3. May 2026:Acquisition closes; initial engineering for power and connectivity begins.
4. Q3 2026:Major site selected to be first for high-density AI pod deployment, per Stocktitan.
5. Q1–Q3 2027:Phase 1 of major retrofits brings substantial new capacity online.
6. End of 2027:Completion target for $1 billion spend with all ten sites operational as AI inference centers.

AI Infrastructure Funding: The Broader Picture

Channelnewsasia notes that stiff competition for operational data centers and technical talent in US metro hubs is creating labor shortages. Market rates for engineers with advanced power and AI experience have hit new highs. Meanwhile, cloud-based AI SaaS models funneled tens of billions of dollars in revenues to US providers over the previous quarter, driving rapid year-on-year growth.

Cogent Fiber’s Strategic Pivot

Per Realassets, Cogent Fiber’s decision to sell its 10-site US data center portfolio is a calculated move to refocus on building international fiber and wholesale connectivity, especially across Latin America and Asia-Pacific where bandwidth needs are growing far faster than in mature markets. The $225 million proceeds will be deployed upgrading long-haul optical infrastructure in those regions, letting the company capitalize on surging data traffic and avoid the high CapEx needed for US AI retrofits.

Selling Entity	Proceeds Use	Refocus Area
Cogent Fiber	Network upgrades	LatAm, Asia-Pacific fiber
Peer operators	Debt reduction, new subsea cables	Divesting non-AI US assets

AI Inference Platforms and Industry Implications

Stocktitan explains that I Squared frames its effort as building an “AI inference platform”—a combination of hardware and software layers optimized to deploy trained AI models in real time. The inference stage of AI means moving from massive, energy-consuming training to running models in production, servicing queries in real applications such as chatbots, self-driving vehicles, or trading algorithms.

Unlike traditional leasing, I Squared’s platform will let customers buy AI inference on demand, backed by provisions for liquid cooling and modular power delivered across all retrofitted racks.

• On-demand service:Monetize AI inference beyond property rental, targeting modeled workloads.
• Technical upgrades:Expand liquid cooling and modular power throughout every pod.
• Diversified contracts:Combine long-term deals with agile terms for new AI deployers.
• Instant onboarding:Pilot schemes for sectors like healthcare, tuned for privacy and speed as differentiators.

Competitive Landscape and Market Directions

Realassets reports that I Squared faces mounting competition as domestic and global investors launch rival AI infrastructure funds. Two new peer capital pools exceeding hundreds of millions committed within April and May 2026 alone. Traditional cloud data centers are no longer sufficient for the inference end of the AI boom—optimizations for training do not translate directly, pushing clients to seek new specialized hosts.

According to Channelnewsasia, data center investors are developing utility partnerships to address grid constraints as each AI deployment wave demands more power and cooling.

$1B — I Squared’s US infrastructure buildout target.

Key Takeaways: The Future of AI Infrastructure Investment

• $225M I Squared buy:Single largest US edge portfolio sale in 2026, laying the foundation for an AI-centric buildout.
• 10-site national spread:Atlanta, Dallas, Chicago, Phoenix offer key locations for distributed inference clusters.
• Accelerated retrofits:Market moves rapidly toward quick upgrades for high-density, GPU-ready AI hosting.
• Private equity inflection:2026 marks sharp growth for digital infrastructure; sector capital is rising fast.
• Cogent Fiber strategy:Divestment funds shift to international long-haul and wholesale fiber, underscoring a global pivot.
• Model host demand:Cloud and SaaS firms lead the secondary-market for inference-optimized centers in 2026–2027.
• Utility-platform turn:I Squared offers inference as a utility, not just square footage—tailoring contracts and onboarding for rising demand.
• Enterprise priorities:Buyers stress power reliability, sustainability, thermal management, and real-time uptime metrics.
• Competitive escalation:Since April 2026, rival funds launched substantial AI infrastructure pools, increasing sector pressure.

For Deeper Coverage

Coverage of I Squared Capital’s expansion into US AI infrastructure continues as site retrofits accelerate and new tenants commit, according to Stocktitan and Channelnewsasia. Readers tracking Cogent Fiber’s asset divestment and its redeployment toward high-growth international networks can find focused analysis and future updates in More in-depth I Squared Capital buys articles. To inquire about coverage details, connect with editors directly via our site. The most significant milestones still ahead include first AI inference tenant deployments and achievement of the full $1 billion US buildout.