Carlyle is preparing to exit a major slice of the infrastructure that sits behind the AI boom, agreeing to sell a data-centre power business valued at about $2.6 billion to EQT in a transaction expected to generate a roughly fivefold return for the private equity firm. While the headline is about price and profit, the deeper story is about what kind of assets are attracting capital right now—and why “power” has become one of the most strategic bottlenecks in the modern data economy.
For years, investors treated data centres as a relatively straightforward play: buy land or acquire operating capacity, build or expand, lease space, and collect steady cash flows. But the AI era has changed the economics of expansion. Training and inference workloads are driving demand for compute at a pace that is increasingly constrained not by servers or even by chips, but by electricity. That shift has elevated grid access, substation capacity, power delivery systems, and related engineering capabilities into the category of “must-have” infrastructure rather than optional support services.
In that context, Carlyle’s sale to EQT reads less like a routine portfolio exit and more like a signal. Private equity is still under pressure in many corners—fundraising can be difficult, leverage is more expensive, and exits can be slow when buyers are cautious. Yet the market is simultaneously rewarding a narrower set of themes: assets that solve real operational constraints for high-growth industries. Data-centre power is one of those themes, and it is drawing attention from both infrastructure specialists and broader buyout players looking for durable demand.
The deal also underscores a particular kind of investor logic that has been gaining traction: when an asset is tied to a structural bottleneck, the valuation case becomes less dependent on short-term sentiment and more dependent on long-term capacity needs. In other words, the buyer is not simply paying for today’s cash flows; it is paying for the ability to keep delivering power as demand rises and as regulatory and physical constraints limit how quickly new capacity can be brought online.
Why power became the bottleneck
To understand why a power business can command such interest, it helps to look at what happens when data-centre operators try to scale. Even when there is land available and construction capacity exists, the limiting factor often becomes the connection to the grid. Utilities may require long lead times for upgrades. Permitting can take time. Substations and transmission lines are expensive and politically sensitive. And once a site is connected, the question becomes whether it can reliably deliver the required load without interruptions or costly curtailment.
AI workloads intensify all of these issues. The power draw of modern training clusters is enormous, and the pace of deployment means that delays in grid readiness can translate into lost revenue, missed customer commitments, and higher costs elsewhere in the supply chain. As a result, data-centre developers and operators have increasingly sought partners who can help them secure power capacity, manage interconnection processes, and build the electrical infrastructure that turns grid availability into usable capacity.
This is where “data centre power” businesses fit. They sit at the intersection of energy infrastructure and digital infrastructure. They are not just about electricity as a commodity; they are about electricity as a delivered capability—engineered, permitted, and integrated into sites that must perform continuously.
A fivefold return tells a story about timing
Carlyle’s expected fivefold return is a reminder that private equity returns are often driven by entry timing as much as by operational improvements. Buying into power-related infrastructure earlier in the cycle—before the AI-driven surge made electricity constraints impossible to ignore—can create a compounding effect. As demand accelerates, the value of capacity and the scarcity premium attached to reliable power tends to rise.
But the return also suggests that Carlyle likely built or enhanced capabilities beyond simply owning assets. Power businesses in this space typically require technical expertise, relationships with utilities, and the ability to navigate complex permitting and engineering requirements. If Carlyle developed a platform that could execute expansions, secure connections, or improve the reliability and scalability of power delivery, then the company would become more valuable precisely when customers start prioritising speed and certainty.
EQT’s willingness to pay for the business indicates that the buyer believes the demand curve will remain steep. It also implies confidence that the regulatory and operational challenges involved in scaling power infrastructure can be managed at scale—an important point because many infrastructure plays fail not due to demand, but due to execution risk.
What EQT is buying: more than electrons
Although the transaction is described as a data-centre power unit, the strategic value is broader than the physical flow of electricity. In practice, these businesses often provide a bundle of capabilities that include:
1) Grid and interconnection know-how
Securing power is not only a matter of building equipment; it is also a matter of navigating utility processes, timelines, and technical requirements. A platform that has already done this work can reduce uncertainty for future projects.
2) Engineering and integration capability
Data centres require electrical systems that meet strict reliability standards. That includes redundancy, power conditioning, and the ability to integrate with the site’s overall architecture. The ability to deliver these systems efficiently becomes a competitive advantage.
3) Capacity planning and commercial structuring
Power infrastructure is capital intensive. Investors care about how capacity is contracted, how costs are allocated, and how revenue scales with demand. A business that can structure deals in a way that aligns incentives with customers can capture more value over time.
4) Operational reliability
In a world where downtime can be extremely costly, reliability is not a “nice to have.” It is part of the product. Power businesses that can demonstrate performance and resilience tend to be preferred partners.
When buyers pay for a power business, they are effectively paying for the ability to convert grid constraints into predictable delivery. That conversion is what makes the asset valuable during periods of rapid demand growth.
Why this is a bright spot for private equity exits
The private equity market has been uneven. Many funds are sitting on assets that are harder to sell because valuations have become more sensitive to interest rates, and because buyers have become more selective. In some sectors, demand has cooled or competition has increased. In others, the challenge is that the path to growth requires too much additional capital or too much time.
Yet the AI infrastructure theme has created a pocket of relative clarity. Data centres are being built and expanded, and the power component is not easily substituted. Even if compute supply improves, electricity remains a gating factor. That creates a kind of “structural urgency” that can support pricing and deal momentum.
This is why the Carlyle-to-EQT transaction stands out. It is not merely an exit; it is an example of how private equity can monetise a theme that is both capital intensive and operationally constrained. When the constraint is real, buyers are willing to move quickly and pay for certainty.
The unique angle: infrastructure as a service, not just an asset
One reason data-centre power has become so investable is that it increasingly resembles a service layer. Customers do not want to buy a pile of equipment; they want a dependable outcome: power delivered at the right capacity, with the right reliability, and with a timeline that matches their deployment plans.
That shift changes how investors evaluate the business. Instead of focusing solely on asset appreciation, they look at the platform’s ability to win repeat demand, expand capacity, and maintain customer relationships. It also affects how the business can be scaled after acquisition. If EQT can extend the platform’s reach—through new builds, partnerships, or geographic expansion—the investment case becomes more compelling.
In this sense, the deal reflects a broader evolution in infrastructure investing. The most valuable infrastructure platforms are those that can package complexity into repeatable execution. Power is complex. Grid interconnection is complex. Permitting is complex. But if a platform can make those complexities manageable, it becomes a scalable engine rather than a one-off project.
What could happen next for the sector
While the transaction itself is about Carlyle exiting and EQT entering, it will likely influence how other investors think about the AI infrastructure stack.
First, it reinforces that “energy infrastructure” is not a peripheral theme anymore. It is central to the AI build-out. That may encourage more capital to flow into power delivery, grid services, and related engineering capabilities.
Second, it may increase competition among buyers for similar assets. If one large deal demonstrates strong returns, other funds will look for comparable opportunities—especially those with proven execution track records.
Third, it could accelerate consolidation. Platforms that can deliver power at scale may become targets for roll-ups, particularly if fragmentation in the market makes it hard for customers to find end-to-end solutions.
Finally, it may push data-centre operators to formalise longer-term partnerships. When power is the bottleneck, customers want to lock in capacity and timelines. That can lead to more structured contracting, potentially improving revenue visibility for power providers.
The risks are real, but they are different
No infrastructure investment is without risk, and power-related businesses carry their own set of challenges. Execution risk remains significant: grid upgrades can be delayed, engineering projects can run over budget, and regulatory approvals can take longer than expected. There is also the risk that demand projections could soften if AI deployment slows or if customers change their build strategies.
However, the nature of the risk differs from sectors where demand is more discretionary. Electricity demand from data centres is tied to ongoing compute needs, and the AI build-out has been persistent enough to keep pressure on capacity. Even if growth rates fluctuate, the underlying trend toward higher compute intensity suggests that power constraints will remain relevant.
Another risk is political and regulatory. Energy infrastructure is inherently public-facing. Permitting, environmental considerations, and community engagement can affect timelines and costs. Investors will need to manage these factors carefully, especially if expansion requires new transmission or substation capacity.
Still, the fact that EQT is stepping in suggests that the buyer believes these risks are manageable and that the business has the capabilities to navigate them.
Why the market is rewarding “scarcity”
At the heart of this deal is a simple economic principle: scarcity drives value. In