Few investment themes have gripped the purse strings of private capital in recent years quite like the burgeoning AI revolution. When ChatGPT launched in 2022, it quickly became the fastest-growing consumer product in history, attracting 100 million users within two months. In the years following, the pace of computing power to support AI has followed a Moore’s Law trajectory.
Importantly, behind the generation of these AI-driven workloads lies a critical foundation — the data centers that run them. And fueling those data centers? Vast amounts of power. This has led to an emerging symbiosis between two of the most transformative investment themes shaping global capital formation — energy and digitalization.
As the pace of data generation accelerates, energy has become both the lifeblood and the bottleneck. The soaring power demands of AI-driven computing are straining existing energy infrastructure, prompting a wave of investment in new power generation, grid modernization, and technologies aimed at ensuring greater energy reliability.
From data centers to microgrids and battery storage, the underlying infrastructure underpinning the AI boom and the energy to power it require scaled capital and deep sector expertise, both of which infrastructure investors have increasingly stepped in to provide. However, as investors increasingly pursue data center and energy investment opportunities, they often embrace risk-return profiles that, to some degree, diverge from the conventional infrastructure playbook.
Bloomberg projects the generative AI market will reach $1.3 trillion over the next decade. As the adoption of cloud services, 5G technology, and internet-connected tools accelerates alongside this growth, demand for the physical infrastructure, such as towers, data centers, and fiber networks to support this rising tide of data generation has soared in tandem.
Among these assets, data centers have commanded the lion’s share of investor attention. In 2024, over $108 billion of private capital flowed into data center deals, and the momentum shows no signs of slowing. By 2030, data centers are projected to require $6.7 trillion worldwide to keep pace with the demand for compute power.
For infrastructure investors, data centers provide exposure to the powerful tailwinds of the AI boom, while retaining key infrastructure-like traits. These assets provide an essential service in the form of data processing, data storage and computational services that now prove essential to daily life and global economic growth. Land and power constraints, coupled with complex permitting, create scarcity value. Plus, these assets provide predictable revenues from medium- to long-term contracts.
As infrastructure investors eye data center opportunities, they can employ strategies with varying risk and return profiles. The most prominent approaches include acquiring stakes in existing operating assets, platform building, and developing customized facilities designed to meet the expanding needs of hyperscalers.
Some private equity firms are acquiring majority or minority stakes in established “data center platforms” with robust development pipelines and strategic landbanks, accessing the dual advantage of immediate exposure to stable, recurring cash flows from operational assets, alongside the growth potential of future developments. Once ‘stabilized,’ these platforms often sell their development assets to free up capital and recycle it into new projects. Diversification from assets operating across different geographies and end markets can offer additional advantages.
A prime example of this strategy includes Blackstone’s $16 billion acquisition of Airtrunk, the Asia-Pacific’s largest data center developer. In an interview with PEI, Blackstone’s Global Head of Infrastructure, Sean Klimczak, noted that the deal was largely underpinned by the operator’s 800 megawatts of fully contracted capacity and stable cash flows, with a moderate portion ascribed to its forward-looking development pipeline.
Broadly, one criticism of data center platform building is that rising investor interest has pushed valuations into frothy, potentially overinflated territory. In recent years, acquisition multiples for these platforms have ranged from 25x to 30x EBITDA — significantly above the broader infrastructure market average of around 16x. The premium reflects strong investor appetite for platforms with secured land and power access.
Historically, the largest tech giants or ‘hyperscalers’ built and operated their own data centers to keep up with rising digital demand. However, the next generation of data centers needed to support expanding AI workloads demands far greater scale and technical complexity than traditional cloud facilities — often spanning 150,000 to over a million square feet and consuming up to 100 MW of energy.
Faced with these mounting demands, build-to-suit partnerships with third-party developers have become desirable for these companies. Global demand for hyperscale data centers is projected to reach $593 billion by 2030, with a compound annual growth rate of 28.4%. In response, greenfield development of these bespoke facilities has emerged as one of the fastest-growing strategies in private infrastructure.
Several qualities make hyperscale data centers attractive from an investment perspective. These facilities are typically fully tailored to the precise needs of the end user. As a result, tech companies often commit to long-term leases — usually 10 to 15 years — before construction begins. That upfront commitment gives investors forward visibility into long-dated, recurring cash flows from high-quality credit parties. Further, with leasing risk largely off the table, investors can generally secure more favorable financing terms for development.
Concentration risk from single tenant occupancy surfaces as the most obvious risk with hyperscale development. If a highly specialized data center is located in a remote area and the tenant doesn’t renew at the end of the lease, backfilling that space could prove challenging. However, leasing risk remains minimal for now, as churn across the sector is extremely low. Concerns over technological obsolescence are also eased by the fact that tenants typically bear responsibility for their own IT costs.
Speculative data centers offer another greenfield avenue, and some investors are comfortable starting construction without an end tenant secured, particularly given the sharp supply-demand imbalance in many markets and potential for higher market rental rates that drive up returns. For example, build-to-suit deals for hyperscalers typically generate returns in the high single digits, whereas speculative developments can sometimes push upward of 20%.
Regardless of approach, greenfield data center development remains a complex process. Developers must secure suitable land and power, navigate entitlements, obtain permits, and manage regulatory challenges. The construction itself is equally demanding, requiring advanced wiring, cooling systems, and built-in power redundancy systems. Supply chain disruptions have also become a growing concern, adding further considerations to project timelines.
Investors can also purchase and retrofit existing facilities (often into more modern colocation centers). In some tier one or mature markets, where land is limited, power is close to capacity, and regulatory hurdles are high for new development, these conversions can allow developers to repurpose existing facilities to more quickly service demand.
Most often, these colocation centers have a larger tenant base, but on shorter leases that range from five to seven years. While this provides a more diversified revenue stream and often higher return potential, it necessitates consistent leasing efforts to maintain occupancy and poses more risk to tech obsolescence.
All is to say — across a data center portfolio, investors face a wide range of underlying value drivers that influence the risk-return profiles of these assets in distinct ways. As noted, leasing structures vary widely. Some assets offer long-term contracted revenue, while others rely on shorter-term agreements.
Many platforms operate across multiple geographies, blending stabilized assets that generate recurring cash flow with development-stage projects that pose execution risks. Additionally, assets within a single platform can often operate under distinct reporting structures, making it challenging to form a cohesive, high-level view of performance and risk.
Adding to the intricacy, data center unit economics are a function of energy costs, cooling efficiency, and other dynamic variables that demand granular KPI tracking to monitor profitability. In Los Angeles, for example, data centers on opposite sides of the same street have faced different electricity rates because they are served by different utilities.
As data center construction ramps up worldwide to meet surging demand, electricity capacity needs will accelerate significantly. Historically, electricity demand has grown about 0.5 percent annually. This flat demand curve has given utilities the foresight to bring in new power generators and make other investments as needed, with a high degree of predictability. With that curve now forecasted to grow at 3 to 4 percent through 2040, many utilities are unprepared for the step change in load demands.
While not the sole contributor, the rising energy intensity of AI workloads is one driver. Compared to a standard Google search, for example, a single ChatGPT query can consume up to 36 times more power. Image-generation tools like DALLE require roughly 50 times more energy, while video models such as Sora can use up to 10,000 times as much.
Therefore, the pressing question now is how to build the capacity needed to power the growing wave of energy-intensive data centers already online as well as the many more in development. Meeting this demand will require sweeping investment across the entire energy value chain, from modernizing the grid and expanding energy storage to accelerating the deployment of new power sources. For energy infrastructure investors, it represents a massive opportunity across multiple access points.
Grid updates and expanded power supply are one avenue. According to Goldman Sachs Research, an estimated $720 billion in grid upgrades may be needed by 2030 to accommodate rising electricity loads. Yet, the timeline for these projects pose a critical challenge. Creating new transmission projects to accommodate data centers, for example, often take years to secure permits, followed by several more to construct.
Compounding the issue is the growing backlog in the “interconnection queue,” where both new power users — like data centers — and new sources of generation, — such as wind and solar — await approval to join the grid. In many regions, this holding period has stretched to five to eight years. In contrast to the 18 to 24 months typically required to develop a data center, the disparity highlights how access to power has become a major bottleneck in bringing new facilities and their cash flows online.
Ultimately, while adding new power sources to the grid and upgrading existing infrastructure will be essential to meeting rising demand, the urgency of the AI buildout has also sparked a range of faster workarounds. With speed a top priority, some data center developers are turning to alternative solutions that allow them to secure energy more quickly. In this context, the faster deployment timelines and lower costs of renewable energy present a compelling option for bringing data centers online faster.
For example, co-locating a data center with an energy source situated directly on-site or nearby has gained appeal. Google, Intersect Power, and TPG Rise’s recent $20 billion initiative to create “energy parks” — massive co-located solar power and battery installations for powering new and existing data centers — is a prime example. This “behind-the-meter” setup allows power to be supplied straight to the data center without passing through the broader utility grid, avoiding lengthy interconnection timelines and enabling faster deployment.
Broadly, as pressure mounts to bring data centers online faster, a growing set of alternative power solutions is gaining momentum. Microgrids and small modular nuclear reactors are emerging as promising options, offering localized, scalable energy sources that reduce reliance on traditional grid infrastructure and accelerate access to reliable power.
Ultimately, the data center boom marks a structural shift in how energy infrastructure is financed, developed, and managed. Private investors with expertise in renewables, nuclear energy, storage, and grid systems are well-positioned to seize emerging opportunities. Still, much like digital infrastructure strategies, investing across the energy value chain starts to flex the traditional view of infrastructure investing.
Depending on the asset, investors may encounter varying degrees of exposure to revenue stability, market competition, shifting power market dynamics, and other operational risks. Each factor plays a critical role in shaping the investment profile and long-term performance of energy infrastructure assets.
Get updates in your inbox
Learn how Chronograph can streamline your private capital investment monitoring and diligence