How data centers and the energy sector can sate AI’s hunger for power

| Article

Surging adoption of digitalization and AI technologies has amplified the demand for data centers across the United States. To keep pace with the current rate of adoption, the power needs of data centers are expected to grow to about three times higher than current capacity by the end of the decade, going from between 3 and 4 percent of total US power demand today to between 11 and 12 percent in 2030.1 Skyrocketing compute and data demands are being further accelerated by gains in computing capabilities alongside reductions in chip efficiency relative to power consumption. For instance, the amount of time central processing units need to double their performance efficiency has increased from every two years to nearly every three years. And providing the more than 50 gigawatts (GW) of additional data center capacity needed in the United States by the end of the decade would require an investment of more than $500 billion in data center infrastructure alone.2

The power sector is rapidly becoming a protagonist in the AI story. Access to power has become a critical factor in driving new data center builds. As the power ecosystem grapples with meeting data centers’ voracious need for power, it faces substantial constraints, including limitations on reliable power sources, sustainability of power, upstream infrastructure for power access, power equipment within data centers, and electrical trade workers to build out facilities and infrastructure. Currently, for example, the lead time to power new data centers in large markets such as Northern Virginia can be more than three years. And, in some cases, lead times for electrical equipment are two years or more.

Without ample investments in data centers and power infrastructure, the potential of AI will not be fully realized. This article addresses this rapidly evolving space: the prospective growth of AI and demand for data centers, the challenges to scaling data centers, and how investors and incumbents could realize significant gains while helping fulfill AI’s potential.

Register for our upcoming webcast Powering AI: Opportunities in data centers

Join the authors for a live webcast on Nov. 21 at 10 p.m. NY / 3 p.m. LON

The scope of escalating demand for data centers

According to McKinsey analysis, the United States is expected to be the fastest-growing market for data centers, growing from 25 GW of demand in 2024 to more than 80 GW of demand in 2030. This growth is fueled by the continued increase in data, compute and connectivity from digitalization, and cloud migration, as well as the scaling of new technologies—the most important of which is AI. McKinsey research estimates that generative AI (gen AI) could help create between $2.6 trillion and $4.4 trillion in economic value throughout the global economy.3The economic potential of generative AI: The next productivity frontier, McKinsey, June 14, 2023. But achieving just a quarter of this potential by the end of the decade would require between 50 and 60 GW of additional data center infrastructure in the United States alone.

Meeting this demand will require considerably more electricity than is currently produced in the United States. This spike in electricity needs is unprecedented in the United States, where power demand in the aggregate has barely grown since 2007.4 Data center load may make up between 30 and 40 percent of all net new demand added until 2030, with demand growth arising from domestic manufacturing, electric vehicles, and electrolyzers (see sidebar “What makes data center load unique?”). Between 2024 and 2030, electricity demand for data centers in the United States is expected to increase by about 400 terawatt-hours at a CAGR of about 23 percent (Exhibit 1).

1
Demand for power for data centers is expected to rise significantly in the United States.

As demand for data centers climbs, the implications for companies in the power value chain become more apparent.

Constraints across the power value chain may hinder progress

Companies across the power value chain are contending with constraints and shortages, bottlenecking progress. The industry is approaching its physical limits on node sizes and transistor densities, and long lead times have also hindered progress. The time required to get new power connections for data center sites in major data center hubs such as Northern Virginia; Santa Clara, California; and Phoenix has been increasing. Locations outside of the United States, such as Amsterdam, Dublin, and Singapore, have placed moratoriums on many new data center builds in recent years primarily because they lack the power infrastructure to support them.

At the same time, the demand from new AI loads has contributed to a shortage of compute availability. Vacancy rates in large tier-one data center markets are at historic lows, hampering interconnection and grid access, especially in highly penetrated markets in which data center load already represents a substantial portion of the overall demand. In Northern Virginia, for example, vacancy rates were less than 1 percent in 2023.5 McKinsey research shows that time to power is the biggest consideration for data center operators when building new sites. Adding to investor tension, as access to grids has declined, timelines for investing in and further building out grids for regulated utilities have become longer than the development cycle of data centers.

Notably, power unavailability in most markets is driven by limitations in interconnecting to the transmission grid, rather than an inability to generate the power. However, latent capacity in the generation fleet is largely held by fossil-fueled plants operating below their maximum levels. While hyperscalers and utilities work to build out the renewable fleet to support sustainability commitments, there is a continuing need to supply not only new load from data centers but also the growing load from electrification (in transport and industry, for example) as well as backfill for aging thermal plants that will eventually retire. As a result, sustainability commitments are, in some instances, taking a back seat to maintaining operations.

In locations with access to power on the bulk transmission grid, there are further constraints on power equipment, such as transformers, on-site backup generators, and power distribution units (PDUs), with historically high lead times of nearly two years in some cases (Exhibit 2).

2
Accelerated demand and supply chain constraints have increased lead times for equipment, resulting in project delays.

The strained labor force is an additional inhibitor, particularly the emerging shortage of electrical trade workers essential to executing these projects. McKinsey estimates anticipate a potential shortage of up to 400,000 trade workers in the United States based on projected data center build-out and comparable assets requiring similar skills, such as semiconductor fabrication and battery gigafactories.6Generative AI and the future of work in America,” McKinsey Global Institute, July 26, 2023.

Additionally, the industry faces the daunting challenge of decarbonizing its footprint to achieve the goal of 24/7 carbon-free energy usage by 2030. While the carbon emission intensity for power grids is set to drop in the next ten years,7 generation from natural gas nationally is expected to increase (Exhibit 3). At the same time, most grid decarbonization timelines (if they exist) far exceed the targets set by major hyperscalers.

3
The intensity of carbon emissions from grid power is set to drop rapidly in the next ten years but is still far from hyperscale’s clean-power target.

As states, local jurisdictions, and power companies set and work toward their sustainability goals, major corporate customers will also have to demonstrate that they are achieving their self-designed targets. Many public commitments exist, but there is no accepted standard for achieving clean power across the industry. And because grids are synchronized across large geographies, tracking local and time-matched generation and consumption across the shared system becomes a complex accounting activity. As a result, these goals are difficult to measure and often difficult to achieve. Companies across the sector have used many different instruments and approaches to manage their carbon accounting, including unbundled renewable-energy certificates, power purchasing agreements (PPAs), time-matched renewable-energy certifications, carbon matching, offsets, and accreditation activities. Low-carbon power will become an even more important area of investment, but for now, many stakeholders are left to define their own motivations, ambitions, and directions for the future.

Untapped opportunities for investors and incumbents in data center growth

For the data center ecosystem, the tremendous capital deployment and close linkage with the power sector present a significant opportunity.

Across the power value chain, investors can participate in and enable solutions to meet the demand for data centers and accelerate growth. Current progress and limitations alike illuminate three clear areas in which investors may be able to make the most impact: power access and sources, power equipment, and trades and technicians.

Across the power value chain, investors can participate in and enable solutions to meet the demand for data centers and accelerate growth.

Power access and sources

Investors and incumbents have ample opportunity to be involved in creating new solutions for power access and sources. Four areas have the most potential:

Transmission and distribution investments. Because of the rising criticality of power availability in scaling data centers, more utility companies have realized the importance of and potential in data centers—21 utility providers mentioned data centers in their fourth quarter 2023 earnings calls compared with just three providers in 2021. As a result, investors can funnel investments into utility companies to build out transmission and distribution (T&D) infrastructure in key markets. The demand for data centers and power shows no sign of slowing, so T&D markets should grow accordingly. Advances in gen AI will create even more data, increasing the need for data storage centers to avoid issues that come with managing large quantities of data. Investments in T&D infrastructure will allow for better compute and storage systems.

Secondary markets with access to reliable, cheap power. A natural tension exists between the timelines of data center builds (which can reach 18 to 24 months) and those of power infrastructure development—gas and renewables projects typically stretch three to five years, and transmission development can regularly take seven to ten years. This tension offers opportunities to bridge the timing gap creatively. Many hyperscalers are building out capacity in new and atypical locations outside the core data center markets because these areas offer cheaper, available power and have the potential for carbon-free infrastructure. Iowa, Wyoming, Indiana, and Ohio, for example, each houses, or has received investment from, at least two of the top four hyperscalers. Much of data center growth—about 70 percent—is expected to be fulfilled directly or indirectly (via cloud services, for instance) by hyperscalers by 2030, so several of these emerging locations are self-built by hyperscalers or through built-to-suit co-location providers that are helping hyperscalers accelerate time to market.

Investors can seek opportunities to fuel growth in these burgeoning markets by investing in data center developers at the company or site level. They can also invest across the value chain by, for example, accelerating the build-out of fiber or power infrastructure in these secondary locations. With growing use cases in training large models, latency across the fiber network becomes somewhat less important, making expansion to additional geographies even more attractive. Data centers in these locations will still need significant data throughput and may be less flexible in the services they provide in the future, but speed to market and attractive economics will, in many cases, outweigh the more specialized nature of the sites. The opportunity for large-scale power projects increases in markets that are scaling with demand for new data centers because it will likely become easier to find contracted offtake, which enables most projects (Exhibit 4).

4
As power transmission becomes constrained in primary markets, leading players are moving to secondary and emerging markets.

Behind-the-meter solutions. Investors can seek to support behind-the-meter solutions to provide power in areas where utilities providers cannot keep up with pace or reliability requirements as local supply availability or transmission constraints worsen. For example, investors could seek opportunities to build power that can be fully islanded outside the grid, retrofit existing sites or facilities to densify and bring additional capacity, or provide supplemental power to complement the grid (see sidebar “The opportunity for nuclear in data center build-out”). The sites available for these opportunities are limited, but creating more competition and urgency among investors to act sooner than later can help secure the talent, connectivity, and regulatory requirements necessary to run the sites.

Sustainability ambitions driven by renewable-energy providers. With hundreds of terawatt-hours of additional energy demand expected in the coming years, there will be a commensurate demand for clean power for hyperscalers and other large players that have put forth leading climate commitments. This demand should arrive incrementally to state- and utility-level targets, in addition to the numerous corporate commitments outside the data center space. As a result, the desire for additional renewable projects is expected to continue and may help sustain the high PPA rates seen recently.8 While many technologies—offshore wind, fission, fusion, geothermal, gas carbon capture and storage, and clean fuels, for example—may be able to supply this energy in the medium to long term, the bulk of new clean generation is expected to come from solar and onshore wind. Investment in this space has a long track record, but this track record does include some mixed returns in the past. However, emerging opportunities in the domestic supply chain, project sponsorship, and increasing operations and maintenance needs are arising in a different environment and may present new risk profiles and attractive returns.

Power equipment

Creating new power grids naturally relies on equipment supply, which offers opportunities to invest in new or emerging technologies. Shortages of critical equipment have allowed for growth among smaller companies focused on creating generators and among hyperscale-focused providers creating PDUs.9Investing in the rising data center economy,” McKinsey, January 17, 2023. Investors can seek out the smaller companies creating this critical equipment to help them scale.

What’s more, rapidly increasing rack power densities (going upward of 50 to 100 kilowatts per rack) are leading to higher power ratings for equipment across transformers, switchgear, and PDUs, necessitating changes and innovation in product lines and creating opportunity for new entrants. In the same vein, specializing in modularization and the prefabrication technologies of, for instance, mechanical, engineering, and plumbing packages could help hyperscalers build new data centers faster.

Trades and technicians

Given the pace of growth in the data center and associated power infrastructure markets, the gap in the supply of trained talent for electrical and mechanical installation work is widening rapidly. Large specialist contractors, historically relied upon for such projects, are increasingly partnering with regional providers to augment their talent supply. And because more data centers are being built in locations with limited availability of talent experienced in data center infrastructure, contractors are moving talent to installation locations.

This growth in demand can provide a few areas of opportunity for companies in this space and investors looking to increase exposure. First, there is potential to consolidate and scale multiple subscale contractors in regions witnessing high growth in demand. These companies typically lack the scale, capital, and relationships to independently execute large projects, but they may have a secure supply of available talent. Second, scaled companies can invest in increasing their share of off-site manufacturing and assembly, reducing the need for on-site talent and accelerating project execution, likely at a lower cost that could enable margin expansion. Last, reimagining talent sourcing and training is imperative to address shortages, particularly for severely constrained roles such as technical supervisors. The availability of a reliable, qualified labor force will likely be a key differentiator for players in this space.


Gen AI has heightened the need to accelerate the pace and scale of power more than any other technology in the past two decades. The growth of data centers has quickly exposed gaps in both power and data center infrastructure where investment is needed to ensure reliability and affordability across the entire system. At the same time, additional power will be needed to meet clean-power objectives. Amid these circumstances are numerous opportunities to invest in the next generation of power supply and data storage, an area in which owners have a higher willingness to pay. Investors and incumbents that can carve out niches in this power-hungry space can fuel the growth of gen AI and position themselves favorably in the new generation of power technology.

Explore a career with us