Generational Shift—Data Centers Bring Change to Energy Landscape

Deals involving the data center sector are occurring regularly, as tech companies make moves to ensure they have a reliable supply of power for their energy-intensive operations. Several generation types, both thermal and renewable, are being touted and pursued by groups involved with artificial intelligence (AI), data management, and cloud computing. Google, Microsoft, Amazon Web Services, Meta, and others are looking at nuclear power, including small modular reactors (SMRs), along with renewable energy resources. Some companies are signing power purchase agreements (PPAs) with operators of natural gas–fired power plants, and analysts say demand for power from data centers could keep some coal-fired power plants online past their scheduled retirement dates. POWER has talked with several energy industry leaders over the past several months about the ongoing and future impacts of electricity demand from AI and data centers, and much of their insight can be found on POWER's website. An important piece of the puzzle for those in the power generation sector is what fuels and technologies will be used to generate that power, and how utilities and grid operators will manage resources to maintain a reliable and resilient power grid in the face of an exponential increase in electricity demand.

Using AI to Support AI

Supratik Chaudhuri, Power and Utilities Lead of Publicis Sapient, a digital transformation consulting partner and software engineering group, told POWER that power grids can be managed more efficiently through the use of AI, which also could help mitigate issues caused by increased demand from data centers. “The integration of AI into power grid management represents a transformative leap forward. AI can not only make decisions in specific scenarios [such as demand-side management when supply is lower], but it also empowers decision-makers with insights that enable quicker responses,” said Chaudhuri. “This capability is particularly beneficial as future energy supplies become more volatile with an increasing share of renewables in the generation mix. AI’s ability to analyze vast amounts of data from sources such as sensors and smart meters allows for more precise demand forecasting and load balancing. This enables utilities to predict and respond to energy consumption patterns more effectively, reducing the risk of outages and enhancing grid stability.” Chaudhuri continued, “Additionally, AI-driven predictive maintenance can identify potential equipment failures before they occur, minimizing downtime, maintenance costs, and revenue losses. Some clients are also using AI to improve the quality of their asset and maintenance data, which helps work crews perform field jobs more effectively. Finally, as AI and machine learning technology continue to mature, they will also self-evolve, making our systems more adept at handling energy pressures and uncovering new ways to reduce them.”

Nuclear Power

Technology companies themselves can be proactive in supporting specific types of power generation. “Amazon’s direct investment in nuclear technology and the proposed nuclear plant at Energy Northwest is a much-needed boost for the [nuclear] industry,” said Grant Grothen, a principal at engineering company Burns & McDonnell. “While many tech companies have committed to long-term PPAs, these agreements alone do not provide the financial backing necessary for new reactor developers to navigate the lengthy and costly licensing process.” Amazon’s deal with Energy Northwest, a consortium of public utility districts and municipalities across Washington state, was among three the company inked in mid-October to support development and deployment of SMRs. The company also said it had made an equity investment in nuclear technology group X-energy (Figure 1), and had signed a separate memorandum of understanding with Dominion Energy to look at SMR development in Virginia—the state that leads the U.S. in data center sites. [caption id="attachment_227587" align="aligncenter" width="740"]

1. X-energy’s reactor technology, including the company’s advanced small modular reactors, is being touted to provide power for data centers. Courtesy: X-energy[/caption] Grothen told POWER it’s important for reactor developers to have funding in place to further their research. “Delaying investment until a project is largely de-risked puts most of the risk and burden back on the reactor company,” Grothen said. “Challenges remain, such as refining the NRC [Nuclear Regulatory Commission] licensing process, ensuring HALEU [high-assay low-enriched uranium] fuel availability, and securing government funding support, but this investment is a significant step forward and we’re excited by the progress and opportunities.”

Natural Gas

EQT Corp., a natural gas production company operating in West Virginia, Ohio, and Pennsylvania, in a recent post-earnings call said new projected data centers and coal plant retirements could boost U.S. natural gas demand by up to 10 billion cubic feet per day (Bcf/d) by 2030. “This demand will be regional, with more than half likely to come from the Southeast in the PJM [Interconnection] markets,” said Jeremy Knop, EQT’s CFO. PJM Interconnection is the largest U.S. power grid operator. EQT, which is one of the largest U.S. natural gas producers, said the company has restarted all previously curtailed production. It also in late October announced it had sold non-operated natural gas assets in northeast Pennsylvania to Equinor, Norway’s state-owned energy group.

A major buildout of new natural gas-fired generation is planned across the U.S., in part to satisfy the power needs of the data center and artificial intelligence sector. Read "Hundreds of New Gas-Fired Power Units Planned as U.S. Gas Output Soars."

Recent developments also include news from American Electric Power (AEP), a Midwestern utility that said it would use fuel cells (Figure 2) to support the power supply for data centers. The company signed a deal with California-based Bloom Energy in November; Bloom could supply as much as 1 GW of solid oxide fuel cell (SOFC) capacity to AEP over the next several years. The deal includes an initial order of 100 MW in SOFCs in 2025, and the companies called it the largest commercial utility deal for fuel cells to date. The agreement is earmarked for the growing market for data centers in central Ohio. AEP earlier in 2024 said it has about 15 GW of load commitment from data center projects in its pipeline. [caption id="attachment_227589" align="aligncenter" width="740"]

2. Bloom Energy and electric utility American Electric Power (AEP) have an agreement in which Bloom’s fuel cells could help provide power for data centers in the utility’s territory in central Ohio. Courtesy: Bloom Energy[/caption] “The rapid increase in energy demand is a challenge that AEP is tackling by finding innovative solutions to meet the unique needs of our customers,” said AEP CEO and President Bill Fehrman in a statement. “These fuel cells will help provide data centers and other large customers with the power they need to quickly expand in our regulated footprint as we continue to build infrastructure to deliver reliable energy for all our customers.” Bloom Energy in a statement said, “AEP will deploy the SOFCs as a customer-sited resource where there are delays/constraints in delivering 100% of a customer’s power requirements via the grid. The SOFCs can be deployed as supplemental grid power or as a microgrid.” Bloom Energy said the first fuel cells will use natural gas.

Solar Power

Mary Powell, CEO of Sunrun, a California-based solar power and energy storage group, in late October said the company is talking with data center developers about supplying solar power generation for their facilities. Powell, speaking at the Dervos 2024 conference in Brooklyn, New York, said Sunrun is considering “a couple different really cool models” to providing solar power to tech firms. Powell said that could include partnering with utilities to provide bespoke solar power systems for data centers. It also could involve using existing Sunrun systems to provide needed electricity. Powell said her company has non-disclosure agreements and would not talk about possible partners, but did say, “It’s just a radical collaboration that has incredible benefits for everyone.” The Dervos conference is focused on distributed energy resources (DERs), and modeled after the annual Davos economic forum. Ben Rapp, strategic product development manager at Rehlko (formerly Kohler Energy), an energy solutions group, told POWER: “As a use case, data centers represent a sector to watch in terms of growth and successful deployments. In an industry that requires constant uptime to maintain consistency of the technology that powers our lives, data center operators are constantly looking at the best ways to stay up and running in the event of unplanned outages. They’re also searching for the most cost and energy savings. DERs have become good options for facilities like hospitals and military bases, and we see them becoming an option for data centers as well.” Rapp said, “Optimizing usage with DERs can be one of the biggest questions facing data center operators,” noting the DERs could include battery energy storage systems (BESS), or solar and wind power. “They need to understand what their load profile looks like and have quality data to demonstrate where their energy demands are. Unlike some other types of DER applications, data centers have a consistently high load profile. Some renewable sources of energy must be carefully considered because they may not always be the ideal option to provide constant power. Balance between dispatchable and non-dispatchable assets, and understanding how that fits together with a very consistent load profile, is key to understanding how DERs work for data center use cases.”

Energy Markets and Trades

Energy marketers and traders also have an interest in power demand from data centers and its impact on the power supply and electricity prices. Mike Naughton, CEO of Ohio-based energy broker Integrity Energy, told POWER, “The greatest opportunity for SMR applications in Ohio currently is to combat the rising energy demands of data centers. Central Ohio has become a hotspot for data center mega-campuses.” Naughton said, “Throughout the central [U.S.] heartland, tech giants like Amazon, Google, Microsoft, and Meta have each invested billions of dollars to develop new data centers, focusing on areas like Columbus and New Albany [in Ohio]. Between 2020 and 2024, data center energy demands have already increased sixfold, from 100 MW to 600 MW. By 2030, officials estimate that Ohio data centers will demand a staggering 5,000 MW.” Naughton noted, “The energy infrastructure in Ohio is struggling as well. Our recent utility auctions with PJM Interconnection saw an 833% increase in capacity charges because the state is decommissioning older power plants without replacing them. The dramatic increase in costs affects residents and small businesses significantly more than tech giants.” Naughton added, “We’re hoping that the developing data center economy in central Ohio contributes to a more resilient electrical grid by investing in nuclear-powered data centers. SMRs are an ideal solution for Ohio’s current electrical demand challenges. Once reliability and costs are under control, the state can explore additional SMR applications.”

Foreign Investment

DataBank, which provides edge colocation, interconnection, and managed services, and operates more than 65 data centers in nearly 30 markets, in mid-October announced a $2 billion equity raise led by AustralianSuper, Australia’s largest superannuation fund, with a global asset portfolio of about $40 billion. AustralianSuper, which committed $1.5 billion to the equity raise, will become a significant minority owner of Texas-headquartered DataBank and will join the company’s board of directors. The investment in DataBank (Figure 3) would be AustralianSuper’s first in the U.S. data center market and the second alongside existing DataBank investor DigitalBridge, a global digital infrastructure investment firm headquartered in Florida. DataBank in the past year has announced three new data center campuses, one each in Texas, Virginia, and Georgia. [caption id="attachment_227591" align="aligncenter" width="740"]

3. DataBank, headquartered in Dallas, Texas, is a major player in the data center market. The company in October announced an equity raise led by an Australian superannuation fund. Courtesy: DataBank[/caption] “We are delighted to have AustralianSuper join our investors,” said Raul Martynek, DataBank’s CEO. “Along with the continued support of our existing investors, it’s a vote of confidence in our strategy and our proven ability to execute and scale the DataBank platform. This investment, and our new campuses, are a game-changer for DataBank and our customers, allowing us to bring this capacity to market now and seize the incredible opportunity ahead of us.” “Our investment in DataBank comes at an exciting time in its growth trajectory with strong tailwinds across the sector, coupled with DataBank’s ambitious expansion program and diverse business base,” said Derek Chu, head of American Real Assets at AustralianSuper. “We’re delighted to be helping DataBank, and its experienced leadership team, capitalize on the unprecedented demand for cloud and AI infrastructure. DataBank will grow and further diversify our global digital infrastructure exposure, a sector we believe will help deliver sustainable, long-term performance for more than 3.4 million members.” DataBank has raised more than $4 billion in debt and equity over the past year. DigitalBridge, meanwhile, recently announced an agreement to acquire UK-based data center developer Yondr Group. “Yondr’s assets and strong relationships with leading hyperscale clients align with DigitalBridge’s vision to support the future of digital infrastructure,” said Jon Mauck, senior managing director at DigitalBridge. “Yondr enhances our existing data center portfolio and strengthens our ability to support hyperscalers. Together, we are well-positioned to capitalize on the increasing demand for hyperscale data centers—fueled by AI, cloud computing, and the ongoing digital transformation across industries.”

Optimizing Energy Supply and Demand

Chaudhuri, whose company specializes in AI-powered digital business transformation, noted the challenge of satisfying increasing demand for power without compromising energy. He said strains on utility infrastructure must be addressed, and said the increased power needs from AI technologies will require smarter—and more adaptable—ways to build and operate data centers efficiently and sustainably. “As we transition to a more decentralized and renewable energy landscape, managing volatility on both the supply and demand sides will become increasingly complex,” said Chaudhuri. “Customers and regulators value reliability and affordability, and expect these to be achieved in the most sustainable way possible. AI will play a critical role in integrating diverse energy sources—from solar and wind to nuclear and battery storage—and optimizing the operation of these resources, ensuring they work harmoniously to provide a stable and sustainable energy supply. Furthermore, AI can facilitate the development of ‘smarter’ grids that are adaptive and responsive to real-time conditions, enhancing energy efficiency and reducing carbon emissions. The future of energy is one where AI not only supports but drives the evolution toward a cleaner, smarter, and more sustainable energy ecosystem. “It’s true that while AI holds tremendous potential to transform grid management and operations, it also requires substantial energy,” said Chaudhuri. “Some estimate that AI data centers will demand more than five times the power of traditional facilities, with renewables alone unlikely to meet this demand. As the infrastructure supporting AI expands and more energy-intensive data centers are built, it’s crucial to provide not only reliable power but also sustainable and efficient green power.” Chaudhuri continued, “Several approaches are in play to optimize the energy needs of AI, including increasing the share of renewables in the energy supply mix for data centers. From a locational perspective, the industry and regulators should consider siting data centers in areas where power congestion is lower, and infrastructure build-up is faster [or where approvals can be accelerated]. Increasingly, data centers are integrating captive energy supplies, especially solar, which reduces reliance on grid power. “Data centers can also use AI to run more power-intensive computing tasks [like GenAI training models] when supply is higher, and power prices are lower, distributing demand more evenly throughout the day,” said Chaudhuri. “New data centers are exploring novel cooling technologies as well as more efficient computing architectures, such as Nvidia’s recently announced Blackwell platform, which promises to use 25 times less energy and cost than currently available architectures.” Chaudhuri noted that the use of AI to support AI, while important, will not replace a thorough consideration of how to balance power supply with demand. “Of course, it’s essential to recognize that AI will not be necessary for every job or system,” said Chaudhuri. “Expertise is required to prioritize technology investments thoughtfully, balancing rate increases and major capital expenditures.” —Darrell Proctor is a senior editor for POWER.