Powering AI: The Energy Implications of Data Centers Expansion in North America
by Zainab Shah, Lead EVCI Analyst Americas at PTR Inc. and
Mike Sheppard, CEO and Americas Lead at PTR Inc.
Introduction:
Generative artificial intelligence (Gen AI) has not only accelerated demand for computing power but fundamentally reshaped the architecture of modern data centers. While power delivery, distribution, and thermal management have long been critical considerations, they are now becoming the defining constraints in facility design. The rise of dense GPU clusters, unprecedented rack power densities, and real-time AI workloads has elevated energy infrastructure. With U.S. data center market expected to grow by a CAGR of 15% between 2024 and 2030 in the US [1], energy demand will largely be driven by increasing demand for cloud services, digital transformation initiatives, and the proliferation of IoT devices.
According to the U.S. Department of Energy, data centers will account for almost 12% of national energy consumption by 2028 up from 4% in 2022. [2] In 2024, U.S. investments in data centers were nearly four times higher than in 2019, signaling the industry’s rapid pivot to accommodate the computational demands of generative AI models.[3] The U.S. hosts around 3,700+  data centers—more than any other country worldwide.[4] A surge in generative AI usage—through Large Language Models (LLMs) like ChatGPT, Copilot, Gemini, and others—is driving a rapid rise in data center construction.
 Evolving energy needs and infrastructure:
The energy profile of AI-focused data centers is significantly more demanding than that of their predecessors. A decade ago, a 30-megawatt (MW) data center was considered a large-scale facility while today, developments of 200 MW or more are increasingly standard, driven by the growing compute intensity of AI workloads. This surge in demand is reshaping the energy and infrastructure profile of the industry. AI-optimized data centers, in particular, are pushing power requirements to new heights due to significantly higher rack power densities. Average densities have more than doubled in the past two years from around 8 kW per rack to 17 kW and are expected to exceed 30 kW by 2027.[5] As a result, hyperscale AI facilities now frequently demand tens of megawatts in dedicated capacity per site, compelling utilities to engage in far more complex interconnection and capacity planning than before. This infrastructural strain is particularly acute in regions like Northern Virginia, the world’s largest data center hub. Home to over 350 data centers, the region has seen a dramatic uptick in demand for electricity. Dominion Energy, the utility serving Northern Virginia, reported in 2024 that its queue of large-scale interconnection requests is set to experience wait times of four-to-seven-years. Electricity capacity under contract surged to 21 GW by mid-2024 and is expected to reach nearly 40 GW by the end of 2025. As a result, Dominion has proposed over $50 billion investment into grid expansion initiatives from 2025 and 2029, including advanced substations and renewable integration. The utility added 1 GW of capacity via 15 new data centers in 2024 and plans to match that in 2025. [6]
Primary markets saturate, secondary sites surge:
Major data center hubs like Dallas–Fort Worth, Phoenix, Chicago, Northern California, Atlanta, and Los Angeles continue attracting hyperscale investment due to their high-speed fiber networks and favorable tax incentives. Chicago, for example, benefits from Lake Michigan access for water-based cooling and the Illinois Data Center Investment Program, which offers 20-year tax breaks and credits.[7] This is not the case everywhere as growth in regions like Phoenix and Los Angeles is bumping up against regulatory hurdles and natural resource limitations.
As these primary markets reach saturation and utilities struggle with infrastructure constraints, developers are increasingly turning to secondary locations offering cheaper land, flexible grid capacity, and fewer community objections—places like Tennesee, Ohio, Indiana, and Nevada. Notable among these is Elon Musk’s xAI project in Memphis, Tennessee, announced in June 2024. This $400–800 million AI data center draws 150 MW of power and uses millions of recycled gallons of wastewater daily. To support its energy and water needs, Memphis Light, Gas and Water (MLGW) and the Tennessee Valley Authority (TVA) are deploying targeted solutions including local solar, hydro, natural gas backup generation, and a recycled-water plant to ensure resilience and sustainability.[8]
Utilities and big tech are adapting through distributed power:
U.S. utilities are facing mounting pressure to modernize as hyperscale data center campuses strain existing infrastructure. In Virginia—within the Pennsylvania-New Jersey-Maryland (PJM) interconnection region—regulators and utilities are
alike are questioning whether current infrastructure can continue supporting the pace of growth.[9] Pennsylvania-New Jersey-Maryland (PJM) interconnection, which manages the regional transmission of Virginia, has already issued warnings about potential shortfalls in power availability starting as early as next year. The increasing footprint of data centers has also driven up wholesale power prices. In Loudoun County, Virginia, electricity capacity prices surged from $29 to $444 per megawatt-day in 2024. In response, legislators are considering measures to rebalance cost allocation and assess community impact—potentially requiring large-scale energy users to share more of the infrastructure burden.
 In parallel, big tech companies are taking a more active role in reshaping the energy landscape— partnering on power system innovation, helping to create a market foundation for advanced nuclear and storage projects. Rather than remaining passive buyers, firms like Meta, Google, Amazon, and Microsoft are forging direct partnerships with utilities to co-develop clean, scalable power sources. On June 13th, Meta announced a 150 MW geothermal project with XGS Energy in New Mexico to reduce its dependency on the grid for its data center operations. The project, set to be phased in by the end of the decade, will represent around 4% of total U.S. geothermal production.[10] In a similar move on May 13th, Google joined forces with utility NV Energy and geothermal developer Fervo Energy to scale up geothermal power supply for its data centers in Nevada, increasing from 3.5 MW to 115 MW over six years. Notably, Google participated in the utility planning and the creation of the Clean Transition Tariff (CTT)—a first-of-its-kind rate structure for clean energy procurement in regulated markets.[11]
Meanwhile, Microsoft’s Swedish data center is setting the standard for sustainability, featuring 24/7 renewables from Vattenfall, battery storage, rainwater cooling, and a facility that repurposes 12,000 servers per month. These approaches are now being explored for deployment across North America. [12] Amazon, too, is deepening its clean energy portfolio, securing a $20 B power purchase agreement with a Pennsylvania nuclear plant on June 9th 2025—just days after partnering with Talen Energy on June 11th to supply Amazon Web Services (AWS) data centers with power from the Susquehanna nuclear facility in Pennsylvania.[13] This momentum is supported by a broader coalition: in May 2024, Amazon, Google, Microsoft, and Nucor joined Duke Energy to launch the Accelerating Clean Energy (ACE) Tariff, which enables large users to fund small modular nuclear reactors (SMR) and long-duration storage projects in the Carolinas over the next five years.[14] Even startups like Exowatt are entering the fold, raising $90 million on 22nd April 2025 to develop modular solar-battery platforms designed specifically for AI data centers, offering containerized plug-and-play server racks designed to run on low-carbon alternatives that reduce grid dependence. [15]
Federal action accelerating AI-driven data centers:
In a notable policy pivot, the U.S. Department of Energy (DOE) is accelerating regulatory action to address rising electricity demand from artificial intelligence (AI) and data center growth. In April 2025, the DOE released a Request for Information (RFI) identifying 16 federally owned sites—including national labs and former energy production facilities—as potential hubs for co-located data centers and energy infrastructure. The agency is soliciting input from grid operators, developers, and technology providers on development models, power timelines, and interconnection strategies. As hyperscale and AI-driven data centers proliferate, regulators at all levels are responding with new policies aimed at managing grid impact, cost fairness, and clean energy adoption.
·      On 21st January, the Trump Administration announced a $500B initiative called “Stargate” backed by private-sector investment to develop AI infrastructure including data centers in the U.S. The initiative is a joint venture involving OpenAI, SoftBank, and Oracle with plans to build 20 large-scale data centers, 10 of which are under construction in Texas. With $100B committed, the remaining investment will be deployed over the next four years.
·      On 18th February, California Public Utilities Commission shared its plans for a dedicated data center tariff requiring utilities to submit proposed rate structures by mid‑2026. It will also mandate newly interconnected data centers to procure 100% zero-carbon electricity by 2030, shaping how tech giants source their power across the state.
·      Texas passed legislation on 6th June 2025, compelling large energy users—including data centers with demand greater than 75 MW to shoulder full interconnection and grid upgrade costs, submit to upfront fees, and submit backup generation plans to ensure reliability during peak events.
·      Recently, the Department of Energy issued a Request for Information (RFI) on 3rd April 2025 identifying 16 federally owned sites for AI-ready data center development. These sites include DOE national laboratories and former military installations, selected for their co-location potential with surplus or new energy infrastructure. The data centers are planned to begin construction by late 2025, with operations starting by 2027.
A future built on energy intelligence:
Data from PTR suggests that the U.S. data center market is poised for robust growth through 2033, driven by rising investments in data center infrastructure and a growing proportion of hyperscale facilities anticipated by 2030.[16] Electricity demand from data centers worldwide is set to more than double by 2030 to around 945 terawatt-hours (TWh), with AI-related workloads alone potentially accounting for 20% of that total by 2028.[17] The energy required to train and operate large AI models has begun to rival that of entire industrial sectors. A single generative AI request can use 10 times more power than a typical Google search. To meet this demand, hyperscalers, utilities, and regulators must co-develop next-generation energy solutions, including high-efficiency chip technologies, renewable energy portfolios, on-site microgrids, and AI-driven demand-side management tools. Utilities like Dominion Energy are investing heavily in grid upgrades while data center operators are deploying new chip architectures like Nvidia’s Blackwell and AMD’s MI350 designed for greater performance per watt. Tech companies are increasingly taking on proactive roles in shaping utility planning and renewable integration in the future. A standout example is Google’s partnership with NV Energy and the Public Utilities Commission of Nevada, which resulted in the creation of the Clean Transition Tariff (CTT). A first-of-its-kind rate structure designed to facilitate the procurement of clean geothermal power through utility-owned resources while ensuring cost recovery and grid stability for the utility, creating a mutually beneficial model that can be replicated across other markets. Other companies are expected to adopt similar rate structures to leverage renewable resources and support the upcoming power demand for AI-driven computing. In conclusion, the long-term success of AI infrastructure in North America will hinge not just on performance, but on how intelligently and collaboratively these ecosystems are engineered.
About the Authors:
Zainab Shah // Lead EVCI Analyst, PTR Americas
Zainab Shah is Lead EVCI Analyst - Americas specializing in electric vehicles and charging infrastructure at PTR Inc. She brings a robust technical background of automotive industry knowledge, having past experience in product development and vehicle localization at Toyota. Zainab offers insightful, data-driven analysis of evolving new energy markets, with a strong focus on North America. She holds a Bachelor’s degree in Electrical Engineering from the National University of Sciences and Technology (NUST.)
Mike Sheppard // CEO and PTR Americas Lead
Mike Sheppard has 20 years of market research experience designing numerous research practices from scratch while leading over 100 bespoke projects with Fortune-500 companies. In 2016 he co-founded Power Technology Research (PTR) and has since launched new research practices in solar, storage, battery, and e-mobility. In 2020, he co-founded Matos, an intelligence automation company focused on providing powerful AI-driven tools for the market research sector. In 2023, this business was acquired by PTR. Prior to founding PTR, he spent 8 years with iSuppli/IHS Markit in various analyst and consulting roles where he covered a broad range of sectors including mobile, renewable power and electricity transmission and distribution (T&D). In his last role, he led the power technology consulting group. He is an expert on the PV industry and having performed numerous competitive dynamics and opportunity assessment projects, covering upstream, downstream, and supply chain topics. In 2008, he obtained two Bachelor’s of Science in both Financial Services and Corporate Finance from San Francisco State University.
Contact: ([email protected])
[1] Power Technology Research (PTR)
[2] U.S. Department of Energy
[3] International Energy Agency (IEA)
[4] Data Center Map
[5] McKinsey
[6] Dominion Energy
[7] Council of State Governments (CSG) Midwest
[8] Memphis Light, Gas & Water (MLGW)
[9] Pennsylvania-New Jersey-Maryland (PJM) Interconnection
[10] Meta
[11] Google
[12] Microsoft
[13] Amazon
[14] Duke Energy
[15] Exowatt
[16] Power Technology Research (PTR)
[17] International Energy Agency (IEA)