Powering the AI century: Why utilities must lead

The U.S. power grid—and the utilities that operate it—are facing a paradigm shift unlike anything seen in the last century. The driver isn’t policy—it’s compute. AI is accelerating demand for large-load data centers at a pace that’s outstripping traditional utility planning cycles.

This isn’t a future problem. Burgeoning demand for AI data centers is already reshaping load profiles in Virginia, Texas, and other hot spots. Guidehouse Research predicts that Virginia’s data center load will approach half of the state’s overall power consumption by 2030—up from about a quarter today.

This shift is happening faster than most utilities are structurally prepared to handle. In Guidehouse’s tenth annual State and Future of the Power Industry pulse survey, more than 60% of utility leaders said they are ready for some load growth, but that system reliability needs are likely to temper upside.

Accelerated energy demand is coming, particularly over the next five years, but if utilities don’t lead through the data center surge, others will. The grid is no longer background infrastructure; it’s a strategic platform for the AI economy and one on which substantial new utility revenue can be built—if the right preparation and capital planning processes are implemented today. 

Billions in investment and revenue are on the line

Meeting the load demands of mushrooming data centers will cost billions of dollars and likely hundreds of billions. Duke Energy has already announced agreements worth $8 billion for new data center capacity investments; Entergy has committed more than $4 billion in Louisiana and Mississippi.

And in the first half of 2025 alone, there have been nearly 2,000 new interconnection requests made across the U.S. Based on the generation source mix and capacities requested, the cost of these new projects alone could easily reach $50 billion.

Considering that overall data center load is projected to nearly triple by 2035—to more than 700 TWh annually—utilities that don’t get ahead of the curve risk foregoing a significant revenue opportunity. Based on an average industrial rate of approximately ~$0.08/kWh, the potential utility revenue from new data center load in the U.S. could approach $250 billion (cumulative) over the next decade.

These numbers aren’t theoretical, and they get very large, very quickly, but forecasting the actual impact for a given utility at a certain point in time remains challenging. Load forecasting has always been part art, part science, and knowing exactly how and when these new facilities will impact the regional grid is complicated by the rapid evolution of more efficient chip and cooling technologies, as well as lengthening interconnection queues.

To get ahead of the wave, utilities should be investing in advanced scenario planning and AI-driven forecasting tools. Waiting for certainty is no longer a viable strategy. 

A perfect storm

In addition to new data center load growth, electrification trends and resiliency demands are creating a perfect capital investment storm. Many utilities will require a 50-100% increase in capital infrastructure investment over the next three years, and few are ready to deliver.

A transformational change of capital delivery models is needed to scale at the rate of market demand and maintain industry health. If they can’t adapt, utilities face numerous risks:

• Market disruption from AI companies adopting substitute generation for data centers: The number of data center on-site generation announcements in 2024 was greater than that of the previous four years combined.
• Heightened competition for equipment and constrained supply chains, resulting in higher prices, longer lead times, and capacity shortages: Transformer lead times are now up to two years, a four-fold increase from pre-2022, and prices are up four to nine times higher.
• Degraded financial performance and lost market share to competitors: Delayed data center grid connection lengthens the time-to-revenue from new connection investments.
• Data center customer dissatisfaction due to delayed connections to the grid or rising costs: Data center grid connection delays are now as long as seven years in Virginia.
• Delayed advances in grid resilience and modernization: Many transmission capacity upgrades are already delayed beyond FERC Order 881’s July 2025 deadline for the use of ambient-adjusted ratings on transmission lines.

Traditional tariff models won’t cut it

The traditional models for capital planning, rate-based returns, and long-cycle grid investment no longer match the velocity of change demanded by data center developers and hyperscalers.

Utilities were built for long-term capital recovery and the lead times for planning, regulatory approvals, permitting, and procurement (exacerbated by persistent supply chain constraints) make it incredibly difficult to meet the 18- to 24-month time to operation that data center developers are demanding.

This mismatch creates significant opportunity cost risk for utilities built for measured, incremental change. Hyperscalers and developers will quickly move on to the next site in their list—or contract for behind-the-meter (BTM) generation—if the local utility can’t meet its load requirements.

Utilities need new frameworks to manage risk and unlock opportunity. Novel tariff designs are emerging to adapt to the data center load challenge today, but best practices have yet to be standardized, and regulatory signoff has been a mixed bag. Emerging tariff models designed to support data centers as well as, in many cases, the green energy goals of hyperscalers include:

• Green tariffs: Large-load tariffs support clean, firm generation, such as the NV Energy/Google geothermal project.
• Data centers as a grid resource: To facilitate faster interconnection, data centers agree to demand response and/or energy storage programs for the utility, such PG&E FlexConnect, EPRI DCFlex programs.
• Revenue guarantees: Data centers commit to multi-year payments to guarantee utility recovery of grid upgrade costs, such as Dominion Energy’s proposed high energy rate class.
• Premium interconnection tariffs: The developer pays a premium for interconnection and clean energy corridor access, such as the Google/Omaha Public Power District partnership.
• Time-of-use (TOU) pricing: TOU-based rates encourage high-load customers to shift load to off-peak hours.
• Very large customer tariffs: Proposed tariff model is designed for hyperscale data centers, such as WE Energies.
• Bespoke resource tariffs: Custom power purchase agreements combine dedicated (often nuclear) generation for data center reliability, such as the Microsoft/Constellation Three Mile Island project.

Utilities can begin piloting these models while simultaneously monitoring what works best for other utilities. But capital planning in anticipation of the data center surge can’t wait.

Rethinking legacy processes: Setting up for success

Rather than simply scaling legacy capital planning and buildout models, utilities will need to change their ways of working to be more flexible and responsive to shifting landscapes, and they’ll need to break down persistent silos to achieve continuous improvement.

Where new capital projects may have been previously performed in-house, utilities will increasingly need to work with engineering, procurement, and construction (EPC) firms—which may be in short supply. Establishing preferred partner status with major firms will become key.

Furthermore, technology and system integration—and robust data management—will be paramount to maintaining project oversight and achieving efficiencies. Pen and paper or Excel spreadsheets will no longer suffice; with billions of dollars on the line, project governance must go beyond “on time and on budget.”

At the same time, utilities must fulfill their obligation to serve the public good (such as their rate base). They must strike an increasingly difficult balance between making large, concentrated capital deployments and not passing excessive costs on to already sensitive ratepayers. Guidehouse recommends utility leaders take the following actions today as the AI boom ramps up:

1. Partner across sectors—tech, regulators, and other utilities—to shape the future collaboratively.
2. Form a cross-functional internal team focused on data center readiness and breaking down silos.
3. Align operating models against new strategies to gain workforce commitment.
4. Engage regulators early to explore flexible cost recovery structures.
5. Pilot new pricing models offering premium zones; green service level agreements tied to clean energy; and traceable, compute-ready supply corridors. Or offer outcome-aligned tariffs.
6. Reimagine procurement solutions to incentivize strategic alliances with EPC firms and equipment vendors to become valued partners with providers of these resources.
7. Integrate and leverage AI to improve forecasting and scenario planning.

The AI century is here. The grid is no longer just wires and substations; it’s the backbone of America’s digital future and a geopolitical imperative. Utilities have a choice: evolve into adaptive, compute-aligned infrastructure partners—or risk being sidelined by faster-moving players.

With a pro-AI, pro-data center, and pro-nuclear administration, the conditions are favorable. But leadership requires action and proactive, agile capital planning is a key place to start.