Report: Electric utilities can use AI to improve service while – hopefully – meeting AI power demands

In brief

• A new report from Stanford’s Bits & Watts Initiative, developed with Eurelectric, EPRI Open Power AI Consortium, and TERNA, examines the dual challenge of AI for power systems and power for AI.
• AI adoption in utilities will depend on trusted data, secure deployment, validation, workforce readiness, and regulatory confidence, participants emphasized.
• AI-driven data center growth is changing utility planning, with large-load demand arriving faster than traditional transmission, generation, permitting, and equipment supply chains can respond, but this can be managed if utilities and regulators act more aggressively.

Artificial intelligence is reshaping the electric power sector in two ways. Utilities are exploring how AI can improve power supply and demand forecasting, grid operations, asset maintenance, customer service, planning, and internal workflows. Simultaneously, rapid electricity demand growth led by new AI data centers, electrification, and advanced manufacturing, must be served quickly, reliably, affordably, and with public confidence.

However, AI for power and power for AI are no longer separate questions, according to the new report “AI x Energy: A Transatlantic Utility Perspective” from Stanford University’s Bits & Watts Initiative, The two developments are increasingly part of the same infrastructure, governance, and affordability challenge.

The report synthesizes a private executive roundtable held late last year. The meeting was led by the grid research and education program Bits & Watts, the European electric utility association Eurelectric, the U.S. electric power research institute EPRI, and the operator of Italy’s electricity grid Terna. Some 70 executives and engineers from U.S. and European utilities, technology companies, and system operators participated. The report does not attribute comments to individual participants or organizations, as agreed in advance of the meeting to foster open, frank discussion.

“AI is changing what utilities can do and what they must do,” said Liang Min, managing director of the Bits & Watts Initiative. “The opportunity is enormous, but the energy system will need better data, faster planning, more flexible large-load arrangements, and clearer public value to move at the speed AI now demands.”

AI for power systems

The first part of the roundtable focused on where AI is already creating value for utilities and what is needed to move from pilots to enterprise-scale adoption. Participants pointed to emerging applications in search and retrieval, documentation, software development, customer service, forecasting, anomaly detection, maintenance inspection, asset-health monitoring, and operator support.

But participants also emphasized that the limiting factor is rarely the AI model alone. Fragmented data across legacy systems, vendor platforms, market tools, GIS environments, customer systems, and maintenance records remains one of the biggest barriers.

“AI can help utilities improve reliability, customer service, planning, and operations, but scaling those benefits requires trusted data,” Robert Chapman, executive vice president and chief commercial customer officer at EPRI said after the meeting. “The sector needs secure deployment models, benchmarkable datasets, validation methods, and industry collaboration so that AI can move responsibly from pilots into operational workflows.”

The report distinguishes between productivity-oriented AI, such as document search and customer-support tools, and operational AI, which requires greater caution because utilities must protect reliability, cybersecurity, market integrity, customer data, and public safety. The near-term opportunity, meeting participants agreed, is not replacing engineers or operators, but helping them work faster, reason across complex systems, and compress planning and interconnection processes that can currently take months or years.

Power for the AI economy

The meeting also explored how rapid growth in AI workloads and hyperscale data centers – along with electrification of transportation, reshoring of industry, and other drivers – is reshaping electricity demand. Participants described the power-for-AI challenge as a speed problem as much as a supply problem. Large-load demand is arriving faster than traditional grid planning, transmission buildout, generation development, permitting, and equipment supply chains can respond.

Power availability is a major determinant in where AI investment can happen, though data center growth is not only an electricity issue. Data centers also need communication fiber access, land, cooling, water, permitting, local economic development, and broader industrial strategy.

“The AI transformation is not only a technology story. It is an infrastructure story,” said Kristian Ruby, Secretary General of Eurelectric. “Across Europe and the United States, utilities are working to modernize grids, accelerate investment, and preserve affordability while enabling the digital economy.”

The report highlights several near-term strategies: better use of existing grid assets, dynamic or ambient line ratings, improved forecasting, more granular power-flow analysis, batteries and storage, flexible demand, and new interconnection models for large loads. These tools can buy some time, participants agreed, but they cannot replace the need for new infrastructure, including clean power generation, transmission and distribution lines, substations, and assets that back up intermittent solar and wind power.

Data centers are not all the same. Training, inference, enterprise cloud, and mission-critical digital services have different power demands, flexibility potential, backup capabilities, and reliability needs. A more differentiated understanding of large-load behavior could support earlier service for some projects while preserving reliability.

Affordability, legitimacy, and collaboration

Across both sessions, affordability and public trust emerged as decisive constraints. AI adoption in utilities will only scale if it clearly improves service, lowers costs, strengthens reliability, or creates value for customers. Similarly, AI-related load growth will only remain politically durable if households and small businesses believe they are protected from unfair cost shifts.

Cost allocation is a core legitimacy question, the report finds. Large-load customers should pay fairly for the upgrades they require, while policymakers and regulators must also recognize that some infrastructure investments may create broader system benefits. Community acceptance will also depend on land use, water, noise, local environmental effects, backup generation, and whether communities are engaged early and transparently.

The transatlantic nature of the roundtable revealed different regional emphases. European participants focused more on physical and permitting constraints, cross-border needs, and the risk of underinvestment. U.S. participants focused more on latent capacity, local congestion, and opportunities to unlock near-term headroom through better use of existing assets. But the report concludes that both regions face the same structural challenge: how to expand and optimize the power system fast enough to support AI, electrification, decarbonization, and industrial competitiveness while preserving affordability and public trust.

System operators are at the center of this transition. Participants noted the need for better tools to understand grid constraints, accelerate planning, and coordinate innovation across borders – and agreed that collaboration between Europe and the United States can help turn shared challenges into shared progress. Terna, Italy's electricity transmission system operator, was among the participating organizations.

The report identifies several priorities for follow-up collaboration, including benchmark datasets and validation methods, secure AI deployment patterns, queue-management and cost-allocation frameworks, flexibility products for data centers, permitting and public-engagement tools, and transatlantic demonstrations on advanced transmission technologies, storage, clean firm supply, and customer-side flexibility.