Introduction
This roundtable explores how the explosive expansion of AI — and the associated growth in data centers needed to power them — is creating new and complex demands on the electricity grid, exposing limitations in traditional planning approaches. Moderated by Bo Yang, Vice President of the Energy Solutions Lab at Hitachi America R&D with insights from Haibin Sun, Director of Transmission Planning at Exelon, and Bret Toplyn, Senior Director at Hitachi Energy, the discussion examines how AI may help utilities better manage these challenges. The conversation delves into AI’s potential to accelerate analysis, inform decision-making, and promote greater collaboration across the energy industry ecosystem. While there is broad consensus that integrating AI with AI with established engineering practices can help shift utilities from reactive to proactive planning, the panelists also stress the importance of ensuring continued reliability, efficiency, and affordability as the sector evolves.
The growing impact of data center loads
Bo: There’s been a lot of discussion about the growth of AI and the impact data centers are having on electricity demand and the strains they are putting on the grid. How big is that impact? Are you seeing it in interconnection timelines and the issues you face?
Haibin: Exelon is a transmission and distribution company. We connect customers, like data center providers, to the grid. We provide analytical feedback to customers on their proposed projects.
The biggest challenge we face is sheer volume. A few years ago, a data center might have been 50–100 MW. Now, we’re seeing projects approaching gigawatt scale — and we don’t yet have operational history at those scales.
Another challenge is the load profile. Planning studies can’t just look at peak demand — they need to examine seasonal and sometimes hourly variations.
The third challenge is the load ramp schedule. A customer might design for 1 GW, but the time to reach that capacity could be five or ten years.
Finally, there’s cascading risk. Historically, we thought of that in terms of generation — a large generator tripping offline could destabilize the system. With gigawatt-scale loads, a single large customer could trigger stability or dynamic issues. We have limited data to model these accurately, so there are industry-wide discussions about monitoring, protection schemes, and planning for these scenarios.
AI solutions for grid challenges
Bo: How can AI-based tools help address some of these load growth challenges? Today’s issues include inefficient processes, limited predictive capabilities in existing tools, and the slow speed of traditional simulations which require extensive sensitivity analyses and aren’t easily adaptable to changing grid conditions.
Another challenge is fragmentation. Because so many stakeholders are involved in reviews, decision-making often becomes slow and disjointed. AI can help by addressing bottlenecks across the entire workflow — from intake and automation to analysis. By integrating AI with existing IT and operational technology systems, using advanced computing to accelerate simulations, and connecting with legacy software to enhance performance, we can build AI-driven tools that improve collaboration among Regional Transmission Organizations, Transmission Owners, Independent System Operators, and developers. The result is fewer delays and greater overall efficiency.
Bret, what is Hitachi Energy’s proposition to customers?
Bret: Hitachi Energy enables the delivery of efficient, flexible, and secure design and implementation, bridging data centers and the grid to transform them from liabilities into valuable grid assets. We assist with siting — putting data centers where they cause the least disruption to the grid and placing generation where it adds the most value. Upgrade costs can be significant, so efficient siting is essential. And of course, we’re working with you and your team at Hitachi America R&D to integrate AI into our simulations and optimization solutions so customers can run more scenarios faster.
Bo: Haibin, what expectations are you seeing from customers?
Haibin: Customers increasingly expect faster service — not years of waiting. Traditionally, Transmission Operators processed interconnection requests one by one, but that approach often created bottlenecks, delays, and speculative queue positions. To address this, we’ve moved to cluster-based studies, grouping requests so that solutions can serve multiple projects more efficiently.
Investors also want a fair playing field. Serious developers prefer competing against other committed investors rather than speculative entrants. This has prompted the introduction of tariff designs aimed at discouraging speculative projects, though these still require regulatory approval.
Flexibility is another area of innovation. Some customers propose pairing on-site generation — for example, 500 MW — with a 1 GW data center load. Studying the net load, rather than gross load, can reduce the need for upgrades. While there’s not yet a standardized approach, this is a promising direction for planning and operations.
Bret: We totally agree, Haibin. On-site generation integrating multiple energy sources, including Battery Energy Storage Systems, or BESS, within a microgrid and in coordination with the larger grid is key. This is the focus of Hitachi’s “Energy Hub for Data Centers,” a collaborative initiative that brings together the energy, generation, data center, and utility sectors to explore new solutions and partnership models.
Scaling AI-based tools
Bo: Grid operators can be conservative and slow to adopt new tools. How receptive are they to adopting innovative AI-based tools? What are the obstacles? Bret?
Bret: One is comfort with the cloud — grid data is sensitive. Customers often face two choices: keep buying hardware and relying heavily on internal IT, or move to a SaaS model. SaaS can be highly efficient if done well, but expensive if not.
Another challenge is data management. Scenario generation pulls from multiple sources and formats, and all that data needs to be clean and structured before meaningful analysis can happen.
Bo: We’ve run multiple tests and proofs of concept on AI’s effectiveness. For example, SPP will use our AI-based analytics tool with the goal of achieving up to 80 percent reduction in interconnection analysis times.*1
*1 https://spp.org/news-list/spp-partners-with-hitachi-to-develop-advanced-ai-solution/
To be adoptable, affordable, reliable and scalable, AI should act as a co-pilot to physics-based models. When those models are too slow or compute-heavy, AI accelerates them, combining speed with reliability.
We don’t discard legacy investments — AI complements them. Traditional grid studies can require tens of thousands of scenarios. By training AI alongside those studies, we can let AI handle more of the workload over time. Benchmarks show runtime reductions of 30–40 percent even on single-core CPUs, and up to 100× faster in benchmark tests using parallel computing or GPUs. That shortens delays and accelerates decision-making.
Haibin: For our part, we of course welcome innovation. I’ve reached out to Bo and others because we need solutions that improve efficiency. But I’m also realistic. There’s promising research, but I don’t yet see a tool that can fully replace the manual engineering work involved in transmission planning. One challenge is that power-system topology varies greatly, which makes AI training difficult.
Ideally, instead of our engineers manually creating all the scenario files, an AI platform could generate them automatically. We’re not there yet, but I hope vendors like Hitachi will lead the way.
Bo: The grid is said to be one of humanity’s great inventions. The challenge isn’t a lack of innovation — it’s that new solutions must meet the safety and accuracy requirements unique to grid operations.
Many cutting-edge AI algorithms perform well on public datasets, but they can’t meet the reliability standards required for grid planning. That’s why Hitachi’s approach is to pair AI with physics-based tools — whether legacy or new — so AI augments rather than replaces them. That ensures safety, accuracy, and adoption.
Bret: As a provider of these tools, our focus is flexibility. We design solutions to be cloud-native, with flexibility in mind, but also with the ability to deploy on-premise. Customers increasingly want to complement their existing tools rather than replace them.
Five years ago, new solutions sometimes felt ahead of their time and were harder to adopt. Now, demand for innovation is so strong that it feels like we can’t build fast enough.
Looking ahead
Bo: What do you think will be mainstream in the next three to five years?
Bret: I expect more standardization and some centralization — both within organizations and across companies. Today, generation and transmission planning are separate, and even within transmission, reliability and economics are handled separately. Centralizing and standardizing these efforts will improve efficiency and make it easier to adopt new technology.
I also expect more collaboration — grids and customers learning from each other, sharing what works, and finding solutions together.
Haibin: From a large-load customer’s perspective, interconnection is challenging. Supply chain issues affect both data-center construction and the utility’s ability to complete system upgrades — even something as basic as a circuit breaker can take years to procure.
From the Transmission Operator’s perspective, we want clearer regulatory rules and tariffs to reduce speculative requests and wasted effort. We’re also eager for AI tools that help us work smarter and more efficiently.
I agree with Bret on integrated, collaborative planning. The industry is starting to create long-term plans — 10 to 20 years out — that combine load growth, generation siting, and grid enhancement.
We should also be more open to grid-enhancing technologies that can buy time while still maintaining reliability.
From the residential customer’s perspective, if large loads take on more of the cost over time, that could lower their bills, but only if those loads ramp up as projected. Affordability will remain a critical constraint on investments.
Bo: I see AI playing an increasingly important role alongside traditional decision-making engines. It will improve reliability, affordability, and efficiency, and help smooth collaboration among all stakeholders.
Haibin: To Bo’s point, in 2000, the National Academy of Engineering named the electricity network the greatest engineering achievement of the last century. If AI achieves its potential in the grid space, maybe it will make the list for the next century.
Bret: I agree. We’re on the verge of a new era, merging AI and energy. Large loads come from AI and data operations, but those same technologies will help us analyze, plan, and operate the grid more effectively.
Bo: On that positive note, I’d like to thank both of you for participating in this enlightening discussion.
The roundtable highlighted that the rapid expansion of data centers is creating significant challenges for electricity grids, but also stimulating the development and adoption of new tools and approaches. AI has the potential to improve interconnection studies, accelerate workflows, and foster deeper collaboration among utilities, developers and regulators — provided it complements, rather than replaces, established engineering models. The path forward requires both innovation and caution, with solutions meeting the strict standards of safety and reliability that define grid operations. While ongoing trends suggested a sustained growth in energy demand — driven by accelerating data center buildouts, electrification, and greater renewable integration — its trajectory will depend on how quickly the industry can modernize its planning and operations. If developed and applied with appropriate diligence, AI may become not just a response to today’s load crunch, but an important pillar of the evolving energy landscape.
Profiles
Hainbin SUN, Ph.D.
Director, Transmission Strategy Analysis
Exelon Corporation
Haibin Sun is a director of Exelon’s Transmission business unit. He oversees planning studies, economic analysis, technical modeling and technology evaluation for Exelon’s transmission investment, competitive solution proposal development, and related policy advocacy initiatives. Before his current role, Haibin took various leadership roles within Exelon and Constellation’s Commercial business unit and Risk organization including serving as the Director & Chief of Staff of Commercial & Industry Power, Director of Quantitative Modeling & Deal Review for energy commodities, M&A and capital investment businesses, and VP of Risk Analytics. Prior to that, Haibin worked in the capital market division of a bank, the logistics solution practice of a consulting firm and the energy economics unit of a multinational conglomerate. Haibin is a Senior Member of IEEE and volunteers in various industry committees.
Haibin received his doctoral degree in economic decision analysis from the School of Industrial & Systems Engineering, and a M.Sc. in quantitative & computational finance, at the Georgia Institute of Technology, and a M.Eng. and B.Eng. in electrical engineering from the Southeast University in China.
Bret TOPLYN
Director, Product Management
Hitachi Energy Ltd.
Bret Toplyn is a director of product management and energy analytics at Hitachi Energy. He began his career as a regulatory analyst for Goldman Sachs. He then started working for ABB Power Grids (now Hitachi Energy) in 2012. Through his career he has worked with customers in energy regulation, renewable generation, transmission planning, grid operations, market pricing and fossil generation. In his current role, Bret works with customers and Hitachi Energy teams to produce best-of-bread solutions focused on market analysis, grid simulation and forecasting to enable better operational and investment decisions. Bret has a M.Sc. in business analytics from the University of Colorado, Boulder.
Bo YANG, Ph.D.
Vice President and Lab Manager, Energy Solutions Lab,
Research & Development Division,
Hitachi America, Ltd.
Bo Yang heads the Energy Solutions Lab at Hitachi America R&D. Her team has pioneered the integration of AI/ML techniques in energy applications, developing several innovative AI and IoT platforms for the utilities industry. Using her extensive professional and academic experience in transmission & distribution energy resource integration and control, grid automation, smart grid, AI/ML and enterprise system architectures, Bo represents Hitachi in energy projects funded by U.S. federal and state agencies.
Bo received her doctoral degree in electrical engineering from Arizona State University. She is a Fellow of the Institute of Engineering and Technology (IET).
