Q&A: Developing a sustainable power grid in the era of AI

Le Xie, Professor of Electrical Engineering at the Harvard John A. Paulson School of Engineering and Applied Sciences (SEAS), wants to know how we can modernize the electric grid to support rapid electrification and the growing demands of AI infrastructure. His research at SEAS focuses on the intersection of power systems, artificial intelligence, and decarbonization. We spoke to him about his work and the challenges facing the electric grid today. This interview has been edited for clarity and length.

Why is this a pivotal moment for the power grid and energy research?

We are living through one of the largest infrastructure buildouts in human history: the buildout of artificial intelligence infrastructure. Like past infrastructure revolutions, it presents enormous societal and economic opportunities—but also profound technical challenges.

When people think about AI infrastructure, they often focus on compute, GPUs, fiber optics, and communications. But electric power is just as fundamental. Without a reliable and scalable power system, none of this infrastructure can function.

What is limiting AI infrastructure growth today?

In many parts of the world, the bottleneck is a lack of chips or technical talent. In North America, the primary constraint is increasingly electric power.

Electricity demand in the U.S. was largely flat for decades—growing just 1% to 1.5% annually since the late 1970s. Now, almost overnight, we are seeing gigawatts upon gigawatts of new demand, driven by hyperscale data centers, electrified transportation, and building heating.

The grid was simply not designed for this pace or scale of growth.

Why is operating the power grid becoming more complex?

The grid must balance supply and demand in real time, on a second-by-second basis. That task is becoming harder for several reasons.

On the supply side, a growing share of electricity comes from variable renewable sources such as wind and solar. On the demand side, electrification and AI computing are driving unprecedented growth. These changes are layered on top of aging infrastructure, making grid operation one of the most complex engineering systems in existence.

How does climate change factor into this challenge?

Electrification is not only about supporting AI—it is also central to addressing climate change. Decarbonizing the economy depends on our ability to electrify transportation, buildings, and industry using clean and reliable power.

In the Xie Lab, we think about these challenges together: how can we expand the electric grid, make it low-carbon, and keep it reliable at the same time? Since renewable energy and AI infrastructure both represent fast-growing supply and demand of electricity, decisions about how we build and operate the grid will simultaneously shape both climate outcomes and the future of AI. Therefore, climate mitigation and AI infrastructure are deeply intertwined problems.

What role can AI play in addressing these challenges?

AI is both a driver of electricity demand and a powerful tool for managing the grid more effectively.

Research in generative and agentic AI can significantly improve the productivity of the electricity sector. These tools can reduce the number of human engineer hours needed to maintain reliability, lower operational costs, and help accelerate the interconnection of new generation and data centers—one of the biggest bottlenecks facing the grid today.

The goal is to augment human expertise with real-time intelligence, not replace it.

Can you give an example of AI-enabled grid solutions in practice?

A powerful example comes from my time working in Texas. Over the past 15 years, Texas grew from virtually zero renewable generation to roughly 50 gigawatts of wind and solar—enough to rank among the top electricity producers globally. On average, renewables now supply more than half of the electricity Texans consume.

But that growth created a serious operational challenge. Fluctuating wind and solar power introduced instability and oscillations in the grid, forcing operators to underutilize transmission capacity—like building an eight-lane highway but only being able to use four lanes.

Using AI-driven real-time detection, localization, and control tools, we were able to mitigate these oscillations as they occurred. This allowed operators to fully utilize transmission capacity and deliver gigawatts more clean power from West Texas to load centers like Dallas and Houston. These tools are now deployed in ERCOT's control rooms, delivering tangible, large-scale societal benefits.

What is the Power and AI Initiative at Harvard SEAS?

The Power and AI Initiative at Harvard SEAS brings together researchers from electrical engineering, computer science, climate science, materials science, and the social and behavioral sciences to take a holistic approach to grid modernization.

The initiative is designed to bridge the power sector and the technology sector—helping the grid support rapid AI growth while using AI to make the grid cleaner, more reliable, and more efficient.

What impact do you hope this work will have?

This work is about enabling the next generation of infrastructure—one that supports AI innovation, accelerates decarbonization, and delivers reliable electricity at scale.

If we get this right, AI and power systems can reinforce one another. AI can help modernize the grid, and a modern grid can support the computing demands that will shape the future of science, industry, and society.