Developing Energy-Efficient Solutions for AI Data Centers

The development of artificial intelligence (AI) data centers is on the rise nationwide. Major tech companies are eyeing Pennsylvania as a desired location for these centers in particular because of the state’s significant energy production. This expansion includes the development of two large-scale data centers, and an additional $90 billion in recently announced energy investments.

While this growth can foster job creation, workforce development and an increase in economic activity, energy-related concerns remain.

AI data centers host computing resources needed to meet AI demands. These centers use high levels of water and energy, and depend on local and regional electric systems. This raises questions as to whether electric grids are prepared for the increase in power, if water usage can handle the pressure from the cooling requirements and what the implications are for consumers of energy.

Lehigh’s Center for Advancing Community Electrification Solutions (ACES) is addressing these concerns by taking a full-circle approach in its data center and energy research to answer these outstanding questions, influencing policy and the way AI data centers are developed.

ACES innovates energy independence strategies through community-led electrification—combining cutting-edge technology, practical workforce solutions and common-sense policy to put American communities in charge of their energy future while advancing national strength. This work includes delivering energy solutions to guide and accelerate the growth of AI data centers.

“Pennsylvania is a net energy producer. We are in an attractive location to tech companies that want to get power for their AI data centers,” explains Shalinee Kishore, Iacocca Chair, professor of electrical and computer engineering and director of ACES. “There’s new technology, innovations and solutions that can be brought into the data centers entering our Commonwealth. Our work can help shape Pennsylvania’s commitment to this sector.”

Creating Energy Solutions

Current estimates indicate that by 2028, the electricity needed for AI data centers could be close to 12% of the total electricity demand in the U.S.—a significant increase from 4.4% in 2023.

"We are looking at opportunities to use AI to improve the efficiency, reliability and security of energy systems,” says Kishore. “That’s AI for energy. And then there's the reverse, which is energy for AI. Our team is designing smart, sustainable energy solutions to meet the growing demand for AI infrastructure like data centers.”

With support from ACES faculty and student researchers, Javad Khazaei, assistant professor in the electrical and computer engineering department and a core faculty member of ACES, is leading the research thrust in AI data centers.

Using small scale data center models in the lab and benefiting from real-time data from Lehigh’s High Performance Computing Center, the team is looking at the systems level to understand data center components and energy solutions from start to finish.

Take cooling, for example.

At the device level, there’s direct-to-chip cooling, which removes heat from computer chips, Khazaei explains. Beyond this at the systems level, there are thousands of GPUs running in one data center.

The cooling of this system is energy intensive, he says.

ACES research looks at the overall combination of resources, such as renewable energy from the grid or from wind, solar or battery power, that can be utilized to optimally manage the energy required and provide an energy-efficient and cost-effective way of supporting this component of data centers.

“We have interdisciplinary faculty focused on different aspects of these data centers working together on solutions,” Khazaei says.

Other areas supported by ACES include modeling thermal loads of machine learning models and their impact on power requirements for data centers, microgrid and virtual power plant solutions for managing interconnected AI data centers, heat exchange solutions, waste heat recovery and scalable demand response models to manage peak load and load variability.

"We take all of our focus expertise and we know how to make the ‘LEGO pieces’ click together to solve the bigger problem. That's what I think is a real benefit of ACES,” says Kishore. “Our team draws on expertise on energy management systems, thermal management systems, energy generation, storage and grid interface solutions. This allows us to look at how a variety of energy-saving solutions can be integrated together.”

In addition to Kishore and Khazaei, the ACES research team includes Arindam Banerjee, Paul B. Reinhold Professor and department chair of mechanical engineering and mechanics; Farrah Moazeni, assistant professor of civil and environmental engineering; Carlos Romero, research professor of mechanical engineering and mechanics; Alberto Lamadrid, professor of economics; and Shamim Pakzad, professor and chair of civil and environmental engineering.

Conducting Student Research

Civil and environmental engineering student Maddie Burns ’27 and electrical and computer engineering student Saimonth Muñoz ’26 work with Moazeni on data center direct-to-chip cooling solutions.

This research is very new and exciting, and only continues to grow in importance as AI advances rapidly occur, says Burns.

“I am keenly interested in improving operational efficiency and reducing water consumption in data centers,” says Burns. “Traditional air-cooled systems often rely on water-based cooling towers in a secondary loop to dissipate heat, even though the primary cooling appears to be air-based. In contrast, direct-to-chip liquid cooling allows heat to be absorbed more efficiently at the source, minimizing the volume of coolant and potentially eliminating the need for water-intensive cooling infrastructure, thus reducing overall water usage.”

In an age of rapidly growing computing demands, improving how we cool chips isn’t just technical, but it’s essential for sustainable innovation, Muñoz adds.

At the systems level, electrical engineering student Refia Temizkan ’28 is applying advanced modeling tools to develop integrated sizing strategies for backup generators and distributed energy resources.

“Data center electricity demand is accelerating, and even brief power interruptions can trigger six-figure losses,” she explains. “Integrating cleaner, distributed resources alongside traditional backup is therefore becoming a strategic imperative rather than a nice-to-have. By coupling these tools, I will generate techno-economic scenarios that explicitly balance cost, reliability and sustainability.”

The outcome, she hopes, will be a decision-support framework that helps data center operators meet surging power demands, cut outage risk and progress toward carbon-reduction targets—all while maintaining competitive total cost of ownership.

Taking the Next Steps

ACES is at the forefront of developing technology and policy solutions that enhance energy efficiency while ensuring grid resilience and cost.

In October, the center will host a symposium bringing together research universities, community colleges, industry leaders, municipalities and the state government to discuss the energy challenges of data centers.

ACES will bring its expertise in optimized electrification technology and policy solutions across a range of sectors to secure the future of energy systems in everyday communities, including those that are, or soon will be, homes to data centers.

The ACES vision centers on communities and partnerships with stakeholders in a variety of areas to develop electrification pathways to achieve targeted goals for cost, water use, emissions, energy security and reliability, and grid efficiency. Through the center’s goals, ACES aims to enable energy independence and resilience for all kinds of communities by reducing time and risk while accelerating the impact of electrification.