The solution to AI's massive energy appetite won't come from one revolutionary invention, but from thousands of engineers making small, strategic improvements across every layer of computing systems. That's the key insight from two Colorado State University Ph.D. students whose research on data center thermal optimization just earned the Best Paper Award at the Association for Computing Machinery (ACM) International Green and Sustainable Computing Conference.

What's Driving the Need for Energy-Efficient AI Computing?

The rapid expansion of artificial intelligence is placing unprecedented strain on data center infrastructure. Modern AI systems demand massive amounts of power, cooling, and water, creating an urgent need to rethink how we design and operate computing facilities. As AI and cloud computing continue to scale globally, improving energy efficiency has become essential not just for environmental reasons, but for the economic viability of next-generation infrastructure.

Jason Crop and Hayden Moore, the award-winning researchers, are tackling this challenge from different angles. Crop balances his Ph.D. research with a full-time role as a power thermal design engineer at Broadcom, where he leads efforts in power and thermal design for advanced three-dimensional integrated circuit (3DIC) based AI systems. Moore's passion for green computing began during his undergraduate studies and has driven him to pursue advanced research in the field.

How Can Engineers Make AI Data Centers More Efficient?

Rather than waiting for a single technological breakthrough, the researchers emphasize that progress will come from coordinated improvements across the entire computing ecosystem. This means optimizing everything from the semiconductor devices themselves to processor architecture, system design, cooling systems, and data center operations.

• Semiconductor and Device Level: Improvements in how chips are designed and manufactured to reduce power consumption at the source.
• Processor Architecture: Enhancements to how processors are structured to execute AI workloads more efficiently.
• System Design and Cooling: Better thermal management systems that prevent heat buildup and reduce the energy needed for cooling.
• Data Center Operations: Smarter management of how data centers run, including workload scheduling and resource allocation.

"The computing demands of the next decade will not be solved by a single breakthrough in architecture or semiconductor technology. Instead, progress will come from thousands of engineers making small improvements across every level of the computing stack," said Jason Crop, Ph.D. student at Colorado State University.

Jason Crop, Ph.D. Student and Power Thermal Design Engineer at Broadcom

Crop's award-winning paper, titled "Revisiting 'Cooler is Better': ITD-Aware Per-CPU Thermal Optimization for Sustainable Data Center Operation," focuses specifically on thermal optimization. The research demonstrates how careful attention to temperature management at the processor level can yield meaningful energy savings when scaled across entire data centers.

Why Does Collaboration Matter in Solving the Energy Problem?

Both researchers emphasized that solving complex engineering challenges requires more than individual brilliance. Collaboration across academia and industry, combined with creative problem-solving and sustained commitment, drives the innovations needed to advance the field. Crop's unique position, working both in industry and academia, allows him to identify real-world challenges while exploring new ideas that could benefit future computing systems.

"It was a very rewarding experience working with other passionate people to come up with creative solutions. If you want to make an impact on the state of your field and on the wider world, research is a way to accomplish that," said Hayden Moore, Ph.D. student at Colorado State University.

Hayden Moore, Ph.D. Student at Colorado State University

The research was conducted under the guidance of Electrical and Computer Engineering Professor Sudeep Pasricha, whose mentorship helped shape the direction and rigor of their work. This collaborative model, combining academic research with industry expertise, appears to be essential for translating theoretical improvements into practical solutions.

What Does This Mean for the Future of AI Infrastructure?

The implications of this research extend beyond academic recognition. As AI systems become increasingly central to business operations and scientific research, the energy efficiency of data centers directly affects both operating costs and environmental impact. The message from these researchers is clear: there is no silver bullet. Instead, the path forward requires sustained effort from thousands of engineers working across different specializations and organizations.

Both Crop and Moore are calling for more researchers to enter the field of energy-efficient computing. Crop noted that the industry needs engineers who are excited to tackle these challenges and help build the next generation of sustainable computing systems. Moore encouraged future students to pursue advanced degrees in electrical and computer engineering to push the frontiers of what's possible in this critical area.

The recognition of their work at the ACM conference signals that the computing industry is beginning to prioritize incremental efficiency gains as a legitimate and valuable research direction. Rather than waiting for transformative breakthroughs that may never arrive, the field is embracing the reality that sustainable AI will be built through countless small improvements, each contributing to a larger shift toward greener computing infrastructure.

" }