October 8, 2026, (Inside AI) — Data centers worldwide are consuming electricity at an unprecedented pace, and the environmental fallout is becoming impossible to ignore. Power-hungry facilities are straining electrical grids and forcing utilities to lean on fossil fuels to meet demand. But one researcher at the Massachusetts Institute of Technology is betting that artificial intelligence can help solve the very problem that AI itself is accelerating.
Christina Delimitrou, a newly tenured associate professor in MIT’s Department of Electrical Engineering and Computer Science, is leading a group that applies machine learning to make large-scale cloud computing systems more efficient, secure, and reliable. Her work focuses on squeezing more computational power out of existing hardware rather than building new data centers, a strategy that could significantly reduce the industry’s energy footprint.
“If data centers are not utilized to the best of their capabilities, then they will burn much more power than they need to meet growing user demand,” Delimitrou says. “There is a lot of bloating, especially on the software side of these systems. If we can remove that bloating in a way that doesn’t compromise performance, then we won’t need to build as many new data centers.”
Delimitrou holds the KDD Career Development Associate Professor in Communications and Technology chair and is a member of the Computer Science and Artificial Intelligence Laboratory (CSAIL). Her research tackles a problem that has become more urgent as tech companies race to build massive data centers for AI training and inference. According to the International Energy Agency, data centers consumed roughly 460 terawatt-hours of electricity in 2022, and that figure could double by 2026. Much of that power still comes from fossil fuels, making efficiency gains a critical lever for reducing emissions.
Read: AI Data Center Backlash Reaches San Jose, Silicon Valley's Biggest City
From Underutilization to Intelligent Management
Delimitrou’s interest in cloud efficiency began during her graduate studies at Stanford University. Working with her mentor Christos Kozyrakis, the Leonard Bosack and Sandy K. Lerner Professor of Engineering, she discovered that many large computing systems were running at only about 15 percent capacity.
“You would expect, with all the demand for these systems, that they should be running close to 100 percent capacity. But we found that most were running at only about 15 percent capacity,” she says. “This is not a resource-efficient or sustainable way of scaling these systems.”
To address this, Delimitrou turned to machine learning. At the time, applying ML to large-scale system problems was a novel and risky approach. “Applying machine learning to solve a large-scale system problem was a novel approach at the time. It was a bit risky because people had not yet shown that these techniques would work,” she says. “But empirical approaches require a lot of expertise, and the scale of the system is so large that it is difficult for users to manage. This is why machine learning is often the best solution.”
One tool her group developed, called Seer, uses deep learning to predict and prevent performance problems in web applications before they occur. This proactive approach avoids the widespread slowdowns that can happen when developers try to fix issues manually. Delimitrou also uses AI to help programmers find and fix bugs in cloud-based applications, reducing downtime that wastes computational resources.
Her work has evolved as cloud applications have changed. Developers now split applications into smaller microservices spread across multiple servers, a shift that increases deployment speed but often clashes with hardware designed for older architectures. “But the servers were not built for this new style of application design. So, I rethought some of my earlier work to build machine-learning systems for this new class of applications,” she says.
Read: Nvidia, Broadcom Shielded as AI Power Crunch Hits Chip Supply Chain, Says Morgan Stanley
Delimitrou joined MIT in 2022, drawn by the opportunity to collaborate with leading hardware and software engineers. She teaches 6.191 (Computation Structure), a popular undergraduate course with about 350 students each semester. “I want the students to learn how to think and learn on their own. Part of that involves shifting away from formulaic assignments and making classes more open-ended. I’d rather give the students something to make them think more deeply,” she says.
A key challenge in her research is that AI models used to manage systems are often opaque. “One of the challenges when it comes to applying AI to these systems is that the AI is not interpretable,” she says. “A lot of the work we are doing now involves adding explainability into these AI tools so people can get useful feedback from the system.”
That explainability matters not only for trust but for future design. By understanding why an AI made a particular decision, engineers can refine both the AI and the underlying hardware. Delimitrou’s group also creates clones of proprietary systems, such as a tool called Ditto, which mimics an application’s structure and performance. This allows academic researchers to study real-world cloud environments without access to confidential corporate hardware and software.
“But you still have to use AI carefully. While it can greatly accelerate the application development side, we still need to audit it and be especially careful about how these models are applied so we don’t lose the ability to gain insights out of the solutions AI is giving,” she says.
Delimitrou’s path to MIT began in a midsized town in northern Greece, where she was inspired by the country’s ancient mathematical tradition and her parents, a chemical engineer and a pharmacist. She studied computer engineering at the National Technical University of Athens and later earned her PhD at Stanford. Outside the lab, she enjoys gardening, painting, and playing classical piano. She and her husband have a 1-year-old daughter.
As AI models grow more advanced, Delimitrou expects her work to keep shifting. The environmental stakes are high, and the pressure to make data centers more sustainable will only intensify. Her research offers a pragmatic path: use AI to optimize the systems AI itself relies on.
