Project aims to put AI to work reducing its own energy needs

Overview

UC Merced is spearheading a research effort to make artificial intelligence more energy efficient by using AI itself to optimize the way modern computing systems use power. The initiative focuses on reducing the energy intensity of training and running AI models—work that carries growing environmental and cost implications as AI adoption accelerates across industries.

The project aims to put AI to work reducing its own energy needs.

What the Project Will Do

The research centers on strategies that allow AI systems to learn when, where, and how to consume less energy without sacrificing accuracy or speed. Expected directions include:

Intelligent workload placement and scheduling that matches computation with available, lower-carbon power.
Software-level efficiency improvements—such as model optimization and dynamic precision—that reduce compute cycles.
System-level orchestration that coordinates processors, memory, and cooling to avoid waste and throttle power draw in real time.

By treating energy as a first-class optimization target, the team aims to cut consumption during both model training and inference, the two most power-hungry phases of AI deployment.

Why It Matters for AI and Technology

The power required to build and operate advanced AI models is rising quickly, straining data center capacity and complicating climate goals. Making AI less energy intensive:

Lowers operating costs for organizations deploying AI at scale.
Reduces greenhouse gas emissions associated with electricity use.
Enables AI to run more feasibly on edge devices and in resource-constrained environments.
Helps cloud and hyperscale providers meet efficiency and sustainability targets without curbing innovation.

For the broader tech ecosystem, energy-aware AI can unlock wider adoption in sectors like healthcare, agriculture, logistics, and public services by shrinking the hardware footprint and power budgets needed to deliver results.

Stakeholders and Collaboration

Led by researchers in computer science and engineering at UC Merced, the project is positioned at the intersection of AI, high-performance computing, and energy systems. Academic-industry collaboration is likely to play a central role, ensuring that techniques devised in the lab can be integrated into real-world software stacks and data center operations.

Central Valley Relevance

Located in California’s Central Valley, UC Merced serves communities that both stand to benefit from AI-driven innovation and are directly affected by energy reliability, water availability, and climate resilience. Energy-efficient AI can:

Ease pressure on regional power infrastructure during peak demand.
Reduce cooling and water needs associated with compute-intensive workloads.
Support local industries—especially agriculture and logistics—with smarter, lighter-footprint AI tools that operate effectively at the edge or in smaller facilities.

What to Watch Next

Key milestones to look for include prototype tools that automatically tune AI workloads for energy efficiency, benchmarks demonstrating reductions in power use without accuracy loss, and pathways for integrating these methods into common machine learning frameworks and cloud platforms. Successful demonstrations could inform best practices for both public research institutions and private data center operators.

Central Valley AI is produced by the CVAI Newsdesk team and developed by Kaweah Tech, a regional firm that builds, deploys, and integrates AI solutions for businesses across California's Central Valley.

Source

https://news.ucmerced.edu/news/2026/project-aims-put-ai-work-reducing-its-own-energy-needs