Bits & Watts adds eight new research projects to its sustainable AI infrastructure research portfolio

Stanford University's Bits & Watts Initiative has selected eight new seed grant projects, expanding its growing Sustainable AI Infrastructure Research Portfolio. The projects bring together researchers from engineering, computer science, business, environmental science, and the social sciences to address one of the defining challenges of the AI era: building the infrastructure needed to support AI while ensuring reliable electric power, environmental sustainability, and community benefit.

Over the past two years, Stanford and the Bits & Watts Initiative have built a nationally recognized Powering AI research program that has informed policy discussions on electric grid utilization, catalyzed more than 20 interdisciplinary research projects, and strengthened collaboration among industry, academia, and government. The new projects expand this portfolio with research spanning resilient electrical infrastructure, advanced data center technologies, environmental stewardship, and community-centered approaches to AI infrastructure development.

“Powering AI is one of the defining energy challenges of this decade,” said Liang Min, managing director of the Bits & Watts Initiative. “No single discipline can solve it. These projects bring together expertise across engineering, energy systems, economics, environmental science, and the social sciences to develop practical solutions that strengthen our electrical infrastructure, improve the sustainability of AI, and inform decisions by utilities, technology companies, policymakers, and local communities. We’re particularly excited to expand our portfolio with research that incorporates community perspectives, recognizing that successful AI infrastructure depends not only on technological innovation but also on public engagement and social acceptance.”

Community and sustainability perspectives for AI infrastructure

Data Center Deliberation: Streamlining Development with AI-Driven Stakeholder Engagement

Principal Investigator: Sarah Billington, Civil & Environmental Engineering
Co-Principal Investigators: Ashish Goel (Management Science & Engineering), Rishee Jain (Civil & Environmental Engineering), Alice Siu (Freeman Spogli Institute for International Studies), Gabrielle Wong-Parodi (Earth System Science)

As AI infrastructure expands across the United States and globally, communities are increasingly raising questions about electricity demand, water use, environmental impacts, and local economic benefits. Yet many public engagement processes provide limited opportunities for residents to understand complex technical issues or contribute meaningfully to planning decisions. This project will adapt Stanford's AI-assisted Online Deliberation Platform to support informed discussions about data center development. Participants will receive balanced educational materials before joining AI-supported deliberative discussions designed to encourage equitable participation and informed dialogue. Working with local and international partners, the research team will evaluate how deliberative engagement influences public understanding, trust, and policy recommendations. The project aims to create a scalable model that helps governments, developers, and communities work together to make more informed, transparent, and broadly supported decisions about future AI infrastructure.

Evaluating Carbon, Water, and Biodiversity Trade-offs of Co-locating Data Centers and Pumped Storage Hydropower

Principal Investigator: Sebastian Heilpern, Earth System Science

As demand for AI computing grows, so does the need for reliable, low-carbon electricity. Pumped-storage hydropower offers an important source of firm energy, but its environmental impacts vary considerably depending on where projects are built. This project will evaluate opportunities to co-locate data centers with pumped-storage hydropower while minimizing impacts on carbon emissions, water resources, and biodiversity. By integrating nationwide datasets on hydrology, energy infrastructure, projected data center growth, and freshwater ecosystems, the researchers will identify locations where new infrastructure can deliver reliable electricity with fewer environmental trade-offs. The resulting planning framework will help utilities, developers, and policymakers better understand how future AI infrastructure can support both energy reliability and long-term environmental sustainability.

Building a resilient electrical infrastructure for AI

Flexible Queues for the AI Era: Designing Location- and Flexibility-Based Interconnection Mechanisms for Large Loads

Principal Investigator: Omer Karaduman, Graduate School of Business

As utilities receive growing numbers of requests to connect large AI data centers to the electrical grid, existing interconnection processes are becoming increasingly congested. Long queues can delay both new computing facilities and other electricity customers while increasing costs across the system. Building on previous research into power system interconnection queues, this project will develop new market and operational mechanisms that prioritize well-sited and operationally flexible AI facilities. The research will examine how pricing, location, and flexibility commitments can improve the efficiency of interconnection queues while making better use of existing electrical infrastructure.

A Decision-Informed Tool for Reliable, Affordable, Secure, and Sustainable Power for AI

Principal Investigator: Inês Azevedo, Energy Science & Engineering
Co-Principal Investigators: Steve Davis (Earth System Science), Carlos Díaz-Marín (Energy Science & Engineering), Edgar Virgüez (Energy Science & Engineering)

Choosing where to build data centers—and how to power them—will significantly influence electricity costs, emissions, reliability, and future investments in the electric power system. The research team will develop an interactive decision-support platform that allows users to compare alternative data center locations and electricity supply strategies across the United States. The tool will evaluate multiple energy technologies while assessing costs, greenhouse gas emissions, resilience, air-quality impacts, and reliability.

Investigating the Interconnection of Power Grid Capacity Expansion with Data Center Siting and Operation

Principal Investigator: Meagan Mauter, Civil & Environmental Engineering Collaborator: Bolun Xu, Columbia University

The rapid growth of AI data centers is expected to reshape future investments in electricity generation, transmission, and storage. This project will integrate data center siting, operational flexibility, and energy efficiency with advanced capacity expansion models to better understand how AI demand drives changes throughout the electrical system. The work will identify strategies that reduce costs, emissions, and water use while improving long-term reliability.

Advancing efficient AI infrastructure

Cooling and Waste Heat Utilization Co-Design for Sustainable Data Centers

Principal Investigator: Carlos Díaz-Marín, Energy Science & Engineering

Data centers generate enormous quantities of heat that are typically treated as waste. This project explores how advanced cooling technologies can transform waste heat into a valuable resource by enabling carbon capture, atmospheric water harvesting, and district heating while reducing energy and water consumption. The research seeks to identify new approaches that could make future AI data centers significantly more sustainable.

Full-Stack Optimization of AI Power Delivery through High-Voltage Transmission and Wide-Bandgap Semiconductors

Principal Investigator: Srabanti Chowdhury, Electrical Engineering

This project will develop a comprehensive model of power delivery from the electric grid to AI processors, evaluating how higher-voltage transmission systems and next-generation wide-bandgap semiconductor devices can improve efficiency across the entire power delivery chain. The resulting framework will help guide future electrical infrastructure for AI.

A Power Distribution Unit to Protect the Grid from Machine-Learning Training Transients

Principal Investigators: Philip Levis (Computer Science & Electrical Engineering), Ram Rajagopal (Civil & Environmental Engineering & Electrical Engineering), Juan Rivas-Davila (Electrical Engineering)

Building on an earlier Bits & Watts seed project, this research team will complete and evaluate a new power distribution unit designed to smooth the rapid power fluctuations created during large-scale AI model training. The technology offers a practical, energy-efficient approach to reducing stress on both data centers and the electrical grid as AI computing continues to scale.

Learn More

Bits & Watts will host a webinar on August 26 featuring the principal investigators from the newly funded seed grant projects. The webinar will highlight each project and explore how Stanford researchers are advancing practical solutions for sustainable AI infrastructure.