AI and climate: Cornell backs eight fast grants to cut compute costs and advance environmental research
As AI energy use climbs and climate pressures intensify, Cornell has launched an inaugural round of AI and Climate Fast Grants to fund work at the intersection of AI and environmental science. The 2030 Project: A Cornell Climate Initiative, the Cornell AI Initiative, and the Cornell Atkinson Center for Sustainability are coordinating the effort.
"AI can be part of the solution for climate sustainability, but it is also part of the problem," said Thorsten Joachims, Jacob Gould Schurman Professor of Computer Science and Information Science and director of the Cornell AI Initiative. The call addresses both sides: reduce AI's footprint and apply AI to climate challenges.
Recent Cornell Engineering analysis shows that at current growth rates, AI could add 24-44 million metric tons of CO2 by 2030 and consume 731 million-1.125 billion cubic meters of water. For researchers building and deploying models, those numbers make latency, throughput, and energy trade-offs a front-line design choice.
"Artificial intelligence is changing society," said Benjamin Houlton, Ronald P. Lynch Dean of the College of Agriculture and Life Sciences (CALS). He emphasized a people-centered approach that weighs risks and benefits, trims AI's energy demand, and delivers outcomes for the public good.
CALS and Cornell Atkinson jointly oversee The 2030 Project, which is disbursing $10,000-$25,000 per team. The Cornell AI Initiative, launched in 2022, supports core AI research and cross-campus collaborations.
What's being funded
One recipient, Udit Gupta (assistant professor, electrical and computer engineering, Cornell Tech), is developing "EcoGPT," a generative AI interface that lets users accept slightly slower responses in exchange for lower carbon and water use. Industry benchmarks suggest that relaxing response generation by even a couple hundred milliseconds can improve overall system throughput and energy efficiency by up to 2.5x-evidence of a strong latency-efficiency trade-off.
- AI to strengthen forest integrity in Cambodia - led by Dena Clink (K. Lisa Yang Center for Conservation Bioacoustics, Cornell Lab of Ornithology).
- Low-cost active climate monitoring with underactuated agents - led by Sarah Dean (Computer Science, Cornell Ann S. Bowers College of Computing and Information Science) and Alexander Terenin (Center for Data Science for Enterprise and Society, Cornell Research & Innovation).
- AI-enabled synthesis of unconventional solar-cell materials - led by John Marohn (Chemistry and Chemical Biology, College of Arts and Sciences) and Michael Lawler (Physics, College of Arts and Sciences).
- Confidence in additive construction - led by Sriramya Nair (Civil and Environmental Engineering) and Nils Napp (Electrical and Computer Engineering), both in Cornell Engineering.
- Robotic eDNA sampling of waterways - led by Kirstin Petersen (Electrical and Computer Engineering, Cornell Engineering) and Peter McIntyre (Ecology and Evolutionary Biology, CALS).
- Motivating climate action using human-AI dialogues - led by David Rand (Information Science; Marketing and Management Communication, SC Johnson College of Business; Cornell Bowers; College of Arts and Sciences) and Gordon Pennycook (Psychology, College of Arts and Sciences).
- Resilience scanner for municipal climate adaptation - led by Anthony Townsend (Cornell Tech).
Why this matters for researchers
- Treat latency as a carbon knob: small response delays can yield large energy savings without hurting user outcomes. Test where your users tolerate delay and quantify efficiency gains.
- Measure before you optimize: track GPU-hours, model size, data center PUE, and, where possible, water use. Report per-inference and per-training-run footprints.
- Prioritize model and system efficiency: right-size models, batch requests, cache results, and select hardware appropriately. Consider mixed precision and sparsity where feasible.
- Design service tiers: offer "green" modes with slightly slower responses for research workflows that don't need instant results.
- Apply AI where it moves the needle: monitoring, materials discovery, construction quality, biodiversity sampling, behavior change, and city-scale adaptation.
Learn more
Program details and collaborations are coordinated through the Cornell AI Initiative and The 2030 Project.
Cornell AI Initiative
The 2030 Project: A Cornell Climate Initiative
Your membership also unlocks:
