Aviv Kesar | Mustang News
by Kaylin O'Connell / April 18, 2025
Artificial intelligence is breaking new ground for Cal Poly under the Cal State AI initiative, which announced its collaboration with OpenAI to make AI tools free for Cal State students. However, the initiative raises environmental concerns due to energy use, water consumption and a lack of transparency from big tech companies.
The Cal State AI initiative negatively contributes to the environment due to the influx of new customers supporting large language models (LLMs), which are AI systems like ChatGPT that use high amounts of energy and water, according to Foaad Khosmood, a computer science & software engineering professor.
“If you have half a million new customers, that’s more incentive for those companies and other companies to create new LLMs,” Khosmood said. “Now they think they can sell it to all these universities and make even more money.”
The Cal State initiative encourages students to use AI without knowledge on its environmental effects, according to Amara Zabback, a computer science & software engineering graduate student. Zabback gave a presentation titled “Environmental Impact of LLMs and Gen AI” in a graduate class on AI.
“As soon as universities or university systems are encouraging [AI use], it’s just putting way more support behind it,” Zabback said. “They need to try a lot harder to hold these people accountable and hold companies accountable.”
Cal Poly remains committed to its environmental goals, according to Cal Poly Spokesperson Matt Lazier.
“As this is a new initiative, Cal Poly and the CSU will be looking closely at AI’s positive and negative impacts,” Lazier said in an email to Mustang News. “The CSU AI Strategy seeks to be a leader in the ethical, social and responsible use of AI in education, and our commitment to sustainability will be a part of those discussions.”
The bulk of AI’s environmental impact stems from the LLM training, as opposed to individuals’ ChatGPT searches, as the training creates higher carbon emissions and requires exponential amounts of water and power, according to Khosmood.
LLMs use small hardware engines to process significant power, which leads to excessive heating and requires water and refrigeration systems to keep the engines cool, Khosmood said.
Data centers can use up to five million gallons of water daily, intensifying water shortages in hot, dry climates like Arizona and Texas, Zabback said. These areas are typically water scarce, so the data centers negatively impact nearby communities and ecosystems, she explained.
Researchers in 2019 found that an LLM could produce as much [carbon dioxide] as five cars over their lifetimes.
“It’s a really cool and exciting technology, but I think that it needs to be used responsibly,” Zabback said. “I think people who use it need to have a profound understanding of what is actually happening.”
Major companies like Microsoft, a leading stakeholder in OpenAI, have pledged to become carbon negative by 2030, Zabback said. However, Khosmood is skeptical of the company’s claim, as carbon offset and water recycling efforts typically have little follow through or confirmed impact from companies, he explained.
“It’s really hard to pinpoint and exactly qualify [LLM’s] carbon emissions,” Zabback said. “It’s not like these data centers are out here advertising the resource use.”
Khosmood believes that increased transparency, both from big tech companies and the Cal State system, is necessary to increase public understanding about the environmental impacts of LLMs.
“When you go to the supermarket and you buy something, the ingredients are listed on there, right?” Khosmood said, holding up a bottle of Coke Zero. “This is all by law. We have to do this, right? Well, they should have to do the same thing for digital products.”
Specifically, Khosmood believes that LLM sites should be open about their training locations, carbon intakes and fine-tuning methods. He emphasized that this should exist at every level, from federal to the Cal State system.
“[LLMs] should publish all the ingredients that went into this and what impact it’s having,” Khosmood said. “It’s not too much to ask, I think. This could be making us sick. This could be making the planet sick.”
Zabback believes that innovation does not need to come at an environmental cost; the balance is more nuanced.
“I think that we can have a lot of really good innovations that can change peoples lives and change things for the better,” Zabback said. “But I don’t think that we need to steamroll the environment in the meantime.”