Is DeepSeek really better for the environment than ChatGPT and Gemini?

When Chinese entrepreneur Liang Wenfeng revealed DeepSeek’s newest app to the world earlier this month, most people had never even heard of the artificial intelligence company. But the new app took the world by storm, as many in the tech community marveled at how DeepSeek functioned at a fraction of the cost of other large language models like OpenAI’s ChatGPT and Google’s Gemini. That’s presumably good news for the environment, as many have criticized the AI craze as being extremely taxing on electrical grids — so much so that some tech companies like Google and Meta have reopened coal plants.

In theory, any AI alternative that consumes fewer resources should be better for the environment. Yet when Salon reached out to experts about the potential promise in DeepSeek’s potential “Sputnik” moment (to quote billionaire software developer Marc Andreessen), they expressed cautious optimism.

“There is almost no information available about either ChatGPT or DeepSeek, so any numbers [about their environmental impact] are speculation,” David Rolnick, an assistant professor in the School of Computer Science at McGill University, told Salon. “It's unlikely that there is a massive difference between these two algorithms, but it's hard to say the biggest differences in terms of the energy consumption aspects of large AI algorithms are how much you use.”

That said, DeepSeek could still represent a step in the right direction, at least in terms of sustainability. Climate scientists  are rightfully worried about the huge energy drain from data centers in the U.S., which is expected to either double or triple by 2028. Rolnick explained that the environmental impacts of very large AI models include the energy use associated with developing training and querying them, water usage for cooling data centers and impacts associated with manufacturing hardware like servers and chips.

Yet Rahul Sandhil, the vice president and general manager for global marketing and communications at the semiconductor company MediaTek, told the Associated Press that DeepSeek offers hope for reducing those impacts. It is known as an “open-weight” model, which means it can be downloaded and run locally, assuming one has the sufficient hardware. If their claims hold up, some routine AI queries in the future may not need data centers at all and could instead be shifted to phones. The fact that the LLM is open source is another plus for DeepSeek model, which has wiped out at least $1.2 trillion in stock market value. Meanwhile, companies like OpenAI, which have been operating for a decade, have yet to be profitable.

What it ultimately comes down to is whether people actually use it over other models. It doesn’t matter if it’s “better” for the environment if it doesn’t see much adoption. Additionally, as noted by Prof. Anthony Cohn at the University of Leeds and Foundation Models Lead at the Alan Turing Institute, said in a statement “the DeepSeek models are still language only, rather than multi-modal – they cannot take speech, image or video inputs, or generate them.  No doubt the future will see such a release, though the computational demands of handling multi-modal data are much greater than when just handling language.”

Want more health and science stories in your inbox? Subscribe to Salon's weekly newsletter Lab Notes.

That could explain it’s seemingly lesser impact on the planet. On a financial level, how DeepSeek functions with fewer resources will raise unavoidable sustainability questions when other AI companies attempt to succeed using more consumptive models.

“The DeepSeek announcement should prompt regulators all across the country to feel empowered to rigorously interrogate forecasts used to justify lock-in of fossil fuel infrastructure,” Julie McNamara, deputy policy director with the Climate & Energy program at the Union of Concerned Scientists, told Salon. In contrast to claims that AI demands massive investments of money and electricity — often hand-waved away as a “trade off” for the bright future this tech will allegedy usher in — policymakers and investors should instead acknowledge that AI could be effectively developed without a catastrophic carbon footprint.

“There is no question that electricity demand will be ratcheting up over the years and decades to come, regardless of data center needs,” McNamara said. “But this is a transition that must be approached with a necessarily long view, built on strong foundations that ensure the needs of the public are elevated throughout.”

Additionally, policymakers and investors should also ask which AI programs are truly necessary. If many routine AI functions can be performed without massive data centers, then limiting AI use to situations when it is necessary and irreplaceable could at least somewhat mitigate its environmental harms.

“Fundamentally, society has to decide how much you want to use an immense generative AI algorithm, and I think that as a whole, we haven't yet worked out whether they are useful in the vast majority of cases in which they're being used,” Rolnick said. People can often perform Google searches or ask humans to perform tasks that AI business leaders argue should be outsourced to AI.

“It is unclear to me that the vast majority of uses of algorithms like DeepSeek and ChatGPT are providing benefits in many places,” Rolnick said. “If we are to think about environmental benefits associated with one model or another, these will pale in comparison to the decision whether or not to use such AI algorithms as compared either with non-AI approaches.”

McNamara acknowledges that — with the fossil fuel industry’s political allies in control of the presidency, legislature and courts — this will not be easy.

“Recently, the massive energy requirements associated with AI development have been invoked to justify the build-out of scores of new gas-fired power plants,” McNamara said. “But such a turn is entirely out of step with the true needs of a forward-looking economy.” That is why, above all else, the unveiling of DeepSeek is relevant for inhabitants of Earth as an opportunity to challenge narratives about humanity’s use of fossil fuels.

“Fossil fuel companies are seizing this moment to attempt a blatant end-run around critical climate and public health standards to lock-in new gas infrastructure for decades to come,” McNamara said.