Even if it’s just a bubble, AI development is already very costly. Harms to the climate – already under a lot of strain – could wipe out years of progress on emissions reductions. Are there ways to make the current bonanza more sustainable?
Almost every new study on AI’s environmental price tag reaches a similar conclusion: powering the generative models, which are being embedded everywhere, unleashes a huge amount of carbon emissions.
Just this week, The Guardian quoted Sharon Wilson, a former oil and gas worker who has documented methane releases for more than a decade, saying that xAi’s Colossus data center in Memphis is spewing more emissions than a large power plant.
And late last year, a comprehensive study by scientists at Cornell University, published in Nature Sustainability, suggested that AI server deployment across the US alone could generate between 24 and 44 million metric tons of carbon dioxide annually by 2030.
That’s the same as adding 5 to 10 million cars to US roads. Water consumption should also rise massively and equal the annual household water usage of 6 to 10 million Americans.
Is it surprising? Not really. AI computing is expanding rapidly – as is the demand for specialized (and enormous) data centers equipped with power-hungry graphics processing units.
Data centers powered by diesel generators
Cornell scientists also say that ambitious net-zero emissions targets by tech giants are unrealistic. Indeed, Google or Meta usually claim carbon neutrality after doing some creative accounting, when in fact, their emissions could already be surging.
Some are urgently calling for a pause. In October, the United Nations’ special rapporteur on the human right to safe drinking water called for a moratorium on the development of new data centers.
And in December, more than 230 environmental groups in the US demanded a national moratorium until they were regulated. With US President Donald Trump now busy with America’s favorite pastime, regime change, one can only wish these activists good luck.
Anyway, Trump recently signed an executive order that creates a single national AI law and overrides any existing regulations in separate states. Big tech firms have applauded the move – after, of course, frantically lobbying for it.
Sasha Luccioni, climate lead at AI company Hugging Face, told The Guardian she was frustrated by selective disclosures from big companies allegedly caring about the climate impact of their products.
“I still believe that AI can do good in terms of fighting the climate crisis – designing the next generation of batteries, tracking deforestation, predicting hurricanes,” said Luccioni.
“There are so many good things for which we can be using it – and instead we’re creating social media websites filled with AI slop while data centers are getting powered by diesel generators.”
Google or Meta usually claim carbon neutrality after doing some creative accounting, when in fact, their emissions could already be surging.
The International Energy Agency projects global data center electricity demand will more than double by 2030 to around 945 terawatt-hours, slightly exceeding Japan's total energy consumption.
And indeed, Goldman Sachs Research forecasts that approximately 60% of this increased demand will be met through fossil fuels, potentially adding 220 million tons to global carbon emissions. A sort of new fracking boom, it seems.
Choose your state, choose your chips
According to the report by Cornell University scientists, the industry energy consumption could double already by 2026, “motivated by AI and other sectors, threatening decarbonization targets under the Paris Agreement, which include a 53% reduction in data-centre emissions by 2030 and net-zero goals for the AI sector.”
However, there’s room for improvement, experts add. For instance, cheap and effective innovations such as DeepSeek can lead to low power requirements for AI computing tasks.
Furthermore, to mitigate climate risks specifically for the US, Cornell experts – while recognizing that AI advancement is important – say that concentrating AI server deployment in Midwestern states such as Texas, Montana, Nebraska, and South Dakota would be optimal.
That’s because of their “abundant renewables, low water scarcity and favourable projected unit water and carbon intensities.”
Besides, these states possess substantial untapped wind and solar resources, enabling robust green power portfolios and reducing competition with other sectors. Their lower water stress also helps ease public concerns.
In December, another new study by Cornell Tech, IBM, and Rensselaer Polytechnic Institute explored one more promising solution: analog in-memory computing (AIMC), utilizing analog chips.
Midwestern states possess substantial untapped wind and solar resources, enabling robust green power portfolios and reducing competition with other sectors. Their lower water stress also helps ease public concerns.
Unlike traditional architectures, which constantly move data back and forth between memory and processors, AIMC stores and processes data in a single location.
“This leverages physics to perform the math calculation instantly without moving the data, potentially slashing power consumption by 1,000 times and making the next generation of AI sustainable,” said Tyanyi Chen, associate professor of electrical and computer engineering at Cornell Tech.
This breakthrough could be transformative, Chen added, since it would make it practical to train and fine-tune large AI models with far less energy and cost.
That could enable the use of applications that are currently out of reach for power-hungry systems, from healthcare devices and wearable technology to industrial sensors and autonomous robots.
Unlock exclusive Cybernews content on YouTube.
Your email address will not be published. Required fields are markedmarked