Stay up so far with free updates
Simply register for Technology sector myFT Digest – delivered straight to your inbox.
One of the boldest guarantees made about artificial intelligence is that it’ll enable us to tackle the world's best challenges, similar to climate change. AI may also help run smarter power grids, develop more efficient electric vehicles, and Plastic pollution in our oceansBut the information centers that host the most recent AI models eat staggering amounts of energy and water. Given the climate emergency, is AI an issue relatively than an answer?
How opposing technologies interact is demonstrated by the experience of Microsoft, which in 2020 made one in every of the boldest environmental commitments in its history. By 2030, the technology company promised, it might be carbon negative and by 2050 it might have offset all of the emissions it has caused since its founding in 1975. But Microsoft still has an extended method to go. Last month, the corporate reported that its emissions had increased by 29 percent since 2020 because it continued to speculate heavily in data infrastructure.
This week, the corporate announced that it’ll invest $3.2 billion over the subsequent two years to expand its cloud computing infrastructure in Sweden. In total, Microsoft plans to speculate greater than $50 billion in data centers this yr, One analyst said “the most important infrastructure expansion ever seen by mankind.” Data consulting firm Gartner predicts that global spending on data centers will increase by 10 percent to $260 billion this yr.
In addition, Microsoft and OpenAI, the AI startup heavily backed by the corporate, plan to speculate as much as $100 billion to construct a supercomputer and data center within the United States, in accordance with The information.
The oversized data infrastructure required to run energy-hungry generative AI models is already causing environmental damage. As Microsoft explains in its current sustainability reportthe infrastructure and power required to run the most recent technologies are creating “recent challenges for meeting sustainability commitments across the technology sector.” While the corporate remained optimistic about meeting its long-term goals of becoming carbon negative, water positive and 0 waste, it acknowledged that it was not yet on course to scale back indirect emissions and replenish more water than is utilized in its data centers.
Researchers are working hard to make AI models do more with less energy, which could be very attractive each financially and environmentally. As a part of its green commitment, Microsoft can also be investing heavily in renewable energy. Last month, the company giant committed to bringing 10.5 gigawatts of renewable energy online within the US and Europe in partnership with Brookfield Asset Management. The additional capability, such as powering 1.8 million homes, is estimated to cost $10 billion.
The company is making some wilder bets on small modular nuclear reactors to supply carbon-free energy. The company has also agreed to purchase fusion-generated electricity from Helion Energy until 2028 (assuming it may be produced by then). Helion is a generously funded startup backed by Sam Altman, CEO of OpenAI.
The major AI corporations argue that the energy requirements of the digital economy remain comparatively low on a world scale. According to the International Energy AgencyData centers account for lower than 1.5 percent of world electricity consumption. What is worrying, nonetheless, is how quickly they’re growing. The IEA predicts that data centers could eat 1,000 terawatt hours of electricity by 2026. That can be greater than double the quantity consumed in 2022 and concerning the same as Japan.
Given the apparent environmental downsides of AI, it's all of the more vital to also harness its benefits. One of essentially the most exciting areas is weather forecasting, which may also help us adapt to climate change. In November, researchers at Google DeepMind unveiled GraphCast, an AI model that may produce more accurate (and energy-efficient) 10-day weather forecasts than traditional methods.
As more countries consider geoengineering in response to climate change, similar to cloud seeding and carbon sequestration, AI researchers are also specializing in modeling the potential impacts. “The national security facets of weather are an enormous topic right away,” the CEO of an AI company tells me. “People are assuming geoengineering goes to occur, and so they need to know what the impacts shall be.”
For now, the environmental costs of AI are real while the advantages remain vague. It's time for the industry to deliver on its ambitious guarantees.
