Why the climate promises of AI sound a lot like carbon offsets

Even without any “penetration discoveries”, the estimates that IEA suffers from, the adoption of artificial intelligence applications on a large scale can reduce emissions by 1.4 billion tons in 2035. These cuts, “if achieved”, will be like three times emissions from data centers by that time, under the most optimistic development scenario of IEA.

But this is “if”. It takes a lot of belief in technical progress, widespread publishing, and bonuses of changes in practices over the next ten years. There is a big gap between how artificial intelligence Be able to Use and how will It is used, a difference that depends a lot on economic and regulatory incentives.

Under the Trump administration, there is no good reason to believe that American companies, at least, will face a lot of government pressure to use these tools specifically to reduce emissions. In the absence of the necessary carrots or political sticks, it can be said that the oil and gas industry will publish Amnesty International to discover new fossil fuels instead of determining methane leakage.

To be clear, IEA numbers are a scenario, not prediction. The authors have easily acknowledged that there is a great lack of certainty in this case, saying: “It is necessary to note that there is currently no momentum that can ensure the adoption of AI applications on a large scale.

In other words, we can definitely rely on artificial intelligence to reduce emissions more than it drives, especially in the time frame that climate change is now required.

As a reminder, it is already in 2025. The growing emissions now pushed the planet close to exceeding 1.5 degrees Celsius completely of warming, risks of heat waves, dehydration, sea level rise and climbing of forest fires-and global climate pollution is still rising.

We are dealing with the mid -century, only 25 years shy when climate models show that every industry in each country needs to approach the clear network emissions to prevent a temperature of 2 ° C from pre -industrial levels. However, any new plants of natural gas that have been built today can remain for data centers or any other purpose, operating easily after 40 years from now.

Carbon dioxide remains in the air for hundreds of years. So, even if the artificial intelligence industry ultimately provides ways to reduce emissions than produced in a specific year, these future cuts will not eliminate emissions that the sector will pump along the way – or the warming they produce.

It is a comparison that we do not need to make if artificial intelligence companies, facilities and regional organizers make more wise options on how to operate the data centers they build and operate today.

Some technology and energy companies take steps in this direction, by stimulating the development of solar farms near their facilities, or helping to restore nuclear plants online, or signing contracts to obtain new thermal energy factories.

But such efforts should become the rule more than the exception. We no longer have a carbon budget to keep emissions on the promise that we will take care of it later.