Google has released its Environmental Report 2024, a document of more than 80 pages that describes all the efforts of large companies to apply technology to environmental problems and reduce their own contribution. But he completely avoided the question of how much energy the AI ​​uses – probably because the answer was “more than I can tell you.”
You can read the full report here (PDF), and to be honest, there’s a lot of interesting stuff. It’s easy to forget how many plates a large company like Google keeps spinning, and there’s some really impressive work here.
For example, it has been working on a water replenishment program, which hopes to balance the water used in facilities and operations, ultimately creating a net positive. This is done by identifying and funding watershed restoration, irrigation management and other work in the region, with dozens of these projects around the world at least partially funded by Google. Already 18% of water use is recharged (by any definition of the word used here) and it increases every year.
The company is also careful to reveal the potential benefits of AI in the climate, such as optimizing irrigation systems, creating more fuel-efficient routes for cars and boats, and predicting floods. We’ve highlighted a few that are already in AI coverage, and can be useful in many areas. Google doesn’t have to do this, and many big companies don’t. So credit is due.
But then we get to the section “Responsibility for managing the consumption of AI resources.” Here Google, so that every statistic and estimate so far, suddenly spreads its hands and shrugs. How much energy does AI use? It can be anyone indeed sure?
But it must be bad because the first thing the company did was reduce the entire data center energy market, saying that it is only 1.3% of global energy use, and the amount of energy Google uses is only 10% at most – so it is only 0.1. % of all energy in the world powers servers, according to the report. A trifle!
In particular, in 2021, they decided they wanted to achieve net-zero emissions by 2030, although the company admitted there were many “uncertainties,” as they called it, how that would happen. Mainly because emissions are increasing every year from 2020.
In 2023, total GHG (greenhouse gas) emissions will be 14.3 million tCO2e, represent A 13% year-on-year increase and a 48% increase compared to the 2019 target base year. This result is mainly due to the increase in data center energy consumption and supply chain emissions. As we further integrate AI into our products, reducing emissions may become a challenge due to increased energy demand from greater AI computing intensity, and emissions associated with increased investments in our technical infrastructure.
(Emphasis mine on this and the quote below.)
But the growth of AI is lost among these uncertainties. Google has the following reasons why the company is not specific about the contribution of AI workloads to the general data center energy bill:
Predicting the future environmental impact of AI is complex and evolving, and our historical trends cannot capture AI’s future trajectory. As we integrate AI deeply into our product portfolio, the difference between AI and other workloads will not be useful. So we focus on data center metrics because it includes the overall resource consumption (and therefore, environmental impact) of AI.
“Complex and evolving”; “the trend cannot be fully implemented”; “difference … won’t help”: This is the kind of language used when someone knows something, but would rather not tell you.
Can anyone believe that Google doesn’t know, down to the penny, how much AI training and inference adds to energy costs? Is it not possible to break these numbers into part of the company’s core competence in cloud computing and data center management? It has all these other claims about the efficiency of dedicated AI server units, how to do all this work to reduce the energy required to train an AI model by 100x, and so on.
I have no doubt there are many green efforts in Google, and you can read all about it in the report. But it is important to highlight what is not denied: the enormous and increasing energy costs of AI systems. The company may not be the main driver of global warming, but despite its potential, Google doesn’t seem to be positive yet.
Google has every incentive to reduce and confuse these figures, which even in reduced conditions, very efficient, cannot be good. We’ll definitely ask Google to be more specific before we know if things will get worse in the 2025 report.
