Google’s heavy reliance on AI for its services has contributed to its carbon footprint. In the words of a Times headline, “Google blames AI as its green carbon targets go up in smoke”. Global warming is the biggest single threat to wildlife on the planet and the single biggest negative impact on wildlife through human behaviour.
We are told that Google’s artificial intelligence bot named Gemini has increased Google’s carbon emissions by 13% last year because the company has switched to AI to power its services.
AI demands more energy which has resulted in Google omitting about 14 million tonnes of carbon dioxide equivalent, about a third of the emissions from the UK’s entire electricity grid last year. A remarkable statistic.
At the base of Google’s homepage it currently states: “Our third decade of climate action: join us”. I think they had a claim in the past that they were carbon neutral for many years. It seems that AI has damaged that very important image that Google projected to the world.
Google had the aim of halving emissions but the company is up 40% on its baseline year of 2019.
The growth in electricity use comes from its data centres and this is where the problem is in terms of trying to be carbon neutral.
Google introduced its AI part Gemini in 2023. In the past, an environmentalist praised Google for its target of halving emissions by the end of the decade.
There are plans to invest in new technologies such as machinery to suck carbon dioxide from the air.
The company has now admitted that the targets may be challenging because of the “increasing energy demands from the greater intensity of AI”.
The language models used by Gemini and Microsoft Copilot are very energy intensive.
Eight years ago Google painted a glowing picture of AI praising the technology for reducing by 15% the energy use of data centres and by applying an algorithm to optimise cooling.
Microsoft is also suffering the same problem. They said in May that its greenhouse gas emissions were up 29.1% on a 2020 baseline.
It produces 15 million tonnes of carbon dioxide equivalent. The company blames the “construction of more data centres” rather than AI per se.
Microsoft’s climate targets are even more ambitious than Google’s. They want to reduce emissions to 0 and remove carbon dioxide and other greenhouse gases from the atmosphere.
Globally, data centres account for up to 1.5% of electricity. By 2026 demand could double according to the International Energy Agency.
“Fuelling the AI boom with renewable energy is already proving to be a challenge. But even if successful the planet is not necessarily better. Using limited renewable energy to feed a rapidly growing number of AI servers will lead to increased fossil fuel consumption elsewhere.”
Alex de Vries founder of the Digiconomist website
It is hoped that AI will help to unlock new ways to tackle climate change. One report suggests that using AI across supply chains and in energy grids might lower global emissions by 4% in 2030. But the benefits remain hypothetical.
Mark Zuckerberg the head honcho at Meta said that to train an AI model (1 GW datacentre) requires the amount of energy produced by a nuclear power plant.
I asked Google’s Gemini two questions:
Is AI very energy hungry?
Response:
AI can be quite energy hungry, yes. There are two main reasons for this:
- Training: Training complex AI models requires massive amounts of computing power, which translates to a lot of electricity use in data centers.
- Hardware: The specialized hardware used for AI, like GPUs (graphics processing units), is generally more power-hungry than traditional CPUs (central processing units).
Here’s a breakdown of the situation:
- The Concern: The computational demands of AI are growing rapidly, and some fear this could lead to a significant increase in global energy consumption.
- The Hope: Researchers are developing more efficient AI algorithms and hardware, which could help to reduce the energy footprint of AI in the future. There’s also optimism that AI itself can be used to optimize energy use in other sectors.
Overall, it’s a complex issue. AI has the potential to be both a drain and a benefit for our energy consumption.
Has Google’s reliance on AI damaged its green credentials?
Response:
It’s true that Google’s heavy reliance on AI for its services can contribute to its carbon footprint. Here’s a closer look:
- The Impact: Training and running AI models requires significant energy use in data centers. This can contradict Google’s goal of achieving net-zero emissions.
- Google’s Efforts: However, Google is aware of this challenge. They’ve made strides towards sustainability:
- Matching their global electricity consumption with renewable energy purchases since 2017.
- Setting a goal to achieve 24/7 carbon-free energy by 2030 (meaning using clean energy sources every hour of every day).
- Mixed Results: A recent report highlights a nearly 50% increase in Google’s carbon emissions over the past five years [9to5Google article on Google’s carbon emissions]. This suggests AI growth might be outpacing their sustainability efforts.
Overall: It’s a work in progress. Google’s reliance on AI can be at odds with its green goals, but they are actively working on solutions.
Here are some resources for further reading:
- Google Sustainability: https://sustainability.google/
- Our 2024 Environmental Report: https://blog.google/outreach-initiatives/sustainability/2024-environmental-report/
