We use our little friend everyday, whether that be asking a question, researching or talking about our relationship problems to it, and some can confidently say that it’s become their ‘go to’ hub for anything and everything, but do we know the impact that this is having on our global footprint.

I’d like to be clear that by no means am I a technophobe. At Carbon Neutral Group, we use data, analytics, and yes, even AI tools in our work. But as someone whose entire professional mission is built around honest accounting and helping businesses with a realistic approach to sustainability and  environmental impact, I have an obligation to say what the industry is largely whispering: AI has an enormous and largely unacknowledged carbon problem.

The hidden cost of chat

Every time you type a question into AI, whether you’re asking it to write an email, summarise a report, or check something, we are  triggering a huge amount of computation happening. This is all in real time, across powerful GPU groups in data hubs around the globe. I’m also not at all kidding too, when I say they are nothing like your laptop processors quietly working behind the scenes, these are large scale machines drawing a serious amount of power around the clock.

Each question to AI uses up to ten times the electricity of a standard Google search, yes ten times. Now you can imagine if hundreds of millions of queries are being fired at AI every single day, you would be multiplying that figure by the hundreds of millions. It’s safe to say that we are under no illusion that it stops here.  As you can see AI is dramatically developing and fast, so this will increasingly  grow in the years ahead.

The water that’s spilling

The server farms that power modern AI don’t just consume electricity, they generate an enormous amount of heat, and that heat evidently has to go somewhere. The answer in most cases is water. Evaporative cooling systems quietly drain local water supplies at scale, and the quantities involved are quite extraordinary.

Generating just 100 words of AI output can consume the equivalent of nearly a full bottle of water. Let’s just say something as simple as “Hi” can cost around 1.5 ounces of water. These figures sound trivial in isolation, but really shocking when multiplied by millions of interactions  by us, per hour, each day, in data centres that are often located in regions already under serious water stress.

I find it remarkable that we’ve built global frameworks to track Scope 1, 2, and 3 carbon emissions, yet the water consumed by digital infrastructure barely registers in most corporate ESG disclosures and I feel this is something we need to look into.

Accountability where are you

Here is what frustrates me the most as a  sustainability practitioner: we are finally living through a period of extraordinary corporate ambition around net zero targets. Thousands of organisations have made public commitments, with frameworks that have been built to hold them accountable, and yet, the AI tools those same organisations are now enthusiastically adopting sit largely outside those accounting frameworks. The emissions AI is creating are quite real. The water consumption is also real, but because they’re being absorbed into muddied supply chains, hidden in Scope 3 emissions of technology vendors, I ask myself are they even being accounted for.

AI and I are not having a moment

I am not arguing that we should abandon AI. This platform and the technology behind it has genuine potential to accelerate sustainability solutions and some of the most exciting work in environmental science is being powered by machine learning. But let’s get one thing straight, potential does not excuse a lack of accountability. The automotive industry hasn’t been given a pass on emissions because cars are useful to us everyday . The aviation industry isn’t exempted from carbon reporting because flying gets us around the globe. We measure,we set targets, and we drive innovation toward cleaner solutions. I feel like AI deserves exactly the same treatment with the same conscientiousness.

What sustainability consultancies need to do

The sustainability community needs to treat the AI energy and water consumption it uses as an issue. Technology businesses must be required to disclose the energy and water intensity of their AI products with the same granularity we expect from manufacturers reporting on product emissions. “AI-powered” can no longer be an advertising claim without an accompanying environmental footnote I think.

Also businesses who use AI whether that be within a team software or individual usage by team members for work purposes should be accounted for and must be included in Scope 3 accounting. If you’re outsourcing computation to a cloud AI provider lets say, that consumption belongs in your organisational footprint. There is no logical argument to why it shouldn’t be accounted for.

Moment of honesty

We have built our reputation I should hope anyway, on  honest accounting. We measure things that are inconvenient at times, we report on numbers that might be uncomfortable and we hold businesses to the same standards we apply to ourselves.

AI isn’t off on its travels  around South East Asia, it’s sticking around. This means the energy and water costs of AI are real, it’s growing, and it’s currently flying under the radar of the framework we’ve worked so hard to build as experts in the sector.

Let’s bring it to light.

(Image: Pexels.com_Angel Rkaoz)

The post The vampire in AI: Why AI’s energy sucking should have every sustainability leader worried at night first appeared on ESG Foundation.

The post The vampire in AI: Why AI’s energy sucking should have every sustainability leader worried at night appeared first on ESG Foundation.