AI’s climate impact is getting attention – what happens next?

Why should companies be more transparent about how they’re using AI? There are plenty of reasons, but a pressing one is that it could be a major source of hidden greenhouse gas (GHG) emissions – and soon it’s going to be everywhere.

Take, for example, the news that Apple is continuing to integrate generative AI functionality into Siri. It’s the latest step in AI’s march into everything – reflecting, perhaps, the fact that about half of the US’ entire VC funding is going into AI and machine learning. That’s a lot of investment to find returns on, so it’s safe to say that this will be the shape of things to come for most things digital.

As we’ve progressed through the AI hype cycle, most of the controversy and discussion has focused on the digital – its impact on intellectual property, its threat to creative jobs, and the tantalising question of whether any actual thinking is going on back there.

But recently, the physical impact of AI is starting to gain notice. According to our research, media conversations about AI’s climate impact are up 30% in terms of reach vs the same time last year, and up 10% year on year overall. AI’s climate impact is becoming more salient.

This is a big deal. The push to incorporate AI into everything is creating a massive demand for computing power, and therefore energy. Datacenters have faced increasing local opposition thanks to their water and power demands – but the issue is becoming global, putting Microsoft’s climate commitments at risk, and the renewable transition overall. Bill Gates has tried to combat these fears by arguing that, because of the efficiencies it will create, AI will actually reduce energy use overall. This might be some way off – and is at odds with Sam Altman’s statement in January that scaling of AI will require so much energy, we’d better hurry up and invent nuclear fusion.

To put this energy use in context – 16 ChatGPT queries uses as much energy as boiling a kettle. Maybe not that much individually, but it adds up; a nation prone to boiling kettles at the same time is part of the reason we’ve relied on hydroelectricity so much in the UK for decades! In Ireland, it’s added up to the point where datacenters now use more electricity nationally than homes. This isn’t all being used for AI – but as AI integration grows, so does the need for datacenters.

At the same time as AI’s environmental impacts are rising, the industry is becoming less transparent about them. In a recent podcast, Hugging Face Climate Lead Sasha Lucconi outlined how new research, in the name of trade secrecy, is far less open about the resources used to train new models. She also describes how, in contrast to other companies, for whom lifecycle assessments are becoming standard, few are available for NVIDIA’s manufacturing – a key part of the AI industry – while other major tech firms have supported lawsuits to prevent journalists releasing data on their water use, claiming that it would be a trade secret.

This goes against the direction of travel for industries around the globe, who are spending ever more time on reporting to comply with climate disclosure policies such as EU’s CSRD (Corporate Sustainability Reporting Directive) and the UK’s TCFD (Task Force on Climate-related Financial Disclosure). The push to include generative AI in everything complicates this it drives up energy use in a huge range of processes, hidden behind multiple layers of technology, few of which are becoming transparent. As Lucconi argues, as AI is integrated into different sectors – to whom do its emissions ‘belong’? If AI is to become an integral part of, say, the transport and logistics vertical, should its emissions from these activities count as part of the IT sector?

A combination of regulation and public awareness means that, far more than it was a few years ago, compliance is a competitive advantage. Companies that can better document their environmental impact will find it easier to work with large clients, and – not always, but more than before – realise a green premium.

Responsible companies have a chance to lead the way when it comes to communicating transparently about their AI and emissions. A company using AI might not be able to change how NVIDIA performs its LCAs, but it can be transparent about where it’s using resource-intensive, generative AI versus extractive AI or other, lighter options. Brands can also be transparent about what they don’t know – and can’t report on – and push for change. They can also look for innovative, climate-friendly computing options, such as Deep Green, a datacenter provider that uses waste heat to heat swimming pools. Brands don’t have to give up using valuable tools – but they can be open about what they are using, why, and what they know about the impacts.

The growing ESG software sector is well-placed to lead this conversation. As reporting requirements multiply, the sector is predicted to grow to become a $2.1bn industry by 2029. As it grows, it’s already rapidly adopting AI in its own software to help streamline reporting processes. Doing so responsibly, and communicating transparently, must be a part of this process if the sector is to retain its status as a trusted, purpose-driven part of the green transition.

The alternative, of course, would be to activate the ‘somebody else’s problem field’, avoid looking at this exponentially growing source of emissions, and hope it goes away. I’d argue that we’ve seen where this road leads with the proliferation of substandard offsets, facilitated by a similar lack of transparency.

The increased focus on AI and sustainability won’t be around forever, but for now there is a chance to change the narrative before another opaque climate problem is normalised. It starts with smarter, bolder, and transparent communication.

If you’d like to learn more about how your cleantech brand can start leading essential conversations, contact Peter at [email protected].