Is the generative AI boom a potential bust for the environment?

For many, 2023 has been about the ‘AI Boom’. Excitement around ChatGPT and other generative AI technologies has meant that businesses in this space have been subject to significantly increased investment levels and shareholder interests. Rapid advancements in computing power has driven the ability of AI companies to push the envelope, and generative AI in particular is now being rolled out to the masses, and this will be a huge boon for businesses globally. But, this increased demand for generative AI could be a massive blow to the environment.

Whilst ESG has, at times, been derided and dismissed in the city as a buzzword – events this year have served as a severe reminder that environment matters are being linked to financial performance. There is no better example than TUI’s market capitalisation being materially impacted by heatwaves and fires across Europe during the summer.

It’s increasingly hard for businesses to ignore the impact of climate change. 93% of middle market business leaders have said they have plans to incorporate generative AI into their business. As boards tackle the cost of implementing generative AI solutions into their business, they may want to consider the potential environmental cost resulting from increasing use of AI.

Data centres – the factories of the digital economy

One of the beneficiaries of the AI boom is semiconductor chip maker Nvidia - who became a $1tn company in 2023. Excitement around Nvidia is centred on their A100 and H100 chip, which are installed in rows at data centres across the globe. These data centres act as digital factories - powering the AI compute demand. Nvidia and market analysts are forecasting significant growth against a backdrop of unprecedented demand for more and more data centres.

'The computer industry is going through two simultaneous transitions — accelerated computing and generative AI,' said Nvidia CEO, Jensen Huang. 'A trillion dollars of installed global data centre infrastructure will transition from general-purpose to accelerated computing as companies race to apply generative AI into every product, service and business process.'

An increase in demand for AI compute triggers increasing demand for more and more data centres that in turn need the latest super-powered chips. Nvidia and competitors such as Intel and AMD are not alone in their excitement as the chip industry gears up to capitalise on an estimated trillion dollar of assets deployed and expectations of a further $250bn of assets being built a year.

The impact on the environment

Our increasing appetite for computing power has raised some significant environmental concerns. Data centres can be significant capital projects that require a physical footprint. They need to be powered with electricity that generates enormous amounts of heat, which in turn requires cooling.

These concerns are not new. Graphics Processing Units (GPUs) have been central to cryptocurrency for years where they are used to crank out the complex algorithms that underpin the blockchain. Rocky Mountain Institute, a non-profit organisation of experts looking to accelerate clean energy, note that Bitcoin alone is estimated to consume 127 terawatt-hours a year, more than Norway. US crypto activity is estimated to emit 25 to 50 million tonnes of CO2 a year – on par with annual emissions from diesel fuel on US railroads.

Whilst debate still ensues on the ways blockchain technologies can be deployed, artificial intelligence is increasingly expected to be a once in a lifetime game-changing general-purpose technology. The environmental concerns are compounded against a backdrop of massive expansion in the computing capacity. Training a single AI model can emit more than 626,000 pounds of carbon dioxide equivalent, according to MIT - nearly five times the lifetime emissions of the average car (and that includes manufacture of the car itself). McKinsey estimate that the power consumption from data centres will be 35 gigawatts by 2030, rising from 17 GW in 2022. They note that one hyperscale data centre can use equivalent power to 80,000 households.

Some businesses are looking at solutions to address this. Data centre company Crusoe is looking to develop data centre solutions that tap into stranded energy, such as flared methane or excess production from renewable sources. By doing this they hope to align the future of computation with a sustainable future for the planet. Regulators and governments recognise the challenge too as they look to deploy sustainability standards on newly built data centres, with the Netherlands and Singapore being early adopters. The 64% of business leaders who believe AI should be governed by a regulatory framework are likely to perceive nations who are monitoring the impact on the environment as taking positive step forward.

The impact on middle market business leaders

For the average mid-market business, the environmental impact of greater use of generative AI will be a downstream consideration. Whilst some businesses will work in this space or develop their own proprietary digital technologies, the vast majority are likely to use software licensed from third parties, configured for use in their business. Considerations about the environmental impact of how these tools have been developed or what is powering them today may feel removed from the boardroom agenda.

However, the changes in the environment are becoming increasingly hard to ignore. Environmental matters can have a direct impact on the share value of a business – as TUI found this summer. We have talked in our report how generative AI is expected to become a key pillar for middle-market businesses – but businesses are also being expected to extend environmental reporting and monitoring to cover the resultant impact on the environment.

Finally - from an operational perspective, as businesses become more reliant on AI, boards may want to monitor the location and stability of the data centres that ultimately underpin their businesses, just as they would the rest of their supply chain. Many business leaders will be actively monitoring the location of their data storage to comply with regulations such as GDPR. Geopolitical and reputational concerns may be other areas to consider alongside environment as the compute becomes part of the organisations digital supply chain