Three ways financial institutions can mitigate the hidden carbon cost of artificial intelligence

Once considered a novelty, the role of Chief Sustainability Officer (CSO) is now commonplace in financial institutions, with the UK’s top three banks – HSBC, Barclays and Lloyds – each employing a head of sustainability. 

Today, financial services companies aren’t just looking to improve their ESG performance – namely, investments made based on environmental, social and corporate governance criteria. As concerns about climate change continue to take centre stage, the finance sector is also working to embed sustainability down to the core of their business. 

But many companies aren’t considering the significant environmental cost of the technologies they are employing. Instead, it’s often employees themselves who are concerned with the sustainability of AI deployments, and are beginning to demand greener AI – and voting with their feet if firms aren’t working towards it.

The carbon cost of AI

AI is powering ever more technologies, and the finance sector is one of its greatest adopters. Some 85 percent of financial institutions report they are currently using some form of AI (University of Cambridge and World Economic Forum Survey). Financial services firms use AI for a broad range of purposes, from risk management and fraud detection to asset portfolio optimisation and customer communication. 

AI is particularly useful for quantitative trading, which uses large data sets to identify patterns that yield insights and generates predictions that inform strategic trades. AI’s machine learning models analyse vast and complex data, and are not only data-hungry, they are power-hungry too.

The supercomputers that train and test mountains of data for AI models can run for hours, days, or even weeks, and these applications consume huge amounts of energy. The carbon emissions from training a single AI model for language translation is roughly equivalent to 125 round-trip flights from New York to Beijing (AI Now 2019 Report).

The carbon cost of AI is even higher when you consider the energy required to keep the computing equipment housed in data centres cool. Cooling is crucial because overheating can impact performance and damage equipment. In fact, in a traditional data centre, at least 40% of all energy consumed goes towards cooling.

But the use of AI can still be compatible with sustainability goals if financial services organisations take the right steps to minimise AI’s environmental impact.

So how companies can minimise AI’s carbon footprint?

1. Commit to sustainable AI – and a plan of action

The first step for financial sector firms looking to minimise the carbon footprint of their AI is to make a firm commitment to sustainability, backed by concrete plans.

Like tech giants Amazon and Google – both of which have made pledges to reach certain renewable energy goals – many financial firms have made specific sustainability commitments, such HSBC providing $100 billion in financing and investments by 2025 to develop clean energy and lower-carbon technologies, and investment firm Blackstone’s pledge to lower the electricity consumption of companies in their real estate portfolio.

In addition to benefiting the environment, sustainable business practices can save on costs and resources, improve reputation, and even strengthen employee recruitment and retention. By setting a target and making a plan, the finance sector can ensure it’s on a path to sustainable AI.

2. Choose cloud providers carefully

Cloud providers may boast that they are carbon-neutral, meaning their carbon footprint is a ‘net zero.’ But it’s important to understand whether a cloud provider actually produces few carbon emissions, or whether that cloud provider is carbon neutral through carbon offset, where carbon emissions are counteracted by funding an equivalent carbon dioxide savings somewhere else.

Despite the “green” label, there’s no guarantee that a cloud provider is powered entirely, or even partially, by green energy. And while carbon offset programmes are a step in the right direction, given the high carbon cost of AI, it would be even better to find a cloud service provider that doesn’t generate large amounts of carbon emissions in the first place.

3. Location, location, location 

Companies must also evaluate whether the data centres that power their AI applications are sustainably powered. The servers that train AI models can easily be housed in data centres powered by renewable energy sources, such as in the Nordics, where data centres are largely powered by sustainable energy. 

Renewable energy is much less harmful to the environment because, unlike fossil fuels, it’s non-polluting and does not generate greenhouse gases. Iceland, for example, produces 100% renewable hydroelectric and geothermal power – with no nuclear power sources – and is connected to a reliable power grid.

Further, when data centres are located in cooler, more temperate climates, natural air cooling of powerful AI servers minimises energy usage significantly. Over 80% of computers don’t need to be near the end-user, and in those situations, choosing data centre locations in cool climates has a substantial impact on carbon emissions, due to year-round, free cooling.

By making a real commitment to sustainable AI, choosing cloud providers carefully, and selecting green data centres, companies can substantially reduce energy consumption and expenses, as well as benefit from long-term energy cost predictability. 

When it comes to achieving overall sustainability goals, financial services companies can’t afford to ignore the environmental impact of AI, especially since what’s good for the planet is also good for the bottom line. 

Nick Dale is EVP of business at Verne Global.

The views and opinions expressed in this Viewpoint article are solely those of the author(s) and do not reflect the views and opinions of Fintech Bulletin.