Artificial intelligence (AI) has become so ubiquitous, many of us don’t realise we are benefiting from its use every day. In the financial sector, it has helped streamline numerous processes, including fraud prevention measures, customer services, risk management, day-to-day administration and more. But how does the use of AI impact a firm’s green credentials and is there enough clarity from cloud service providers about the environmental impact of their data centres? In our latest free report, we consider how firms might overcome the environmental risks involved with AI and the impact of other energy-intense technologies such as cryptocurrency and Blockchain.
How bad is AI for the environment?
When businesses and individuals began (consciously) using services powered by AI, it is probably safe to say that for most, their first concern wasn’t the potential environmental impact of the technology. How can a software program or algorithm – something intangible that is inside your computer, phone, or in the Cloud – have a physical impact on the environment? “When we talk about ‘The Cloud’, we think of it as something that requires no physical presence – because for the businesses using Cloud services, that’s broadly what it means: no on-site servers,” says Mike Finlay, CEO of RiskBusiness. “But in reality, storing large volumes of data – whether on site or elsewhere – always requires hardware. And hardware needs energy to operate. Herein lies the environmental impact.”
How much energy does AI use?
The complex Large Language Models (LLM) used for AI systems require huge volumes of data in order to function. This data all has to be harnessed and stored in a physical server in a data centre. Data centres rely on massive amounts of electricity as they are running around the clock. Those data centres also generate a lot of heat, which then has to be reduced using air cooling systems. These also use a lot of electricity and require high volumes of water to function.
According to estimates, using a generative AI platform such as ChatGPT for an online search query uses 10 times as much energy as carrying out a Google search. The models used by these AI systems also have to be trained – a process which is extremely energy intensive. According to the Association of Data Scientists (ADaSci), training OpenAI’s ChatGPT-3 required approximately 1,287 megawatt hours of electricity. This is the same amount of energy used by an average household in the US – not per year, but per 120 years. “The type of hardware used for training and running LLMs significantly impacts energy consumption” says ADaSci. “High-performance GPUs (Graphics Processing Units) and TPUs (Tensor Processing Units) are commonly used for these tasks due to their ability to handle large-scale, parallel computations. For instance, NVIDIA’s A100 GPUs, used in many modern AI training setups, have a maximum power consumption of around 400 watts each. Training a large model across 1,000 A100 GPUs could consume up to 400 kilowatts per hour.” This demonstrates the volume of energy required and also how hardware-heavy data centres need to be.
Draining natural resources
The hardware described above relies on materials that are heavily mined, including rare earth elements, metals and other critical minerals. These have significant environmental and social impacts associated with their extraction. Some of the impacts of mining these materials include: land degradation and habitat loss; high volumes of water consumption and pollution; greenhouse gas emissions; deforestation and soil erosion; human health impacts such as respiratory issues from dust particles and toxic chemicals released during the mining process and displacement of communities living in areas being mined. Once the minerals are mined, they then have to be processed, which is also energy intensive.
Are banks building this into their sustainability initiatives?
Many banks claim to be carbon neutral – or working towards this status. But are they taking the energy required for AI usage into consideration?
This feeds into supply-chain/third-party/Nth party risk because calculating how much energy is being used by suppliers is a complex process. Environmental activists, investors and consumers in general, are becoming more astute at assessing a bank’s links to fossil fuels. Currently, in most countries, companies like Amazon and other data centre owners are not universally required to make their energy consumption public. However, regulatory pressure is increasing, especially as data centres become significant consumers of electricity due to the global rise in digital services. For now, most energy consumption reporting by data centre companies is voluntary and often generalised at the corporate level rather than by individual facilities. However, with increasing regulatory interest, particularly in the EU, and mounting public and investor pressure, data centre energy consumption disclosure may become more standardised and public in future.
Understanding carbon neutrality
Banks have an important role to play in creating a world that is carbon neutral. They have the power to decide which companies receive capital and should be working towards phasing out financing for fossil fuels and deforestation.
Being “carbon neutral” means to achieve an equal balance between the volume of carbon dioxide released into the atmosphere from a firm’s activities and the amount removed. Companies can achieve this by using renewable energy sources to power their operations, but also through investing in green initiatives such as carbon offset programmes. Carbon offset programmes include things like reforestation projects and renewable energy projects such as solar transportation.
Most carbon-neutral claims follow the Greenhouse Gas Protocol, which separates emissions into three scopes:
Scope 1: Direct emissions from owned or controlled sources.
Scope 2: Indirect emissions from the generation of purchased energy (e.g., electricity).
Scope 3: All other indirect emissions, such as from supply chains and digital services, which can include AI.
It is becoming increasingly difficult for even the most powerful of companies to achieve carbon neutrality. Google claimed to be carbon neutral from 2007 until 2023 by purchasing carbon offsets to match the carbon emissions from its data centres, office buildings and travel. But in the tech giant’s 2023 environmental report, it confirmed it had ceased to be carbon neutral due to a 50% increase in its greenhouse emissions since 2019. The increase was largely driven by the growing energy demands of AI development and infrastructure. The firm has said it aims to achieve net-zero emissions across all of its operations and value chain by 2030.
“The path to get to these goals is difficult, and we’re committed to working through the challenges we face with the ultimate aim of driving larger systems change to create a more sustainable future,” said Google’s Senior Vice President, Benedict Gomes. “Further, predicting the future growth of energy use and emissions from AI compute [sic] in our data centres is difficult. Despite this, we remain focused on developing new ways to make AI computing more efficient while leveraging the opportunities that AI presents to have a positive environmental impact.”
The counter argument: using AI for green initiatives
AI is such a powerful tool so couldn’t its benefits be funnelled into making banking greener? The answer is yes. On a data-centre level, AI can optimise data centre operations to prevent energy waste in ways that manual or traditional methods cannot. By using AI, data centres can dynamically adjust their cooling systems, server workloads and even resource distribution in response to demand fluctuations, all of which reduces their overall energy use.
Another argument is that cloud-based computing – which AI relies heavily upon – involves multiple companies using the same site to house their data, helping to save on energy usage. “Cloud computing’s multi-tenant approach puts more applications, data sets and users on a smaller amount of hardware, at approximately 85% to 95% utilisation of capacity,” explains David Linthicum in an article for InfoWorld. “Compare this to traditional approaches where we own the servers and data centres, and the hardware resources are often utilised at a very low capacity, often 3% to 7%.”
Examples of banks using AI for green initiatives
AI is also being leveraged by banks to optimise their green initiatives, which arguably contributes towards offsetting the carbon emissions created by AI processes.
HSBC: HSBC launched an AI-powered index in May 2023 which tracks companies that are expected to benefit financially from improvements in ESG-related risk management. By analysing ESG metrics in this way, HSBC has been able to shift investments towards greener, more sustainable companies and provide greener investment options to its clients. This AI-driven shift could encourage more businesses to adopt sustainable practices and help investors make eco-conscious choices.
JPMorgan: JPMorgan uses AI to optimise its energy consumption and renewable energy sourcing for its office buildings and data centres. “Software will be embedded in JPMorgan Chase’s building management systems at its larger commercial real estate properties across the globe to help each building run more efficiently by using artificial intelligence and sensor technology to track and improve the firm’s energy use,” said Nantum AI, who provided the software used by JPMorgan. The AI-driven platform claims to help reduce unnecessary energy usage and optimise the integration of renewable energy sources. This reduces JPMorgan’s carbon footprint and could encourage a trend within the wider banking industry to adopt cleaner energy practices.
BBVA: BBVA is among a handful of banks that has specifically addressed the environmental problems of AI usage. “Large language models are indeed very power-hungry, and this should be clearly a concern with regards to their environmental impact,” Javier Rodriguez Soler, global head of sustainability and corporate and investment banking at BBVA, told American Banker. In an interview with Global Finance in October 2024, he addressed the issue again when he said that one of the biggest changes occurring in the banking sector right now is its use of AI: “One of the main changes will come from data and artificial intelligence, and its environmental impact should be a concern,” he said. “Large language models are indeed very power-hungry. That is why we are looking into a new generation of models that achieve comparable performance with fewer parameters, significantly reducing energy consumption. One of our primary strategies at BBVA is adhering to the principle of data minimisation when building AI models. By using the data that is absolutely necessary, we create smaller, more efficient models that require less computational power – and consequently, less energy.”
BBVA employs AI to assess climate risks in its loan portfolio, especially in industries sensitive to climate change. The bank developed a tool that uses AI to identify which loans and investments could be at risk due to climate-related factors, such as extreme weather events or regulatory changes. By adjusting its lending policies based on these findings, BBVA is able to reduce financing to high-risk, high-emission industries. This helps redirect capital away from industries that may contribute heavily to pollution and climate change, steering it instead towards greener initiatives.
Data centres and their impact
According to the International Energy Agency (IEA), data centres account for about 1-1.5% of global electricity use. Despite rapidly growing demand for AI and digital services, the IEA says emissions from data have grown only modestly since 2010, which it says is “thanks to energy efficiency improvements, renewable energy purchases by information ICT companies and broader decarbonisation of electricity grids in many regions.” However, the IEA warns that emissions must drop by 50% before 2030 in order to get on track with the Net Zero Scenario, which envisions global net zero emissions by 2050. Net zero refers to “cutting carbon emissions to a small amount of residual emissions that could be absorbed and durably stored by nature and other carbon dioxide removal measures, leaving zero in the atmosphere,” according to the United Nations’ definition.
“Not in my back yard”
Many large tech companies have data centres based in Ireland, in some cases for tax and data regulation reasons. The resultant strain on the electrical grid in the Dublin area in particular, has resulted in a near-freeze on new data centre approvals. According to official figures, data centres in Ireland consumed more electricity in 2023 than all of its urban homes combined.
Similar restrictions on building new data centres have been put in place in the Netherlands, Germany and the US. Virginia, which is home to one of the largest data centre markets in the world known as the “Data Centre Alley”, has faced growing scrutiny over energy consumption and land use. In Texas, local government officials are pushing for regulation to force data centres to build their own power plants to power their activities in the state. While no outright ban on data centres is in place in the US, environmental groups and local governments have been pushing for more sustainable practices, including renewable energy usage and limits on water consumption.
DORA regulation: data centre recovery
DORA, the Digital Operation Resilience Act, is a piece of EU legislation aimed at boosting the operational resilience of the financial services sector. One of the key provisions of DORA is a requirement for firms who outsource their data storage to the Cloud to have in-house backup or recovery capabilities. Specifically, firms must be able to recommence data processing in-house in the event of a disruption to their outsourced data centre services. This means that even if a firm relies on third-party providers for their data storage and processing, they are still required to have the ability to switch operations to their own data centres (or a backup facility) if the outsourced service is disrupted.
This may lead to additional energy consumption by many firms as they may need to have functional data centres in-house running in standby mode for contingency purposes.
AI washing
It is also important to consider the other elements of ESG risk associated with AI use – i.e., not only the environmental risks, but also the social and governance issues. The so-called “AI race for arms” has seen companies of all industries scrambling to take advantage of the potential benefits of AI – or at least in some cases, to appear to do so. “Firms need to be wary of over-promising and under-delivering when it comes to AI,” warns RiskBusiness’ Mike Finlay. “The potentials of these technologies may appear to be limitless – but only when utilised in the correct way and targeted in the right areas. Banks should be considering what investors and consumers are actually gaining from the use of AI and whether or not they can demonstrate these benefits to regulators as their focus shifts to this emerging area of oversight.”
In March 2024, the US Securities and Exchange Commission (SEC) took action against two investment advisers, Delphia USA and Global Predictions, for making false and misleading statements about their purported use of artificial intelligence. “We’ve seen time and again that when new technologies come along, they can create buzz from investors as well as false claims by those purporting to use those new technologies,” said SEC Chair, Gary Gensler. “Investment advisers should not mislead the public by saying they are using an AI model when they are not. Such AI washing hurts investors.”
The Trump effect
Donald Trump’s reelection as president of the United States will undoubtedly have an impact on how climate-related risks are regulated in the financial industry. Trump, who has for many years denied the existence of climate change, is unlikely to prioritise regulation around green banking initiatives. However, his close links with the world’s richest man and Tesla CEO, Elon Musk, may muddy the waters somewhat. Musk’s business interests lie within the renewable energy markets, particularly solar energy and electric vehicles (EVs). Trump’s administration is likely to continue his championing of fossil fuel industries and reduced subsidies for renewable energy. However, Musk’s influence has already been demonstrated by Trump’s about-face on the EV industry in recent years. Having previously described EVs as creating a potential “blood bath” for the automotive industry, conversely, on the campaign trail in August this year he said: “I’m for electric cars; I have to be because Elon endorsed me very strongly.”
Trump has suggested Musk may play an unofficial role in his new administration, though the specifics are yet to be confirmed. Their collaboration could lead to a notable shift in how climate-related policies impact financial institutions, particularly around banks’ obligations concerning fossil fuel investments and energy consumption.
Cryptocurrencies
Cryptocurrency is another energy-guzzling technological advancement in the financial services sector. According to the International Monetary Fund (IMF), due to the electricity used by the hardware needed to mine crypto assets, one Bitcoin transaction uses roughly the same amount of electricity as the average person in Ghana or Pakistan consumes in three years.
The UN has also raised its concerns. According to a United Nations University study, in the year from 2020 to 2021, the global Bitcoin mining network consumed 173.42 terawatt hours of electricity. “This means that if Bitcoin were a country, its energy consumption would have ranked 27th in the world, ahead of a country like Pakistan, with a population of over 230 million people. The resulting carbon footprint was equivalent to that of burning 84 billion pounds of coal or operating 190 natural gas-fired power plants.” it said.
Water consumption is also a huge problem with cryptocurrencies. During the same time period, the UN estimates Bitcoin used the equivalent of more than 660,000 Olympic-sized swimming pools’ worth of water – enough to meet the domestic water needs of more than 300 million people in rural sub-Saharan Africa.
Norway, Sweden, Thailand and the UK are among the countries that make it to the top 10 list when the water or land footprint of their Bitcoin mining activities is taken into account. Those top 10 Bitcoin mining countries (for environmental footprint) are responsible for 92–94% of Bitcoin’s global carbon, water and land footprints.
“Technological innovations are often associated with unintended consequences and Bitcoin is no exception,” said Professor Kaveh Madani, the Director of the United Nations University Institute for Water, Environment and Health (UNU-INWEH), who led the study. “Our findings should not discourage the use of digital currencies. Instead, they should encourage us to invest in regulatory interventions and technological advancements that improve the efficiency of the global financial system without harming the environment.”
US regulators on crypto
Banks’ exposure to cryptocurrencies is relatively modest, but considering the environmental impacts of this technology, it is still something that should be considered when mitigating ESG risk. Global financial regulators have been warning banks to approach cryptocurrency with caution for a number of years, particularly US regulators.
The Federal Reserve has taken several measures to limit the growth of cryptocurrencies, including rejecting crypto-focussed banks such as Custodia Bank from joining the Federal Reserve System. The Fed’s concerns about crypto are largely focussed around criminal activity and the use of Bitcoin and other cryptocurrencies by illicit actors such as terrorists and fraudsters. However, a 2022 study released by the White House, entitled Climate and Energy Implications of Crypto-Assets in the United States, does address concerns about crypto’s insatiable appetite for energy. It is worth noting that this document was published under the Biden administration. Trump has already expressed a desire to establish the US as the “crypto capital of the world.”
Proof of Work vs Proof of Stake
Most of the energy used by cryptocurrencies is during what is known as the Proof of Work (PoW) process , i.e., the Digital Ledger Technology (DLT) used to mine and verify crypto assets. This energy consumption could be massively reduced by switching to an alternative DLT model known as Proof of Stake (PoS) – which the White House report highlights. Most crypto currencies still use PoW, including Bitcoin, but some have begun switching over to PoS. In September 2022, Ethereum transitioned from PoW to PoS in a process it called “The Merge”. The transition reduced Ethereum’s energy consumption by a massive 99.84%.
While financial regulators have yet to mandate that crypto exchanges only list PoS cryptocurrencies instead of PoW ones, the idea is not entirely outside the realm of possibility. Many regulators worldwide are pushing for stronger ESG standards across financial services. A mandate for crypto exchanges to only list PoS assets could align with ESG policies and sustainability goals. This idea could be particularly appealing in jurisdictions with strong environmental policies, such as the European Union, where proposals to limit the use of high-energy assets like PoW cryptocurrencies have already surfaced (though they were rejected in 2022.) Although regulators cannot directly dictate the blockchain protocol (PoW or PoS) used by cryptocurrencies, they could influence the types of assets available on regulated exchanges by setting listing standards. This could be similar to securities and commodities laws that limit or approve certain financial products for public trading.
However, there would of course be many legal and logistical challenges to enforcing these types of requirements, given the decentralised nature of cryptocurrencies and their global reach. Regulating at the exchange level might simply shift PoW trading to less-regulated exchanges or decentralised platforms, making enforcement difficult.
FCA on cryptocurrencies and banks
In the UK, the Financial Conduct Authority (FCA) regulates crypto assets. Most of its regulations are focussed on crime prevention and risk management and ensuring that crypto assets are marketed correctly. But the regulator has touched upon the environmental impacts of the technology, if only briefly.
In a speech in 2023, Sarah Pritchard, the FCA’s Executive Director of Markets and International, said that the FCA was collaborating with international standard setting bodies and regulators to work towards an internationally coordinated effort on regulating crypto assets.
“One other area that we are committed to throughout our organisation and work is environmental, social and governance considerations,” she added. “We have set up workstreams to further understand what future crypto standards and requirements may mean when under the ESG lens. We will reflect on how best to factor ESG considerations in the design of the future crypto asset regulatory regime.”
In the UK Treasury’s Consultation and Call for Evidence on the future financial services regulatory regime for crypto assets, it talked (again, briefly) about the ESG considerations in future regulation: “Given parallels between crypto assets and securities markets, applying similar ESG-related reporting requirements may be a proportionate way of achieving our ‘same risk, same regulatory outcome’ principle,” it said. “However, given the nature of crypto assets, this may be more challenging. The way that consumers interact with the product and intermediaries is different to securities markets, not least because of decentralisation. Additionally, unlike in other parts of the sector, there is currently no agreed upon set of indicators or metrics for measuring the environmental impact of crypto assets.”
Financial regulators face an uphill battle in managing ESG risks associated with cryptocurrencies as they fundamentally challenge traditional regulatory structures. The technology’s decentralised, global and rapidly evolving nature means it will likely always be just out of reach of full regulation. However, by working towards more transparent, incentive-based and coordinated approaches, regulators may be able to encourage better ESG compliance over time.
View a full PDF version of this report, here.
