It has become clear to everyone that uses the internet that artificial intelligence (AI) is the next big thing. Nearly every company is releasing an AI chatbot/AI product/AI service to enhance their customer experience and improve efficiency of their products.
It is true that AI can improve the lives of customers in many ways, such as facilitating real-time responses, providing seamless interactions, and overall reducing processing errors within banking and other industries.
However, we are also all aware that there are significant negative aspects to AI – people are worried about AI taking over their jobs, AI has increased the believability and accuracy of fraud and hacking schemes, and there have been major hits to our own
minds in our ability to process and generate information without automation, and to be creative without digital stimulation.
On top of this, AI is drastically damaging our environment. Research indicates that AI chatbots such as ChatGPT demand a massive amount of energy. The computational power needed to power AI is
doubling every 100 days. According to the International Energy Agency (IEA), asking ChatGPT a question uses nearly 10 times the electricity than a Google search. AI is
estimated use up to 6.6 billion cubic meters of water use by 2027, which is two-thirds of England’s annual consumption.
Data centres that have had the benefit of tax breaks and discounted electricity over the years are now becoming increasingly disruptive to the power grids of multiple countries thanks to the electricity that AI demands. Data centres
currently account for 3% of global energy consumption, releasing as much carbon as the entire country of Brazil.
Because of AI, data centres will be consuming
4.5% of global energy generation in 2030. The
IEA also predicts that data centres will be using 1000 terawatts hours (TWh) of electricity annually by 2026, equivalent to the electricity consumed by the entire country of Japan. In just the US, data centres are predicted to consume 6% of power across
the country by 2026.
Google’s carbon emissions rose 14% in 2023 to 14.3 million metric tons, which aligns with their recent decision to abandon carbon neutrality.
Microsoft’s carbon emissions have increased by 30% since 2020 because of data centres.
According to
Forbes, Morgan Stanley reported that global power usage will triple in 2024 from 15 TWh in 2023 to 46 TWh, and Wells Fargo estimates that AI power will surge 550% in the next two years, from 8 TWh to 52 TWh in 2026, and rising 1150% more to 652 TWh by 2030
due to AI training demand.
Though sustainable action has been a priority for businesses throughout the fintech sector, carbon emissions are hiking due to the massive power crisis caused by AI. As more big tech conglomerates race to develop the best AI models, sustainable goals are
being pushed to the backburner.
How are regulators and governments approaching data centres?
Data centre developments need to be regulated to manage the energy glutton that AI is morphing into.
Ireland‘s EirGrid proposed a moratorium on new data centres in Dublin until 2028, yet there are still data centres being built in the country which are disrupting the power grid. AWS has restricted power users in a region of Ireland where it’s data centre
is, and there are politicians are pushing for further developments despite concerns.
While data centres were previously established in more remote areas, there has been a push from companies to make them closer to cities to ensure faster response times.
In the US, utilities are worried about the demand and the impact of more data centres on energy infrastructure. The energy usage from data centres puts pressires on utilities beyond electricity; water prices are on the rise as it is used in hordes for cooling
overheated data centres. The water used to cool data centres needs to be clean and pure to avoid corrupting the machinery, putting data centres in contest with water needed by people to clean, wash, and cook.
Water is being consumed at a rapid rate by big tech data centres, and consumed water does not get returned to its source.
Google’s data centres consumed 5 billion gallons of fresh water in 2022, using 20% more water in 2022 than 2021. In the same year, Microsoft increased its water usage by 34%. In recent years, these numbers can only be going up.
Virginia currently has a huge base of data centres in the northern part of the state, and a few years ago governments of
Connecticut, Maryland, Mississippi, and more were clamouring for data centre developments. However, there is now an increased amount of concern for the burden this will cause on utilities, that may make them move from clean energy to fossil fuels to keep
up with demand.
Along with Virginia, countries that have grown into technology hubs such as Singapore and the Netherlands are pushing for regulation to control data centre development. However, it is clear that in the world we live in that capital gain from data centre
development is prioritised over the impact data centre it has on affordable utilities for everyday consumers, especially from government officials who want to make that bank.
These issues have led to
environmental complaints in Chile and Uruguay, where the governments have allotted land for Google to build data centres where they would be taking water from local water supply, which is essential to the public amidst droughts.
People cannot be left underserved and overpriced on utilities for the sake of big tech development.
What can a sustainable data centre look like?
There are still developments underway to generate software and AI models that consume less energy, but there are significant breakthroughs that need to be made. At the pace at which AI is rising, it seems unlikely that big tech will hit the pause button
to wait for sustainable solutions (again, money).
Big companies are working towards their net zero goals, focusing on building wind and solar farms to generate clean energy, however it is not their sole source of electricity as wind and solar are not consistently reliable, especially in various weather
conditions, thus the power demand that AI has may not be matched by these resources.
Meta has bragged of sustainable energy powering its facilities, but has installed backup diesel generators in Ireland.
AI energy consumption can be reduced from 15% to 12% by cutting down AI training phases of models, with the consequence that GPU tasks
will take 3% longer.
AI making a difference in sustainability
What cannot be ignored is that AI could also make an enormous impact on sustainable action in a myriad of ways. The World Economic Forum states that AI can be used as a tool for establishing a sustainable infrastructure.
AI can be used for everyday needs to find the most eco-friendly groceries for your household, a carbon-neutral way to travel, and such to larger endeavours such as advising how companies can reduce carbon emissions in their supply-chain, modelling climate
changes, tracking endangered species, monitoring deforestation, using data to forecast where homes can use heating more efficiently, and more.
In summation, AI could be a major factor in enabling decarbonisation. Yet, AI needs to be managed in a way where it can be used as a tool to help the world, but also be controlled in its energy consumption.