“Can machines think?”
That was the question posed by the ‘father of modern computing’, British scientist Alan Turing, in a landmark paper published exactly 75 years ago1.
At the time, the question was controversial. However, with the far-sightedness for which he has become known, Turing – who anticipated the rise of artificial intelligence (AI) – also predicted that the question would become irrelevant. Whether or not we would ever accept that machines could truly ‘think’, they would one day, he believed, compete with human creativity.
Today, there is little room for debate. Less than three years since OpenAI’s ChatGPT was launched – the first of many products in a rapidly expanding ecosystem – the disruptive power of generative AI’s ability to ‘think’ is felt everywhere: in our consumer, health, energy and industrial systems, in the defence sector, in social media and politics.
The disruptive power of generative AI’s ability to ‘think’ is felt everywhere
AI’s potential is so great, and the pace of advancement so quick, that governments around the globe are scrambling to secure strategic independence – in AI itself, in related industries, and in numerous AI-enhanced sectors. In tandem, some of the world’s biggest technology companies are building ever more computing power as they compete to get to the front of the AI pack.
As the AI arms race escalates, however, scientists and policymakers are increasingly asking: what impact will this new technology’s demand for resources have on the physical world? And how could investors benefit from – and play a part in – efforts to build a more efficient and sustainable AI future?
Sign up for our newsletter
The concrete cloud
It’s easy to think of AI as having no physical presence, existing only as lighter-than-air electrical pulses somewhere in ‘the cloud’. In the city of Mesa, Arizona, however, it quickly becomes clear that AI is less cloud, more concrete. There, in a municipality of just over 500,000 people, 15 large data centres, covering a total of 1,500 acres, have been built, approved or are under proposal, including campuses owned by Apple, Google and Meta.2
These data centres – and the thousands like them around the world – are the home of the internet, containing the servers and hard drives that power our searches, process our online purchases, and store our text, photos and videos.
They are also the home of AI, which relies on hundreds or even thousands of advanced processors to conduct billions of computations. Put simply, AI demand is data centre demand. Between now and 2030, it’s anticipated that the world’s total data centre capacity will more than triple,3 and that data centre electricity usage will grow four times faster than any other sector to reach 3% of all global electricity consumption.4
Read also: The AI-health revolution – future-proofing our golden years
AI’s demand for computing is making data centres more power hungry. It is also making them thirstier
Competing with computing
This is where cities like Mesa come in. With an average of nearly 300 sunny days per year,5 Arizona’s solar power potential makes it an ideal place to build these vast, electricity-hungry monoliths – Google, for instance, aims to deploy a combination of wind and solar power to achieve 80% carbon-free electricity in its Arizona data centres by 20266.
However, while year-round sun provides the power, Mesa’s dry climate means the city is often beset by water scarcity. This is a particular problem when it comes to large data centres. With water more than 20 times more efficient than air for transferring heat7, these data centres have come to rely on liquid cooling systems to manage the heat generated by advanced AI computations in increasingly heat-dense AI data centres. As a result, some centres now consume up to 5 million gallons of freshwater per day – the equivalent of the usage of a town of 10,000–50,000 people – for both energy generation and cooling.8
Across the world, this conundrum – where locations chosen for their solar potential are often both water-poor and at greater need of cooling – is set to become more intractable. AI’s demand for computing is making data centres more power hungry. It is also making them thirstier – within just two years, global AI computations are expected to use as much as 1.7 trillion gallons of water annually, more than six times the water consumption of Denmark.9
In Mesa, Meta has pledged to support regional water conservation projects… meaning its new USD 1 billion data centre will become water-positive by 2030
In Mesa, though the jobs and investment the data centres brought were initially welcomed, residents are increasingly uneasy, and policymakers have recently tightened the planning regulations, including bringing in greater scrutiny of the centres’ water usage.
Mesa is not alone. In The Dalles, Oregon, for example, a local newspaper won a 13-month legal battle with the city council to publish the water usage of a data centre – they discovered the centre was responsible for more than a quarter of the city’s entire annual consumption10. In Spain, Uruguay, Chile and even the Netherlands there have been protests against new data centres by farmers and local communities worried that they may soon be competing with AI for their water.
Water cooling innovations – the birth of an industry
For data centre owners, efficient cooling has shot up the agenda. In the summer of 2024, for instance, Microsoft announced it would implement closed-loop water cooling systems across all its new data centres, with the company’s first zero-water-consumption data centres to be operational in Arizona and Wisconsin from 2026. By replacing evaporation with a sealed loop, the firm promises to save more than 125 million litres per year at each of its new centres, showcasing the potential of liquid cooling technologies as a key sustainability solution.11
For data centre owners, efficient cooling has shot up the agenda
A flourishing ecosystem of specialist businesses is emerging. US-based Vertiv, for example, offers a hybrid cooling system that combines closed-loop liquid with air-cooling and can be plugged into existing data centres, eliminating water waste and reducing energy use without requiring a full refurbishment.12
Some new data centres are opting for immersion cooling, where specially designed servers operate while submerged in non-conductive coolant. With no fans or pumps involved, the systems – such as those developed by UK- and US-based Iceotope – mean zero or minimal water consumption and energy usage cut by more than 90%.13
US-based start-up Subsea Cloud is targeting deeper immersion still, with its pilot data centre pod – known as the Jules Verne – sited on the floor of the Pacific Ocean. With plans for further test sites in the Gulf of Mexico and the North Sea, the company aims to demonstrate that it can eliminate freshwater consumption and cut energy use dramatically, even in warmer seas.14
Microfluidic cooling is also gaining traction. By etching tiny channels directly into AI chips, liquid coolant can flow inside the silicon itself, removing heat far more efficiently than traditional cold plates. Microsoft recently demonstrated the approach, showing it could cool servers running core services with up to three times the effectiveness15.
Read also: Nature-positive investments gain traction
Investing in the planetary transition
In 2025, Google, Meta, Amazon and Microsoft will spend a combined USD 350 billion on building and equipping data centres – across the US it’s thought that AI investments alone could grow the economy by as much as 0.7% this year, representing half the Federal Reserve’s 1.4% growth forecast.16 Europe is also determined to capitalise on the opportunity – French President Emmanuel Macron has announced an EUR 109 billion private investment in AI in France17, while the European Commission’s ‘InvestAI’ will mobilise EUR 200 billion to support AI development across the EU bloc18.
This vast spend is creating investment opportunities across multiple sectors – from renewable energy projects and battery storage, to construction, to the cutting-edge chips and server hardware that underpin new AI models. According to capital markets firm Jefferies, demand for liquid cooling systems will be supported for at least the next 5–10 years, with the market set to grow to a value of almost USD 28 billion by 2030.
At Lombard Odier, we are convinced that AI will become an indispensable ally in our ongoing efforts to build a sustainable future, and that it will be a key catalyst in the transition to a new economic model – one that is net zero, nature-positive, socially constructive and digitally enabled.
Through our Planetary Transition investment strategy, we aim to capitalise on this shift, seeking out opportunities as businesses work to end their transgression of planetary boundaries (such as excessive emissions and freshwater withdrawals). As part of this strategy we invest in firms accelerating the AI revolution and those innovating the efficient cooling technologies needed by cutting-edge data centres.
In Mesa, Arizona, fear over freshwater consumption is driving change. Meta has pledged to support regional water conservation projects so as to replenish local supplies by more than they use, meaning its new USD 1 billion data centre will become water-positive by 203019; meanwhile the new Edged Energy data centre has been designed to use waterless liquid cooling from day one, saving 94 million gallons of water annually20.
As innovations in liquid cooling technologies gather pace, other AI data centres are sure to follow suit, reducing water consumption and supporting sustainability solutions. For Mesa and the many places like it around the world, residents may once again see data centres as a boon to the local economy, rather than as competition for their precious drinking water.