28 August 2025
The Equitable Path for AI Infrastructure
The rise of artificial intelligence has been compared to the industrial revolution and the internet in its transformative impact. But the infrastructure powering this revolution does not exist in a vacuum - it draws on the same critical resources we all depend on: power, water, and land.
So the question is: if data centres are becoming as vital as our energy sources and water systems, should we be treating them not as AI infrastructure but as climate infrastructure? And if so, how do we ensure it is equitable?
Data Centres as Critical National Infrastructure
Global demand for AI is driving explosive growth investment in data centre infrastructure. Today, there are around 11,800 operational data centres worldwide - with the United States alone hosting 5,388 facilities (as of March 2024). Germany follows with 512, then China (499) and Canada (366). The sector is scaling at breakneck speed: Big Tech companies are expected to invest $350 billion in data centres in 2025, rising to $400 billion in 2026, with global spending projected to exceed $3 trillion by 2029. Unsurprisingly, the number of data centres is set to keep climbing as AI’s footprint expands.
Given this trajectory, and their critical role in underpinning digital services and national security, it's clear why data centres are increasingly seen as critical utilities, fundamental to national resilience - just like energy grids or water systems. In the UK, data centres have officially been designated as Critical National Infrastructure, granting them the same status - and associated government coordination - as other essential systems.
In China, the “Eastern Data, Western Computing” (东数西算) initiative led by the government explicitly frames this as digital infrastructure for national resilience, alongside transport, energy and water systems. It also aims to move energy-hungry data centres out of densely populated regions to areas with abundant renewable energy.
And yet, while data centres are now recognised as being as vital as energy, transport, and water systems, they are largely built and expanded with communities and the climate as an afterthought.
Can Data Centres, although Critical, also be Equitable?
The Energy Problem
With the expansion of AI comes an equally explosive surge in energy demand. Training GPT-3 alone reportedly consumed 1.3 GWh of electricity, or roughly the annual power usage of 120 U.S. homes. Inference (running models at scale) often dwarfs this, locking in sustained demand spikes. The International Energy Agency (IEA) warns that by 2030, data centres could consume nearly 1,000 TWh annually - surpassing Japan’s current total electricity use. In the U.S., data centres already account for around 4% of electricity demand, projected to rise to 8% by 2030 without intervention, with over half still powered by fossil fuels. In Ireland, data centres consume 19% of national electricity, forecast to rise to nearly one-third by 2030; meanwhile, Denmark and the Netherlands have even paused new developments due to concerns over grid capacity and the pace of renewable transitions.
The Water Problem
Water, too, is a critical input for AI infrastructure. Many data centres still use evaporative cooling systems, which lower server temperatures by evaporating water into the air - highly effective, but extremely water-intensive, especially in hot or arid regions. In the U.S., data centres consumed approximately 163 billion gallons in 2021, and in The Dalles, Oregon, a single Google facility drew more than a quarter of the city’s water supply. About 40% of U.S. data centres are located in regions already experiencing high or extreme water stress. Globally, AI workloads could drive 4.2–6.6 billion cubic metres of water withdrawals annually by 2027 - more than half of the UK’s total annual water use.
The Equity Question
Energy and water are critical resources under serious strain, raising urgent questions about resource equity in the era of AI:
Whose grid reliability or water supply is compromised while tech companies scale?
Will upgrades to renewable energy and water infrastructure serve local needs - or be secured exclusively for corporate contracts?
Can energy systems decarbonise while absorbing soaring AI demand?
Who bears the cost of necessary infrastructure improvements?
And how do we ensure communities don’t shoulder environmental burdens while the tech industry reaps the benefits?
If data centres are to become genuine climate infrastructure, equity must be built into every kilowatt and every litre. As must sustainable practices.
What does 'Sustainable Data Centre' Really Mean?
The term ‘sustainable data centre’ gets thrown around - but what does it really mean? At its core, a sustainable data centre is one that minimises its environmental footprint across energy, water, materials, and e-waste, while also supporting the communities it operates in.
On the energy side, this means carbon-free / renewable energy. Backed by over 100 operators and endorsed by the European Commission, the Climate Neutral Data Centre Pact commits signatories to run on 100% carbon-free energy by 2030, achieve a PUE of 1.3–1.4 depending on climate, and integrate waste heat into district heating networks where possible. In Finland and Denmark, data centres already pipe excess heat into municipal systems, warming homes with what would otherwise be wasted energy. Similarly, many new builds are adopting modular construction which can reduce emissions by 45%, with low-carbon concrete and recycled steel, cutting down embodied emissions.
Water efficiency is another key point - traditional evaporative cooling can consume millions of litres a day as mentioned above, but some operators are now shifting towards closed-loop, waterless, or liquid immersion cooling to cut demand dramatically. In England, a 2025 study found that 64% of data centres used less than 10,000 m³ of water annually - less than a leisure centre - and 51% operated with waterless cooling systems. Circular water strategies - such as rainwater harvesting and recycling - are also gaining prominence as framework-level innovations.
Beyond water and energy, sustainability also means tackling the short lifecycle of servers: AI hardware often lasts only 3-5 years, creating mounting e-waste. Analysts estimate that the AI boom could add 1.2-5 million tonnes of extra e-waste annually by 2030, as outdated servers and chips are discarded rather than repurposed. Forward-looking operators now commit to refurbishment, resale, and rare-earth mineral recovery, for example data centre operator Equinix has committed to 100% reuse or recycling of retired IT equipment across its global sites.
We can reduce the emissions of a data centre with renewable energy sources, innovative cooling solutions for efficient water usage, green building materials, circularity and recycling e-waste, but there is so much more that can be done on the plot of a data centre to curb emissions - this is where carbon removal solutions come in.
Data Centres as Carbon Infrastructure
If data centres are to become genuine climate infrastructure, they must go beyond ‘standard’ sustainability and actively contribute to removing emissions with carbon removal embedded. A wave of new technologies is emerging that could transform them from resource-hungry facilities into hubs of negative emissions, leveraging their access to renewable energy, waste heat, and advanced AI-driven innovation:
Direct Air Capture (DAC):
Data centres and DAC could be mutually beneficial as data centres typically have high-volume green electricity connections through PPAs and on-site generation, which could play a crucial role in offsetting DAC’s high energy costs.
Waste heat generated by data centres could be repurposed to improve DAC efficiency.
Microsoft is running a pilot system called DACinDC repurposing 70–85% of data centre waste heat to drive Direct Air Capture modules integrated into facility infrastructure.
BECCS (Bioenergy with Carbon Capture and Storage):
Arbor, a BECCS startup, gasifies waste biomass, burns it with oxygen, and achieves ~99% CO₂ capture efficiency.
By-products include supercritical CO₂ (easy to store) and water that can be reused for irrigation or data centre cooling.
Potential to deliver removals at under $100 per tonne of CO₂.
In the UK, Drax Power is converting former coal units to biomass + CCS and has held discussions with data centre operators about partnerships.
AI-Designed Carbon Capture Materials:
Amazon is piloting AI-developed sorbent materials with Orbital Materials that act as atomic “sponges” for CO₂, improving capture performance up to 10× vs. conventional solutions.
In the UK, Orbital Materials and Civo Cloud are testing these sorbents directly inside data centre cooling and airflow systems, effectively turning the facilities themselves into active carbon removal devices.
AI Infrastructure as Climate Infrastructure - The Path for Equity
To avoid replicating the harms of past industries, data centres must be built as climate infrastructure - powered by renewables, conserving water, designed for circularity, and embedded with carbon removal.
AI has the potential to transform society, but its infrastructure draws on the same scarce resources - energy, water, and land - that communities depend on. Without equity and sustainability at the core, data centres risk becoming the new fossil fuel industry: powerful but extractive.
To commit to equity, the data centre industry has to ensure communities are at the centre of this infrastructure, engaging and shaping them as a fellow neighbour, not simply as an obscure black box draining resources.
For more information on data centres as climate infrastructure, climate solutions and carbon removal, please contact rachael@opna.earth.
