AI energy and resource use: Are companies pricing it into net zero plans?

As AI drives up electricity demand, companies and investors face tough questions: are net zero plans adapting to the reality of higher power use, bigger infrastructure footprints and more complex transition risks?

AI has a complex relationship with climate change, offering both potential benefits and notable environmental costs. These benefits are felt across the economy, particularly through greater operational efficiencies and more accurate climate forecasting. However, AI, with its data centres, networking equipment and cooling systems, is also driving up electricity demand and resource utilisation.

According to the International Energy Agency (IEA), data centres accounted for around 1.5% of global electricity consumption in 2024. Demand is expected to rise sharply as AI adoption spreads – indeed, this figure has grown by 12% each year since 2019, with the IEA expecting electricity demand from data centres to double by 2030¹. And, while the IEA has found that AI and its components are getting more efficient, it’s also being used more widely, offsetting those efficiency gains.

It’s an issue not just affecting the technology sector. If electricity demand from AI and cloud infrastructure rises faster than low-carbon supply, it can make decarbonisation harder by increasing reliance on fossil fuel generation, delaying the clean-up of grids or raising the cost of securing new clean power. It can also increase embodied emissions or the upfront carbon footprint of the building of new facilities and the manufacture of semiconductors, servers and other hardware. For companies with net zero targets, and for investors trying to assess whether those targets remain credible, that creates a new layer of scrutiny.

Additional electricity demand does not automatically translate into high emissions, of course, but nor is it neutral. If new data centre infrastructure development is met by clean power that wouldn’t have otherwise existed, the impact on emissions may be limited. However, if it depends on a slower decarbonisation of the grid, greater use of gas-fired generation or optimistic assumptions about future clean power, the picture changes.

Carbon Brief has argued that, if even just a small amount of the power demand from UK data centres is met by gas, their emissions impact could be much higher than earlier official estimates³. That doesn’t mean that AI growth is incompatible with net zero, but it does suggest the answer depends heavily on infrastructure, timing and transparency.

The success or otherwise of SpaceX will be judged by how it performs over the coming months and years.

For investors, this isn’t just an environmental issue, it’s a question of transition risk and disclosure quality. AI is fuelling growth expectations and attracting large amounts of capital into data centres, semiconductors and related infrastructure. But that same expansion can increase exposure to energy-price volatility, grid bottlenecks, delays in receiving permissions, and tougher scrutiny of climate claims.

That tension is particularly marked in terms of real assets. MSCI has noted that data centres are attracting strong demand but also risk higher portfolio emissions and more difficult decarbonisation pathways⁴.

For equity investors, the key question is whether companies expanding AI capacity are also making net zero pathways more expensive or less credible. Higher electricity use may increase operational exposure, while new data-centre construction can push up Scope 3 emissions.

Some companies are already making the link. In its 2024 Environmental Report, Google said that its emissions had risen since 2019, citing increased data-centre energy consumption and supply-chain emissions linked in part to AI growth. The company emphasised efficiency improvements, clean-energy procurement and its 24/7 carbon-free energy ambition, but it also made clear that the path to net zero has become harder.

While, in its 2024 Environmental Sustainability Report, Microsoft showed that while its direct operational emissions were falling, its Scope 3 emissions had risen significantly, driven largely by data-centre construction and hardware procurement.
This openness it useful, but it’s not universal. One of the problems facing investors is that disclosure can be patchy. This can make it more difficult to consider the impact of AI-related growth on net zero pathways.

Measurement too remains challenging. Many companies report emissions at a company or business-unit level, but don’t separate AI growth from wider digital infrastructure, or disclose the energy use of specific AI services or products. Investors are therefore often left to infer the scale of the challenge from things like rising capital expenditure, data-centre announcements or broad changes in emissions totals.

It’s not just the impact on energy that companies and investors need to consider. The knock-on effect on water consumption also creates challenges. Electricity generation demands vast amounts of water, for example, to heat fossil fuel power plants or nuclear turbines. Another indirect drag on water resources is the manufacture of hardware, such as servers and chips. More directly, data centres have significant cooling requirements. Globally, the IEA estimates that the data centre sector consumes some 560 billion litres of water annually, and projects this to increase to 1,200 billion litres by 2030⁵.

With finite water resources, this can create supply risks for local communities and wider threats to nature and the environment. In the UK, for example, the Environment Agency predicts a shortfall of nearly 5 billion litres of water per day by 2050, the equivalent of over a third of the current public water supply⁶. To address these risks, some hyperscalers have set targets to lower water consumption and invested in technology to diversify cooling systems, create closed loop systems or harvest rainwater.

A growing number of institutional investors are recognising the impact AI can have on all aspects of ESG – impacts we highlighted in our report Governing the Algorithm: Investor priorities for responsible AI. Actions to address these impacts include:

• Integrating AI into investment analysis through evaluating AI-related risks and disclosures.
• Stewardship and engagement; in 2024 we formally introduced AI and ethics as a stewardship theme within our Stewardship Policy and Voting Guidelines.
• Set expectations aligning with evolving standards to drive consistency and transparency.

When it comes to AI, the climate benefits and risks, aren’t straightforward.

But the technology is increasingly in demand and growing at pace. For investors, the question is whether companies, regulators and markets are starting to treat that demand as strategically important to transition planning. The answer will shape not only the credibility of net zero commitments, but also potentially capital flows.