© Adobe
As with any investment, your capital is at risk.
While artificial intelligence (AI) is driving up energy demand fast, it could, in time, also help bring it down.
Much of today’s energy systems still rely on wide safety margins, limited data and slow decision-making.
Applied well, AI could ease those bottlenecks by making energy systems more efficient. And in energy, that usually means lower emissions.
However, used poorly, AI could become another source of pressure in an already difficult transition.
The scale of the bill
Datacentres consumed approximately 415 terawatt-hours (TWh) of electricity in 2024, roughly 1.5 percent of global demand. That translates to about 0.5 percent of global carbon dioxide (CO₂) emissions, depending on how you measure it.
So far, AI workloads have been a relatively small part of that, likely about 15 percent, or roughly 27 million tonnes (Mt) of CO₂ on current estimates. But that share is growing rapidly.
By 2030, total global datacentre electricity consumption is projected to approach 1,000TWh, just under 3 percent of global electricity demand, with AI driving most of that increase.
Datacentre electricity demand could more than double by 2030
What that means for emissions depends largely on the grid. If current trends continue and average grid intensity – the amount of CO₂ emitted for each unit of electricity generated – falls to about 360g CO₂ per kilowatt-hour (kWh) by 2030, then 1,000TWh of demand would result in roughly 360Mt of CO₂.
In a faster-decarbonising system, aligned with a net zero pathway, where intensity approaches approximately 165g CO₂ per kWh, the same demand would translate to nearer 165Mt.
Emissions depend on the carbon intensity of the electricity grid
In either scenario, new capacity must be built before any AI-enabled efficiency gains appear, and those gains are not guaranteed.
In the near term, the most credible case for AI is not in broad claims about ‘solving climate change’, but in areas where computing power has been a real constraint and where the business case is already clear. Two areas stand out: power grids and industrial processes.
Power grids: the missing layer in the energy transition
Grids are engineered for reliability, not flexibility. They are built to handle peaks and outages, so they tend to run well below their true capacity most of the time.
That leaves expensive slack in the system while, paradoxically, local bottlenecks curtail clean power generation.
AI can help close that gap with three applications close to commercial reality:
- Dynamic line rating (using real-time weather and equipment data to work out how much power a transmission line can carry safely) and congestion management (reducing bottlenecks on the electricity network): running transmission lines closer to real-time limits, which can unlock capacity without building new infrastructure.
- Renewables integration: improved forecasting and dispatch – deciding when different power sources should generate electricity – for renewables reduces wasted generation and lowers reliance on gas- or coal-fired generation when renewable output is low.
- Demand flexibility: coordinating millions of flexible loads, whose electricity usage can be shifted to another time, such as electric vehicle (EV) charging, heating/cooling and industrial motors, means they use power when clean supply is abundant.
Even under simple assumptions, the impact can be meaningful. In a modest 2030 case, where overall system use is broadly unchanged, but AI improves the use of renewables and demand flexibility, the system could deliver about 415TWh more clean electricity or avoid roughly 150–170Mt of CO₂, depending on the grid intensity.
With wider adoption, combining better forecasting, smoother grid operation, and more flexible demand could raise that to about 1,600TWh, equivalent to roughly 550–650Mt of avoided emissions.
In a more ambitious case, if grid operators and market designers build AI into grid management and electricity markets, the effect could be larger still, up to about 3,000TWh and close to 1 gigatonne (Gt) of avoided emissions.
These are not forecasts but they illustrate the scale of the opportunity. Small percentage improvements in how the grid is used and how flexibility is managed can translate into very large quantities of clean power delivered to end users.
Industry: shaving losses from energy-intensive processes
Heavy industry, materials and manufacturing account for about 30 percent of global energy use and roughly 12-13 billion tonnes of carbon dioxide emissions.
These sectors are hard to decarbonise quickly because they depend on heat and complex chemical processes. Many plants also run below their best: inputs vary, equipment is often ageing, safety margins are built in, and processes can be long and complicated.
That’s where AI can help. In the near term, the opportunity is straightforward: cut downtime, reduce defects and run processes closer to their ideal operating point.
The first wave is monitoring and maintenance: spotting problems early, predicting when equipment will fail and improving quality control.
Beyond that, the bigger gains come from optimising whole processes. That means using detailed sensor data to model how plants actually behave, and testing improvements before making changes in the real world.
There is also potential in materials science, finding ways to use less energy or to develop entirely new compounds.
These industries use so much energy, that even small improvements can add up quickly. With widespread adoption, the savings could reach hundreds of terawatt-hours of electricity each year, and potentially even more in reduced fossil fuel use.
Under simple emissions assumptions, that could translate to several hundred million tonnes of avoided CO₂ annually, adding up to several billion tonnes cumulatively by 2030, depending on how quickly these tools are adopted.
This is not a single breakthrough. Rather, it is about many steady, incremental gains, but at this scale, those efficiency gains matter.
The longer-term opportunities: bigger gains, greater uncertainty
Beyond grids and making industrial processes run more efficiently, AI’s role becomes harder to pin down. Here, it is less about running systems better and more about discovering new ways of doing things.
Food systems are one candidate. If AI accelerates the development of new proteins, fermentation methods and processing techniques, it could help make alternatives to animal products cheaper and more appealing.
The climate and broader environmental impacts could be significant. But this depends heavily on what people are willing to eat, how regulators respond and whether the economics work at scale.
Direct air capture, a technology that removes carbon directly from the air, is another. The slower emissions fall, the more we may have to rely on removing CO₂ from the air to limit further climate damage.
Making that viable comes down to efficiency in chemistry, materials and energy use. These are all areas where AI could help, but progress is uncertain.
Weather and climate modelling should also improve. More detailed forecasts can help countries prepare for extreme events and help energy systems make better use of renewable power. The emissions impact is indirect, but the benefits for resilience could be substantial.
What could limit the gains?
Two risks stand out:
- ‘Brown AI’: the risk that the same tools used to improve efficiency in clean power production can also make fossil fuel production more efficient. AI can make extraction, refining and distribution cheaper and more efficient, which may prolong the life of high-emitting assets.
- Rebound: the risk that efficiency lowers costs, which then increases demand. Without carbon pricing, regulation or other constraints, some of the efficiency gains may be offset by higher overall consumption.
Conclusion
AI is not a climate solution on its own. It increases energy demand in the near term, and in many regions, that means higher emissions. Over time, it could help reduce both, but only if companies and grid operators decarbonise quickly and apply AI to the parts of the system where inefficiencies are greatest.
Case study: Arm
Arm Holdings is a UK-based semiconductor and software design company best known for developing and licensing energy-efficient processor designs used in almost every modern smartphone.
As AI adoption grows, energy efficiency is becoming more important. AI is power-intensive, particularly in data centres, but it could also help improve how energy is generated, distributed and used.
Many of the advanced technologies needed for a net zero energy system – from cleaner power grids to more efficient industrial processes – are to benefit from faster data processing, better sensors and more responsive control systems.
© Shutterstock / FOTOGRIN
Arm is well-placed within that shift. Its core strength is low-power chip design and its technology is widely used in devices that process data close to where it is generated, known as edge devices.
This gives Arm exposure to two related themes: improving the energy efficiency of data centres, where power consumption is becoming an increasingly important constraint, and enabling the sensing and control systems needed across power grids, industry, buildings and transport.
That does not make Arm a pure climate investment. Its growth will still depend on demand for semiconductors, competition from other chip architectures and the pace at which AI workloads move between data centres and edge devices. But if energy efficiency becomes a more important constraint on AI deployment, Arm’s low-power design heritage could become a valuable advantage.
As AI workloads become more widespread and power consumption becomes a greater constraint, Arm's energy-efficient architecture may also help it gain share in markets such as servers and personal computers.
Risk factors
The views expressed should not be considered as advice or a recommendation to buy, sell or hold a particular investment. They reflect opinion and should not be taken as statements of fact nor should any reliance be placed on them when making investment decisions.
This communication was produced and approved in June 2026 and has not been updated subsequently. It represents views held at the time of writing and may not reflect current thinking.
Potential for profit and loss
All investment strategies have the potential for profit and loss, your or your clients’ capital may be at risk. Past performance is not a guide to future returns.
This communication contains information on investments which does not constitute independent research. Accordingly, it is not subject to the protections afforded to independent research, but is classified as advertising under Art 68 of the Financial Services Act (‘FinSA’) and Baillie Gifford and its staff may have dealt in the investments concerned.
All information is sourced from Baillie Gifford & Co and is current unless otherwise stated.
The images used in this communication are for illustrative purposes only.
Important Information
Baillie Gifford & Co and Baillie Gifford & Co Limited are authorised and regulated by the Financial Conduct Authority (FCA). Baillie Gifford & Co Limited is an Authorised Corporate Director of OEICs.
Baillie Gifford Overseas Limited provides investment management and advisory services to non-UK Professional/Institutional clients only. Baillie Gifford Overseas Limited is wholly owned by Baillie Gifford & Co. Baillie Gifford & Co and Baillie Gifford Overseas Limited are authorised and regulated by the FCA in the UK.
Persons resident or domiciled outside the UK should consult with their professional advisers as to whether they require any governmental or other consents in order to enable them to invest, and with their tax advisers for advice relevant to their own particular circumstances.
Financial Intermediaries
This communication is suitable for use of financial intermediaries. Financial intermediaries are solely responsible for any further distribution and Baillie Gifford takes no responsibility for the reliance on this document by any other person who did not receive this document directly from Baillie Gifford.
197613 10063780
