By Oliver Sartor, Chief Economist at Voltalis
Understanding the world of energy at times can appear to be quite paradoxical. For example, in 2024, for the first time in 50 years, oil accounted for less than 30% of global energy consumption, and low-carbon energy sources, which are growing faster than any other energy source, now exceed 40% of electricity production. Yet, neither oil nor coal consumption, nor greenhouse gas emissions (+0.8% in 2024), are decreasing at this stage.
The green consumption paradox
The answer lies in the fact that even though clean energy production have been growing exponentially, the consumption of energy – and especially electricity, which is growing at around 3% per year – has, until now, been growing just a little bit more.
As a consequence, some are beginning to wonder whether the energy transition is caught in a kind of “green consumption paradox”, whereby the world is developing lots of clean energy sources on the one hand, but, on the other hand, the inexorable rise in electricity demand means that we never get rid of an ongoing reliance on coal and gas power stations.
In this context, a new report published by the International Energy Agency is extremely informative. The Agency examined the expected impacts of AI and the digitalisation of our economies on energy systems. At first glance, it appears to confirm the idea of a green consumption paradox: it finds that global electricity consumption linked to AI and data centres is expected to rise significantly over the next 5 to 10 years.
Depending on assumptions about the industry’s energy efficiency and the speed of its development, electricity demand from AI is expected to more than double by 2030 to reach 945 TWh, which is an increase roughly equal to Germany’s total annual electricity consumption today. By 2035, this demand could triple to reach 1,200 TWh, which is roughly the electricity consumption of India today. This level of growth poses a particular challenge for countries such as the United States, which aim to host a disproportionately large share of this new industry on their soil.
Electricity demand shaped by new uses, of which AI is only one
However, a more nuanced view is also possible. Global electricity demand in 2023 plateaued around 30,000 TWh. In a central scenario where electricity consumption increases by about 800 TWh by 2035, AI’s growth would account for only around +2.7% of global demand today. It is certainly a challenge for the global energy transition, and for specific locations in specific countries, but it’s not an insurmountable problem globally speaking.
Other factors are set to drive much greater growth in electricity demand and alter its nature over the next decade. The rise of air conditioning and electric heating, the mass adoption of electric vehicles and the widespread electrification of industry and other forms of transport. The good news is that this signals the end of natural gas, oil and coal for these purposes. After all this is what the electrification of the energy sector is about: electricity replaces other sources of energy, notably from fossil fuels. Thus, assuming current trends in renewable energy deployment continue, emissions decrease.
What challenges does this pose?
But this also creates the real challenge for the energy transition, which lies in the addition of hundreds of millions of new electricity thirsty devices distributed throughout our energy system. Not only will they increase consumption and shift the times at which we use power, but they will also need to adapt to a far more variable electricity supply. The challenge with renewables is not that they can’t produce enough energy to meet our needs, but rather that they sometimes produce too much, and other times not enough, depending on the time of day, week, or month.
A true missing piece of the energy transition is therefore techniques for managing this diffuse electricity demand and steering it to align with periods when energy is available; using more when there is an over-abundance of renewable generation, and less when solar and wind production is low.
But can we really expect consumers to constantly react to weather variations every time they want to use energy? That would be unrealistic, even if shifting off-peak hours to sunny daytime periods is a good start. So, what’s the solution?
Digital technology to the rescue
The answer lies in digital solutions and AI, which enable smarter and more automated real-time control of devices. While there is a lot of AI hype, it’s also true that a number of key technologies are falling into place at precisely the right time to solve this very problem: falling prices for smart sensors to measure usage, technical advances in the Internet of Things (IoT) which enable automatic control and management of remote devices, cloud computing, powered by ultra-powerful AI algorithms, that allows real-time control of, for example, power demand whilst preserving user comfort using smart thermostats and the like.
Thus, millions of electrical devices such as air conditioners, heat pumps, heaters, water boilers, electric cars and batteries can still provide all their services for consumers just like before, while also being integrated into “virtual power plants” that manage and optimise demand on a large scale from the perspective of the energy system. Thus, they can ease pressure on the power system during peak demand (instead of relying on polluting, expensive power plants), and compensate when renewable energy is less available than usual.
The IEA itself has noted that we are heading into an “age of electrification”. In this new era, the potential of AI in managing global energy demand is not only vast, it is essential. Yes, AI will increase electricity demand, but the success of the energy transition also depends on it.