Demand for electricity from data centres is growing fast– globally, it will more than double in the next five years. It is a real challenge which deserves a serious response. Data centres are placing real strain on electricity grids. Communities raise legitimate objections: changes to land use, higher domestic electricity bills, heavy draws on water for cooling. And nations are waking up to a strategic reality. Whoever controls the data centres controls critical infrastructure. Energy sovereignty is no longer just about oil and gas pipelines. It is also about who runs the computers.
These are serious issues. But the increase in demand from AI data centres, though large – estimated to double from 415 terawatt hours in 2024 to a projected 945 TWh in 2030 – is dwarfed by the power we use for air conditioning, a figure set to rise steeply as global temperatures climb. And efficiency improvements in AI models and data centre design could alter the trajectory significantly.
But AI is also a powerful tool for solving energy problems. And the solutions it is beginning to unlock are extraordinary.The future of energy is autonomous. That is not just an academic prediction. It is something I can see first-hand from the companies I chair and advise.
We are already in the first wave. AI agents are diagnosing the health of physical systems – offshore platforms, liquified natural gas plants, refineries – and prescribing actions to improve their operation. They are optimising whole electricity grids, matching demand and supply in real time and anticipating what will be needed hours ahead. They are reaching deep into supply chains, tying the needs of physical systems directly to planning and operations across manufacturing and logistics.
What probably comes next is full autonomy: agent-to-agent collaboration across the entire energy value chain. AI systems and advanced robotics acting as the primary control layer – negotiating, adapting, and optimising in real time, without human latency slowing them down.
Imagine a living operating system for the energy sector. One that continuously simulates thousands of scenarios and dynamically adapts as weather changes, demand shifts, and markets move. From autonomous refinery operations to the synchronisation of production, trading and delivery – a fundamentally transformed energy system, defined by continuity, speed, and structural efficiency. This will take more development of AI systems – to ensure they can deal with the messiness of reality and know when to refer to humans for guidance.
There is, though, a harder problem that is sometimes avoided in discussions of energy transition: aircraft, container ships and heavy goods vehicles on long routes. These sectors run on liquid fuels, and for the foreseeable future, electricity cannot replace them – the energy density of aviation fuel is roughly forty times greater than the best batteries available today.
We need a new generation of liquid fuels – created from low-cost, nationally controlled plant-based feedstocks, releasing significantly less carbon dioxide than fossil fuels, and cost-competitive with conventional hydrocarbons.
Fuels that give nations sovereign control of their energy supply.
The catch is that creating those fuels requires new organic catalysts – enzymes that do not exist in nature. This is where AI may achieve a breakthrough, using our understanding of quantum chemistry, with AI acting as the experimental scientist – running hypotheses, designing molecules, iterating with a physical laboratory. We may discover these catalysts within years. Success would open new energy sources for Europe, India, and parts of the world wholly dependent on imported fossil fuels today.
Newsletters
Choose the newsletters you want to receive
View more
For information about how The Observer protects your data, read our Privacy Policy
New materials have always driven energy revolutions – from steel to silicon. AI can accelerate materials discovery at a pace that would have seemed like science fiction a decade ago.
An example is the solar cell. The most advanced perovskite-silicon tandem cells have reached efficiencies in the laboratory, 50% higher than for today’s commercial silicon panels. The challenge is durability – perovskite degrades far faster than silicon – and solving that requires new coating materials that preserve performance over decades. AI is helping to design them.
Another example is electrolysis – the foundation of a potential green hydrogen economy – which today depends on rare and expensive platinum group metals. New AI-designed catalysts could change that equation entirely, and AI can also design materials to capture carbon dioxide from industrial processes and make battery recycling dramatically more effective.
The common thread is this: AI is not just optimising existing systems. It is helping us discover entirely new ones. And that distinction matters enormously.
The technology is here. We must invest, build, and make physical products from these discoveries that give everyone, everywhere access to affordable, clean energy. This is the foundation of development, of health, of human dignity. Getting this right may be the most important thing our generation does.
John Browne is the chair of Avathon; chair of Xyme; and former CEO of BP
Photograph: Jason Alden/Bloomberg via Getty Images