“Knowledge is power, but now power is knowledge,” was the theme of Adipec 2025. The global energy conference welcomed 220,000 visitors to Abu Dhabi last week. Plenty were there for the traditional business of buying and selling oilfield kit. But many wanted to explore the AI boom: its reality, its uses, and its energy appetite.
Dr Sultan Al Jaber, UAE Minister of Industry and Advanced Technology and managing director and group chief executive of Adnoc, told the opening audience that “electricity demand will keep surging through 2040, as power for data centres grows four-fold”, liquefied natural gas demand “will grow by 50 per cent … oil [demand] will stay above 100 million barrels per day beyond 2040”.
Part of this energy consumption growth is to meet the need for better living standards in developing countries, part for air-conditioning in a hotter world. But much of the acceleration in growth, particularly in western countries and the Gulf, is for electricity to run data centres.
That is at the core of the proposition that now, “power is knowledge” – power meaning the electricity to run and cool processing for artificial intelligence. The Enact majlis, hosted by Dr Al Jaber for a select group of senior leaders just before Adipec, highlighted this concern.
American attendees from government and corporations debated the right balance of gas, nuclear power and renewables, the reliability of battery systems for AI, and the struggle to get hold of gas turbines fast enough. European representatives worried that high energy costs make them uncompetitive.
In contrast, Gulf speakers were confident the region’s blend of low-cost natural gas and renewable energy with batteries gives them an edge in the AI duel. Last week, the UAE sealed a crucial deal to import advanced Nvidia AI chips, which the US has been trying to restrict to hold on to its technological lead.
Chinese officials were not prominent at the majlis, but would probably have been confident in their blend of electricity from modern and highly-efficient coal power stations alongside a rapidly-expanding and very cost-competitive hydroelectric, solar, wind and nuclear complex.
The current AI and data centre boom has much in common with the railway mania of 1840s Britain, the Roaring Twenties real estate, radio and car rush, or the early 2000s dot-com bubble. Capital spending was enormous, the immediate profitability of new technologies was overhyped, many investors lost their shirts. But the foundations of a new economy were laid; new fortunes emerged from the survivors, such as Amazon and Google today.
Whatever energy plan emerges has to take this into account. The lengthy investment cycles for traditional power generation raise the risk of overbuilding if AI stalls for a few years, if current approaches hit a dead end, or the circularity of spending between a few big players jams. That favours flexibility, optionality, and shorter-cycle developments such as solar with battery backup, which can be up and running within a year or two.
Overall, AI almost certainly means more energy consumption – but it will also improve energy efficiency, and boost output. AI systems themselves will become less energy-intensive and therefore more cost-effective. Simplified and energy-lean models may be adequate for many uses, such as China’s DeepSeek.
The direct value of AI in the energy industry comes less from generative AI – text and image generators like ChatGPT and Midjourney – and more from analytical, and then Agentic AI, understanding and controlling things in the real world.
For instance, at Adipec, Adnoc announced a partnership with oil services giant SLB to deploy an AI-powered production optimisation system across eight of its fields. Running remote and offshore installations or surveying large, dangerous or unreachable areas by drone or robot frees up people to do safer, more productive things.
Data centres can be valuable sources of flexibility in grids. A steady and predictable source of baseload demand, they can also be planned to turn down temporarily at times of maximum demand from other users, or when renewable energy output drops off. For example, in the UAE, solar generation ceases in the early evening while consumption for air-conditioning and home entertainment and meals preparation peaks.
For users, AI can improve energy efficiency and manage loads by scheduling when appliances or factory tasks run, phasing building cooling periods, optimising motor speeds, and diagnosing and eliminating excessive consumption.
The acquisition of more data and experience will add increasing value to such techniques. The industry also needs to build the human capability to work alongside AI without over-relying on it.
Beyond the improvement of existing operations, AI also offers the vista of radical advances. The Nobel Prize awarded last year to the developers of AlphaFold, which predicts protein structures, could be prophetic. The breakthrough might come from synthetic biology to make fuels or capture carbon dioxide, materials science to search millions of compounds for the ideal battery, or physics to control the limitless clean power of nuclear fusion.
Such game-changers would solve AI’s energy problem themselves. That would complete the circle: knowledge creates power, which creates knowledge. The strategic implications of that possibility, and the danger of missing out, not just for the energy industry but for the fate of nations, explain the intense AI interest in Abu Dhabi last week.
