Energy officials were not discussing tanker routes or oil embargoes in Washington this summer. They were discussing gigawatts. In particular, the number of gigawatts that could be brought online quickly enough to power data centers that are educating the next generation of AI models. With no televised pipeline disputes or dramatic OPEC meetings, just warehouses full of servers running around the clock and using more electricity than some small nations, this may be the quietest energy revolution in decades.
The International Energy Agency estimates that in 2024, data centers will use about 1.5% of the world’s electricity. That figure seems small until you travel to northern Virginia, where highways are lined with low, windowless buildings that are all throbbing with air conditioners and gated entry. Data centers now make up over 10% of the electricity demand in some U.S. states. Ireland’s percentage has surpassed 20%. It is difficult to overlook its physicality: substations quietly growing behind chain-link fences, transmission lines rising behind suburban neighborhoods.
| Subject | Global Energy & AI Transition |
|---|---|
| Key Institution | International Energy Agency |
| Data Centre Electricity Use (2024) | ~1.5% of global electricity |
| Nuclear Share of Global Power | ~10% |
| US Policy Initiative | America’s AI Action Plan |
| Key Voice | Doug Burgum |
| Industry Contributor | Deutsche Bank |
| Reference | https://www.iea.org |
It seems as though the new indicator of strategic strength is electricity rather than oil.
Hydrocarbon control was a focal point of energy politics for decades. The focus of the discussion now is on who can produce large amounts of reliable, reasonably priced, low-carbon electricity. Doug Burgum recently framed the development of AI as a race to generate more power as well as smarter algorithms. He declared, “Whoever has the most electricity will win.” It sounded direct. Maybe on purpose.
It’s hard to ignore the logic. Massive amounts of computing power are needed to train large AI models, and computing power almost immediately translates into electricity demand. Investors appear to think that energy security and AI leadership are inextricably linked. The Gulf states are making significant investments in AI infrastructure and LNG exports, China is speeding up nuclear construction and grid expansion, and the United States has announced its AI action plans. As I watch this develop, it seems more like a geopolitical recalibration than a tech story.
Once largely discussed in relation to climate change, liquefied natural gas is currently being reexamined as a bridge fuel for demand in the AI era. Data centers need consistent baseload power, which LNG plants can supply more quickly than new nuclear facilities. Executives openly discuss combining gas and carbon capture to meet sustainability and reliability goals in conference rooms from Houston to Abu Dhabi. Although it’s still unclear if climate activists will be satisfied with this compromise, ideological arguments are often stifled by urgency.
At the same time, nuclear energy is seeing a comeback that was improbable just ten years ago. Almost 10% of the world’s electricity comes from nearly 420 reactors. There are currently over 60 new reactors being built. Technology companies that previously shunned the term “nuclear” are now supporting efforts to increase capacity, claiming that 24/7 AI services require equally consistent generation. The rediscovery of atomic energy in Silicon Valley has an almost ironic quality.
However, AI is changing supply as well as driving demand. Machine-learning systems are being used by utilities in Europe and Asia to predict wind and solar output much more accurately. Turbine downtime is being decreased by predictive maintenance tools. In order to reduce transmission losses, smart grids are rerouting electricity in real time. Digital dashboards that monitor weather fronts and consumption spikes, along with algorithms that modify flows before operators finish their coffee, illuminate control rooms in Germany. The transition from reactive to anticipatory control is difficult to overlook.
However, there are risks associated with this change. Uncomfortable sovereignty concerns are brought up by the concentration of cloud infrastructure in a small number of multinational corporations. Who really has leverage if energy grids are optimized around AI facilities owned by foreign corporations? Governments are already promoting domestic chip manufacturing and strengthening export restrictions on sophisticated semiconductors. Infrastructure related to energy, data, and computation is combining into a single strategic ecosystem.
The sustainability paradox is another. Although AI promises efficiency in areas like agricultural optimization, industrial waste reduction, and extreme weather forecasting, the models themselves require a significant amount of water and energy for cooling. Hyperscale data centers are vying with agriculture for water resources in areas that are vulnerable to drought. Grid managers are concerned about localized price spikes and saturation. If left unchecked, the technology intended to speed up decarbonization might actually make it more difficult.
Hydrogen is also becoming a topic of discussion, especially for remote or off-grid data centers. According to executives, hydrogen fuel cells are robust and adaptable, enabling them to power buildings in areas where demand exceeds grid capacity. Although it is still unclear if hydrogen can scale profitably, governments are increasingly investing in pilot projects in an effort to ensure optionality in a changing energy environment.
At the same time, finance is changing. To support grid modernization, nuclear upgrades, and renewable expansion, Deutsche Bank and other major lenders are extending their transition finance frameworks. Approximately $2 trillion was invested in energy transition last year. However, experts predict that in order to keep up with electrification and growth driven by AI, annual transmission spending must surpass $200 billion worldwide. The alternative—grid bottlenecks impeding economic competitiveness—may be worse than the startling statistics.
The speed of this moment is what distinguishes it from previous energy shifts. Pipelines took generations to build, and oil fields took decades. Data centers may expand in a matter of months. Every week, AI models are updated. In contrast, grid expansion is still controversial and slow. Tension is being caused by this mismatch. Slow infrastructure improvements may end up being the main barrier to digital growth, the International Energy Agency has warned.
There is a sense that geopolitics is being subtly altered as a result of the convergence of AI and energy policy. Not through grand summits, but through export regulations for semiconductors, transmission permits, and procurement contracts. Once taken for granted, electricity is now the foundation of national aspirations.
Whether the world will handle this shift cooperatively or split into rival techno-energy blocs is still up in the air. However, one thing is clear: energy politics is more than just fuel in the era of artificial intelligence. It’s about power in its most literal sense, channeled through grids, measured in megawatts, and increasingly transformed into intelligence.
