AI and energy: the data-centre power panic has the economics backwards

The story of the moment is that AI is about to double data-centre electricity demand and break the grid. The IEA's Energy & AI outlook puts real numbers on it: data centres consumed about 415 TWh in 2024 (~1.5% of global electricity) and reach roughly 945 TWh by 2030 in the base case — a 2.3x rise, around 15% a year, with AI-accelerated servers responsible for about half of the net increase.

But the same report contains the fact the headline drops: data centres are less than 10% of the total growth in electricity demand between 2024 and 2030. The rest is electrification, industry, cooling and EVs. AI is not what tightens the grid — it is a visible slice of a much larger shift.

For an energy, O&G or heavy-industry operator, the economically interesting lever is on the other side of the meter. Turbines, refineries and compressors already generate more data than teams exploit; AI closes that gap without a single new gigawatt. The rare resource has never been the megawatt — it is the ability to turn energy data into decisions faster than a competitor.

The starting point is not a model. It is a map: the energy data you already hold, crossed with the decisions that actually move margin and uptime. Everything downstream — where compute goes, which use case pays first — follows from that map.