Who controls the watts controls the future

Posted on: 16 January 2026

One of the fundamental rules of systems analysis is to observe what actors do, not what they say. What they say is narrative. What they do is strategy. The distance between the two reveals the real structure of the game.

Take the energy transition.

The narrative is familiar: the future is electric, distributed, clean. Solar panels on rooftops, energy communities, democratisation of production. A coherent, attractive story, repeated everywhere.

What is actually happening is different.

American data centres consumed 183 terawatt-hours in 2024. The projection for 2030 is 426 terawatt-hours, an increase of 133%. Globally, data centre consumption doubles over the same period. The United States and China account for 80% of this growth. Artificial intelligence, which currently represents between 5 and 15% of data centre consumption, will reach 35-50% by the end of the decade.

That is the demand side. Now the supply side.

The Trump administration has signed an executive order excluding solar and wind from incentives for powering data centres. The permitted sources are natural gas, coal, nuclear, and geothermal, defined as "dispatchable baseload sources". Data centres for artificial intelligence have been designated "critical defence facilities". The stated objective is to quadruple American nuclear capacity from 100 to 400 gigawatts by 2050.

A superficial observer might read this as energy policy. A systems observer recognises a different pattern: deliberate selection of sources that require capital concentration, specialised expertise, centralised infrastructure, controllable supply chains.

Solar and wind have a structural characteristic that distinguishes them: they are distributable. The barrier to entry is low. Anyone with a roof and some capital can become a producer. Power disperses.

Nuclear has the opposite characteristic. It requires investments in the billions, development cycles spanning decades, expertise that exists in few places, complex state authorisations. It is centralisable by nature. Power concentrates.

This is not a conspiracy. It is structural selection. If you must choose which sources to incentivise, and you systematically choose those requiring concentration, the result is concentration. Regardless of stated intentions.

The second level of the mechanism concerns distribution.

In most American states, electric utilities operate as regulated regional monopolies. Only 13 states out of 50 have programmes allowing generators to sell electricity directly to end consumers. In the other 37, every transaction passes through the local utility.

In 2025, the value of acquisitions in the American energy sector was 141.9 billion dollars, compared to 28 billion in 2024. A factor of five in twelve months. Constellation acquired Calpine for 29 billion. NRG acquired LS Power assets for 12.5 billion. The documentation for these deals explicitly indicates as motivation "securing dispatchable generation to meet data centre demand".

The pattern is accelerated consolidation in generation, on a substrate of existing monopolies in distribution.

The third level concerns industrial-scale consumers.

Google acquired Intersect Power for 4.75 billion dollars, gaining the capacity to develop generation and data centres in an integrated manner. Amazon invests directly in nuclear projects. Microsoft funded the reopening of the Three Mile Island plant to power its data centres.

These are not energy purchases. They are purchases of position in the value chain. The logic is linear: whoever depends on others for a critical input is vulnerable. Whoever controls the input is protected.

The emerging structure has three layers.

First layer: competition among energy incumbents to reposition. TotalEnergies has 19 gigawatts of renewable capacity and aims for 120 terawatt-hours of electricity production by 2030. Shell has 4. BP and the American majors remain focused on fossil fuels. There is no coordination: there is competition to determine which asset configuration survives the transition.

Second layer: consolidation among utilities and independent generators. Whoever controls generation and distribution controls the flow. Existing regional monopolies strengthen themselves through vertical and horizontal acquisitions.

Third layer: vertical integration by large consumers. Tech giants are internalising energy production because they have identified external dependence as a strategic risk.

The result is a funnel structure. Many end consumers, few flow controllers, very few controllers of generation at scale.

There is one final element to consider: the distribution of costs.

Data centres in the PJM electricity market, covering an area from Illinois to North Carolina, have already contributed an estimated 9.3 billion dollars in increased capacity costs for 2025-2026. The impact on residential bills is in the order of 15-20 dollars monthly in some areas. Projections indicate average increases of 8% by 2030, with peaks of 25% in areas of highest concentration.

The mechanism is structural: large industrial consumers sign long-term contracts at stabilised prices. Increased demand raises marginal prices. Network infrastructure investments are recovered through regulated tariffs, distributed across all users. Benefits are concentrated. Costs are distributed.

This is not a flaw in the system. It is the system.

A final note on narrative.

In specialised data centre circles, the reference metric is changing. The primary discussion is no longer energy efficiency, measured in PUE. It is "tokens per watt per dollar": how much computational capacity, translated into language model output, is obtained per unit of energy and cost.

This is a significant frame shift. Energy is no longer a cost to minimise. It is a strategic input. Whoever controls more of it, processes more. Whoever processes more, trains more powerful models. Whoever has more powerful models, captures more value.

The energy transition is happening. Only the direction is not towards distribution. It is towards a new form of concentration, with updated packaging.

Who wins and who loses in this structure is not determined by malign or benign intentions. It is determined by position in the chain: who controls generation at scale, who controls distribution, who controls industrial consumption. Everyone else pays the bill.

This is not conspiracy. It is architecture.