Elon Musk’s claim that “the future currency will essentially just be wattage” is easily read as another Silicon Valley provocation. The infrastructure market is treating it as a scheduling problem. Gas-turbine makers are taking delivery reservations into 2031.

American power projects sit in grid queues for more than five years on average. In Norway, where hydroelectricity has long been shorthand for industrial advantage, data centres already consume more than 3 TWh a year and are seeking far more capacity than the system can readily absorb. Musk made the remark in a Moonshots interview published in January, when asked what would matter in an economy built around artificial intelligence.

He described wattage not simply as electricity, but as power turned into useful work — “intelligence or matter manipulation.” He also called electricity generation AI’s limiting factor for at least the next two years. That is a sharper proposition than the social-media gloss that followed it. Musk did not, in the available transcript, say that money would disappear or describe a self-sustaining loop of robots, chips and solar power.

But his narrower point lands because the race to build AI is colliding with assets that cannot be conjured by a new funding round: substations, transmission lines, transformers, turbines and generation capacity.

The contest for AI capacity is becoming a contest for the equipment and permits that turn capital into electricity. The International Energy Agency expects global electricity demand to grow by 3.6% a year between 2026 and 2030, adding roughly 1,100 TWh annually. Data centres are central to that acceleration.

The agency projects their electricity use will rise from about 485 TWh in 2025 to roughly 950 TWh in 2030; in the United States, they account for about half of expected demand growth through the end of the decade.

Hardware is waiting for power

The bottleneck is now visible in the orders placed by the companies that make the machinery of power generation. GE Vernova said in July that its gas-power equipment backlog and slot reservations had reached 116 GW, including 53 GW of firm backlog and 63 GW of reservations. The company expected at least 125 GW under contract by year-end, while accepting slots that stretch to 2031.

Siemens Energy reported a 69 GW gas-turbine backlog in August, after booking 15 GW of new orders in the preceding quarter. Its executives said lead times were running to three years or more. The strain is not confined to generation: Siemens Energy’s grid backlog stood at €51bn.

America’s interconnection system tells the same story from the other end. Berkeley Lab counted about 1,312 GW of generation and 749 GW of storage in US queues at the end of 2025. The median project took more than five years from request to operation, and only 13% of projects seeking connection between 2000 and 2020 were ultimately built.

That does not make gas the inevitable answer. China added 315 GW of solar capacity in 2025 on the National Energy Administration’s measure, while Europe added 67.2 GW and India 45.7 GW. Yet solar farms, batteries and wind parks still need wires, inverters and firm capacity when the system is short of sunlight or wind.

The IEA estimates that 15 GW to 27 GW of on-site gas generation could be serving data centres by 2030, mostly in the US. The result is an uncomfortable split between the rhetoric of clean, abundant AI and the near-term reality of diesel generators, gas turbines and emergency power plants. It also exposes a weakness in the idea of wattage as a universal currency: a megawatt in the wrong place, or at the wrong hour, is not interchangeable with one beside a data centre.

Norway’s surplus meets a queue

Norway has more electricity per person than almost any European economy and recorded a power surplus of more than 22 TWh in 2025. That has made it attractive to data-centre developers seeking renewable power, cooler air and a stable political setting. It has also made grid access a political question.

Norwegian data centres used an estimated 3.14 TWh to 3.3 TWh in 2025, roughly twice the level of two years earlier. The Norwegian Water Resources and Energy Directorate projects around 8 TWh of demand by 2030. Statnett had reserved 3.5 GW for data centres by March 2026; a parliamentary proposal later put the wider mix of reserved and queued projects at about 12 GW across roughly 120 schemes.

That pipeline is not a forecast of actual consumption. Many projects will fail, shrink or wait. But at 70% utilisation, 12 GW would imply about 70 TWh a year — more than half of Norway’s present electricity consumption.