Energy Power Plays: Who Owns the Stack?

An industrial company on an island in the Baltic Sea wants to expand and cannot. It will get its connection once E.ON, the local grid operator, builds a new transformer station, and until then the expansion waits, perhaps for another three years. A founder who builds behind-the-meter batteries told the story at the Ripple on energy power plays to make a point the table kept returning to. The island does not lack energy. It lacks power at the peak.

The hosts had opened with the scale of demand. Dinesh Ganesan of ArcTern Ventures cited the IEA's expectation that by 2030 US data centres will use more electricity than the production of all US energy-intensive goods combined, from steel and aluminium to cement and chemicals. Ethan Sohn of Toyota's venture arm urged caution. In 2020 forecasters expected 60 to 100 gigawatts of extra demand from electric vehicles by 2030, and it did not arrive. The gap between low and high forecasts of extra demand from AI data centres to 2028 is now about 250 gigawatts on its own.

Where it lands differs by continent. A participant who specialises in Europe said data centres will add perhaps 100–150 terawatt-hours there by 2030, much like each of industrial electrification, electric vehicles and heat pumps. That makes it a local problem, in Dublin or Amsterdam. In the US, another participant said, the interconnection queue stood at 2,200 gigawatts at the end of 2024, with a median of five years from request to commercial operation.

Grid maxing

The battery founder called the confusion between kilowatt-hours and kilowatts one of the greatest in the energy business. People who come from finance see bills in kilowatt-hours and assume that is what matters. "It's power. It's kilowatts," the founder said. Smoothed out, the energy may already be there. The cost of the wait falls on the municipality, on the company and, if the investment moves to another country, on Europe. The founder's answer was to get the most out of the grid that already exists. They called it grid maxing.

A consultant who has built data-centre power infrastructure in the US and across Europe agreed. The problem on both continents is peak power for a handful of hours, or a few dozen, when substations and transmission lines are not reinforced enough. Compute flexibility, which turns loads up and down, and batteries are the cheapest fixes. The consultant added that cost allocation matters too. Grids serve whoever can pay, when households perhaps ought to come first, and Denmark, the consultant believed, had recently moved data centres down its interconnection queue because of their effect on people.

When one participant suggested separate grids for different demand patterns, such as seasonal heating and flat data-centre loads, another rejected the idea firmly. Grids are cheap to run near full utilisation and very expensive at low utilisation, so separate grids would be the slowest and costliest route, and investing in them a fast way to go bankrupt. The proposer then suggested virtual grids instead. Those exist already, came the reply. Spot prices across Europe change every 15 minutes, and under interruptible contracts a data centre accepts about 60 hours of curtailment a year, brings its own battery and is connected faster.

Who gets squeezed

Behind the physics sat a question of fairness. One participant said AI data centres will be allowed only if people who cannot pay as much still get their energy, which is why their company exists. A founder making fertiliser from electricity feared exactly that squeeze. AI companies can pay a lot for power, and prices rising across the board would hurt industrial electrification.

An infrastructure credit investor drew a lesson from European batteries. Many battery platforms were financed with very high levels of debt on hopes of arbitrage. When power prices turned low or negative, the returns did not come, and the assets are being refinanced. Investors in data centres should keep expectations conservative, the investor said, and take a tenant's credit strength with a grain of salt, because tenants turn over.

The hosts then asked where venture capital fits. "Your fund size is your strategy," one investor said, and a $200m fund cannot finance a half-billion-dollar project. Others saw a narrow opening at the first project, which European infrastructure investors will not fund. VCs dislike their capital sitting in assets, and a first project vehicle takes six to twelve months to set up, so many founders have funded early projects with their own equity. The window exists, one investor said, because the technology is de-risked while infrastructure funds are not yet backing developers, and solar and wind developers have seen a great deal of M&A. A participant warned that utilities move at the speed of population growth, and that regulation scares many VCs off.

Near the end, a participant said the conversation had made them sad. Europe was an industrial power built on energy, they said, and the table was admitting defeat before trying to compete. The reply was that the obstacles are policy and regulation, with incumbents resisting more interconnection and more price areas, and not technology. And Sweden, someone recalled, built out nuclear power after the first oil crisis and has been riding on it ever since.

This Ripple was hosted by Dinesh Ganesan (ArcTern Ventures) and Ethan Sohn (Toyota) at The Drop 2026 on 16 September.