The Grid That Runs on Wishes

The electrical grid does not store electricity. This is the fact that most people who think they understand the energy transition do not fully reckon with.

Batteries exist, yes. But at the scale of national electricity infrastructure, storage remains nascent, expensive, and insufficient. The grid is a real-time system: every electron generated must be consumed, almost instantly, or the frequency dips, the voltage drops, and in the worst case, sections of the network begin shedding load in a cascade that can take days to reverse.

This is not a new problem. Grid operators have managed real-time supply and demand balancing for a hundred years. What is new is the pace at which the composition of the grid is changing — and the institutions governing it are not keeping up.

Coal and natural gas plants are dispatchable. An operator can call up a gas turbine and tell it to produce more power within minutes. Wind and solar are not dispatchable. They produce electricity when the resource is available, not when demand requires it. This is not a reason not to build them. It is a reason to build the rest of the system — storage, flexible demand, interconnections, fast-responding backup capacity — at the same pace.

We are not doing that.

The US electricity grid is a patchwork of regional markets, utility monopolies, federal agencies, and state regulators that evolved over a century of largely separate development. The Eastern Interconnection, the Western Interconnection, and ERCOT in Texas operate almost independently, connected by only a few high-voltage lines that limit how much power can flow between them during a crisis. The rules that govern how generators are compensated were mostly written for a world of large, centralized thermal plants. They function poorly in a world where the grid is increasingly distributed, variable, and bidirectional.

The software running many of these systems was written in the 1980s and 1990s. Some utilities are still running on SCADA systems so old that the original developers are retired or dead. The cybersecurity implications of this are not hypothetical: the 2021 Oldsmar water treatment hack demonstrated how vulnerable critical infrastructure remains to even relatively unsophisticated attacks.

Meanwhile, the pace of electrification is accelerating. Electric vehicles are adding load. Heat pumps are adding load. Data centers — fueled in part by AI — are adding load at a rate that is catching grid planners by surprise. The consultancy firm Grid Strategies estimated in 2023 that utility load growth forecasts had shifted from near-zero to potentially doubling by 2035. The wires that carry this power, the substations that transform it, the control systems that manage it — none of this was designed for this scenario.

I am not saying the energy transition will fail. I am saying the energy transition is a software and institutions problem as much as it is a hardware problem, and we are paying much more attention to the hardware.

The grid that runs on wishes is the grid that assumes everything will work out because it has to. The grid we actually need is one where we have done the boring, expensive, procedurally complex work of upgrading the institutions as fast as we are upgrading the generation fleet.

That work is not glamorous. There is no IPO for reforming capacity market rules. But without it, the clean energy future is a system that works most of the time and fails catastrophically when it does not.