Fusion power might not get cheap as fast as you think

Fusion power has been the holy grail of clean energy for decades — essentially limitless, zero-carbon electricity, if we can ever get the reactors to work. But a new paper in Nature Energy throws a bucket of cold water on the hype: even if fusion plants do get built, the power they produce might never be cheap.

Technologies usually get cheaper as we build more of them. Lithium-ion batteries dropped about 90% in cost since 2013. Solar modules have followed a similar curve. But not everything follows that path. Nuclear fission, for example, has barely budged — its cost declines by only about 2% every time global capacity doubles.

The key metric here is something called the experience rate: the percentage drop in cost each time installed capacity doubles. For onshore wind it’s 12%. For lithium-ion batteries it’s 20%. For solar it’s 23%. For fission? A measly 2%.

So where does fusion land? The researchers behind the new study tried to answer that by looking at three factors that historically correlate with experience rate: unit size, design complexity, and how much customization each installation needs. Bigger, more complex, and more customized technologies tend to have lower experience rates.

They interviewed fusion experts — both from public research labs and private companies — and asked them to rate fusion plants on those three dimensions. The results aren’t pretty.

Fusion plants will be large, comparable to coal or fission plants that also rely on generating heat. They’ll probably need less customization than fission plants, since regulations should be simpler (no meltdown risk, no long-lived waste). But they’re still more customized than mass-produced solar panels. And complexity? Almost everyone agreed fusion is off-the-charts complex. Some experts literally couldn’t rate it on the scale the researchers provided.

The final estimate: fusion’s experience rate is somewhere between 2% and 8%. That’s faster than fission but way slower than solar, wind, or batteries. It means electricity from fusion could stay expensive for a very long time, requiring massive deployment — and decades — before costs come down meaningfully.

This matters because a lot of current modeling assumes fusion will follow a much steeper learning curve, with experience rates of 8% to 20%. Those assumptions have been used to justify some serious money. The US government allocated over $1 billion to fusion in fiscal year 2024. Private-sector investment hit $2.2 billion between July 2024 and July 2025.

Lingxi Tang, a PhD candidate at ETH Zurich and one of the study’s authors, puts it bluntly: “If you’re talking about decarbonization of the energy system, is this really the best use of public money?”

That’s a fair question. We’ve seen this pattern before — pouring money into a technology that sounds great on paper but never delivers at scale. But it’s also worth remembering that predictions about future energy costs have a terrible track record.

Egemen Kolemen, a professor at Princeton Plasma Physics Laboratory, points out that in 2000, most analysts thought solar would stay expensive. Then China went all in on manufacturing, production exploded, and prices crashed. “People weren’t exactly wrong then,” he says. “They were just extrapolating what they saw into the future.”

The point is that cost curves aren’t destiny. They depend on regulation, geopolitics, labor costs, and a hundred other variables we can’t predict. And as Kolemen notes: “We haven’t built the thing yet, so we don’t know.”

So fusion might get cheap — but it might not. The honest answer is that nobody knows. The study doesn’t prove fusion is a dead end; it just shows that the rosy assumptions baked into many models are probably too optimistic. If we’re going to bet billions on this technology, we should at least be realistic about the odds.