In the last fortnight, stocks in the nuclear industry have seen their steepest decline for five years. Rising real yields have punished assets with a long duration whose value lies years away, and nuclear has been dragged down with the AI power trade it is now lumped in with.

The price of financing the AI build-out tells its own story. Anthropic is paying Blackstone a 10% coupon on a $60 billion chip financing, against 8.5% on a similar deal just four months ago and 7% for comparable credits in the public high-yield market. Lenders want a fatter premium for AI risk, and nuclear has been sold in sympathy. Yet that same deal pays for 1.5 gigawatts of new chips. Somebody has to power them.

Jim Reid at Deutsche Bank recently circulated a note that makes the point better than any argument. Over 80 years, power has become roughly 50 times cheaper. Computer chips got a hundred billion times cheaper in 60 years. AI has fallen some 200,000 times in barely five, the steepest cost curve ever recorded. Intelligence is now the fastest deflating good in history, and it runs on the slowest deflating input.

William Stanley Jevons taught us in 1865 that when a resource gets cheaper we don't use less of it; we find a thousand new uses and consume vastly more. Cheap intelligence won't shrink the power bill, it will explode it. Value migrates to whatever cannot be compressed, and nothing compresses less than firm, reliable power and the fuel that makes it. Whoever owns the electrons owns the toll road to superintelligence.

There is a second property of electricity that the market forgets in a panic. Demand for it is among the most price-inelastic of any good. Households and hospitals don't switch off when the bill rises, and a data centre training a frontier model will pay almost anything to keep running.

The same holds one link further up the chain. Uranium is a small fraction of the cost of running a nuclear reactor, so utilities will pay whatever it takes to keep their plants fuelled – 2006 and 2007 proved it. When the flooding of Cigar Lake knocked out expected supply, uranium rose sevenfold in six months. Consider the world in which that happened. Half of it was set against nuclear power. There was no iPhone in anyone's pocket, no AI and no data centres drawing a city's worth of power. Even then, with demand flat and politics hostile, a single supply shock was enough. When buyers cannot walk away, price is the only thing left to give, and it gave violently. Inelastic markets don't drift. They gap.

The nuclear industry has been hit by “gamified finance”

So why has the market priced the opposite for the nuclear industry? Look at who owns the shares. This rally was built by a new breed of investor. Finance has been gamified, put on a phone and turned into entertainment, with conviction sourced from Substack, StockTwits and X rather than from balance sheets.

The numbers are startling. Retail flows now account for roughly 20% of US equity volumes, and as much as 55% to 65% of daily turnover in South Korea and China. Seven in ten retail investors are under 45. In 2015 just 6% of 25-year-olds held an investment account; by 2024 it was 37%.

US exchange-traded funds (ETFs) took a record $1.5 trillion of net inflows in 2025. In Britain, retail equity and ETF flows reached £228 billion. This is a younger, less wealthy cohort with thin cushions against loss, and it has learned to borrow. Margin accounts, options and leveraged ETFs are now standard kit on trading apps.

Now the brokers are pulling the ladder away. Last week one of Europe's largest retail trading platforms asked for 110% collateral on positions. At that level, margin is no longer leverage at all, it is a demand for more than the position itself is worth. When the cost of carrying a position exceeds the position, holders don't sell because they have changed their minds. They sell because they have no choice. That is precisely what unravelled in 2008 and 2009. The leverage went first and prices followed, whatever sat underneath.

Which brings me to the nuclear industry trades I find most compelling. and are among a group of nuclear names now trading, in some instances, below the cash on their balance sheets: strip out the cash and the market is assigning the business itself a negative value. The technology, the enrichment capacity, the land, the permits, the people and years of regulatory work are being thrown in for less than nothing.

That price makes sense in only two ways. Either the board is about to set fire to the cash, or the seller doesn't care what he is selling. Any serious investor must ask how fast the money is being spent, what commitments sit against it and how much dilution lies ahead.

Here the answer is reassuring. None of these companies is burning cash at a reckless pace. Lightbridge spends $2 million a month, a modest sum against the cash it holds. When names across a whole sector slip below cash in the same fortnight, and none of them is squandering it, the explanation lies with the holders, not the businesses. The holders were forced.

That is the asymmetry. Buy below cash and the downside is cushioned by money already in the bank, while the upside is whatever the business turns out to be worth. In 2009 the market made that offer on housebuilders and miners that went on to multiply. Today it is making it on the companies building the fuel, enrichment and power that the age of cheap intelligence will depend on.

Nothing about nuclear reactors, fuel or enrichment has changed in a fortnight. What we are hearing is margin calls, momentum funds and macro traders crowding towards the same exit. I have heard that sound before. On that day in March 2009, it told me the equities business would be shut by the following day. It turned out to be the best buying opportunity of a generation.

This article was first published in MoneyWeek's magazine. Enjoy exclusive early access to news, opinion and analysis from our team of financial experts with a MoneyWeek subscription.