Nuclear Stocks Fall as AI Power Demand Keeps Growing

Nuclear shares have fallen sharply over the past fortnight, their steepest decline in five years. Higher real yields have weighed on companies whose potential value depends on earnings far in the future, while nuclear stocks have been swept into a broader sell-off in businesses tied to artificial intelligence and data centre power.

That drop sits uneasily beside a basic feature of the AI boom: computing may be getting cheaper, but the electricity needed to run it remains a physical constraint. A recent financing deal shows the cost of building out AI infrastructure. Anthropic is paying Blackstone a 10% coupon on $60 billion of chip financing. A similar deal carried an 8.5% rate four months earlier, while comparable public high-yield credits were at 7%. Lenders are asking for more compensation for AI risk. The deal will also fund 1.5 gigawatts of new chips, which will need power.

Jim Reid of Deutsche Bank has highlighted the widening gap between the cost of computing and the cost of electricity. Over 80 years, power has become roughly 50 times cheaper. Computer chips fell in cost by a factor of 100 billion over 60 years, while AI costs dropped some 200,000 times in barely five years. The implication is striking: the fastest-falling cost in this equation depends on an input that has become cheaper much more slowly.

That dynamic echoes the Jevons paradox, named for William Stanley Jevons, who wrote about it in 1865. When a resource becomes more efficient or cheaper to use, demand can rise as people find new ways to use it. Cheaper AI could therefore increase electricity consumption rather than reduce it. Data centres need power around the clock, and nuclear generation is one source of firm, steady electricity.

Why power demand can hold up

Electricity demand is relatively insensitive to price. Households and hospitals still need it when bills rise. Data centre operators training advanced AI models also have strong reasons to keep their facilities running. That doesn’t mean every power producer benefits equally, or that demand forecasts will unfold as expected. It does help explain why investors are focused on dependable supply.

The same reasoning extends to uranium, which makes up a small share of the cost of operating a nuclear reactor. Utilities have a strong incentive to secure fuel rather than leave plants idle. In 2006 and 2007, flooding at the Cigar Lake mine disrupted expected supply and uranium prices rose sevenfold in six months. At the time, nuclear power faced widespread opposition, smartphones and AI data centres had not arrived, and demand was flat. Even in those conditions, a supply shock produced a dramatic price move.

That episode illustrates the risk in markets where buyers have few easy substitutes: a supply disruption can move prices sharply. It doesn’t establish that the same rise will happen again. The scale of any future price move would depend on supply, demand and how quickly buyers and producers can respond.

Forced selling can obscure company value

Why, then, are some nuclear-related shares trading as if their prospects have deteriorated? One explanation is the changing mix of investors and the use of borrowed money. Retail investing has become more app-based and socially driven, with investors drawing ideas from Substack, StockTwits and X. The figures cited here suggest retail investors account for roughly 20% of US equity trading volume, and 55% to 65% of daily turnover in South Korea and China.

The investor base has also grown younger. Seven in ten retail investors are under 45. In 2015, 6% of 25-year-olds held an investment account; by 2024, the share was 37%. US exchange-traded funds (ETFs) received a record $1.5 trillion of net inflows in 2025, while retail equity and ETF flows in Britain reached £228 billion. Trading apps have also made margin accounts, options and leveraged ETFs familiar tools for some investors.

Borrowing can magnify gains, but it can also force investors to sell when prices fall or collateral requirements rise. A large European retail platform recently asked clients for 110% collateral on positions. At that level, a holder may need to provide more than the position’s value. Selling can then happen because the investor needs to meet the requirement, regardless of whether the underlying business has changed. When many holders face pressure at once, prices can fall quickly.

Lightbridge Corporation, ASP Isotopes and Fermi America were cited as nuclear-related companies trading, in some cases, below the cash on their balance sheets. The calculation depends on the figures and liabilities used, and a cash balance alone doesn’t establish that a company is undervalued. Investors still need to assess spending, future commitments and the possibility of issuing more shares. Lightbridge’s reported cash spending is $2 million a month, described as modest relative to its cash holdings.

What the gap between price and cash can mean

A share price below a company’s cash balance may reflect doubts about how that cash will be spent, future financing needs, or the value of the business itself. It can also reflect forced selling that has little to do with a fresh assessment of the company. The distinction matters: cash offers no guarantee against losses if operating costs rise, projects fail or further funding is needed.

The case for looking closely at nuclear companies rests on the possible mismatch between near-term selling pressure and longer-term demand for reliable power and fuel. Comparisons with the market conditions of 2009 offer a reminder that distressed prices can precede strong recoveries, but they don’t predict one. For households investing in individual companies, this remains a high-risk area where a convincing demand story must be weighed against financing, execution and price volatility.

The clearest takeaway is to separate the power-demand thesis from the condition of any one company. A fortnight of falling share prices doesn’t change how reactors operate or how much electricity data centres use. It does make cash, spending plans and funding needs essential parts of any assessment.