nuclear-news

The SMR boom will soon go bust

by Ben Kritz, 3 Nov 25, https://www.msn.com/en-ph/technology/general/the-smr-boom-will-soon-go-bust/ar-AA1PJi1U

ONE sign that the excessively hyped concept of small modular reactors (SMRs) is now living on borrowed time is the lack of enthusiasm in the outlook from energy market analysts, whether they are individuals such as Leonard Hyman, William Tilles and Vaclav Smil, or big firms such as JP Morgan and Jones Lang LaSalle. None of them are optimistic that the sector will be productive before the middle of next decade, and the more critical ones are already predicting that it will never be, and that the “SMR bubble” will burst before the end of this one. My frequent readers will already know that I stand firmly with the latter view; basic market logic, in fact, makes any other view impossible.

In a recent commentary for Oil Price.com, one of the rather large number of online energy market news and analysis outlets, Hyman and Tilles predicted that the SMR bubble will burst in 2029. They based this on the reasonable observation that power supply forecasts are typically done on a three- to five-year timeframe. The fleet of SMRs that are currently expected to be in service between 2030 and 2035 simply will not be there, so energy planners will, at a minimum, omit them from the next planning window, and might decide to forget about them entirely. Deals will dry up, investors will dump their stocks or stop putting venture capital into SMR developers, and those developers will find themselves bankrupt.

That is an entirely plausible and perhaps even likely scenario, but the SMR bubble may burst much sooner than that, perhaps even as soon as next year, because of the existence of the other tech bubble, artificial intelligence, or AI, an acronym that in my mind sounds like “as if.” The topic of the AI bubble is an enormous can of worms, too complex to discuss right now, but the basic problem with it that is relevant to the SMR sector is that AI developers need a great deal of energy immediately. It has reached a point where AI-related data centers are described in terms of their energy requirements — in gigawatt increments — rather than their processing capacity. The availability of power determines whether or not a data center can be built; if the power is not already available, it must be within the relatively short time it will take to complete the data center’s construction.

Even if SMRs were readily available, their costs would discourage customers; AI developers are not too concerned with energy costs now, but they will be as their needs to start actually generating a profit become more acute. On a per-unit basis, SMRs are and are likely to always be more expensive than conventional, gigawatt-scale nuclear plants, and for that matter, most other power supply options. Hyman and Tilles estimate that on a per-unit cost basis (e.g., cost per megawatt-hour or gigawatt-hour), SMRs will be about 30 percent higher than the most efficient available gigawatt-scale large nuclear plants. Being smaller, SMRs would — hypothetically, as they do not actually exist yet — certainly cost less up front than large nuclear or conventionally fueled power plants, but their electricity would cost much more in the long run. That might not be an issue in some applications, but it certainly would if SMRs were intended to supply electricity to a national or regional grid.

Some analyses point out that some early adopters of SMRs, that is, customers who have put down money or otherwise promised to order one or more SMR units if and when they become available, may not be particularly price-sensitive; for example, military customers, governments taking responsibility for supplying electricity to remote areas, or some industrial customers. However, they would still be tripped up by the fragmented nature of the SMR sector, which was caused by the “tech bro” mindset of ignoring almost 70 years of experience in nuclear development and trying to reinvent the wheel.

JP Morgan’s 2025 energy report noted that there are only three SMRs in existence, with one additional one under construction; there is one in China, two in Russia, and the one not yet completed is in Argentina. All of them had construction timelines of three to four years, but took 12 years to complete; or in Argentina’s case, 12 years and counting. Argentina’s project has had cost overruns of 700 percent so far, while China and Russia’s projects were 300 percent and 400 percent over budget, respectively.

These are all essentially one-off, first-of-a-kind units, so some of these problems are to be expected, such as regulatory delays, design and manufacturing inefficiencies, and challenges from building supply chains from scratch. These problems would be resolved over time, except that there are literally hundreds of different SMR designs all competing for the same finite, niche-application market.

If the SMR developers listened to the engineers and policymakers who built up nuclear energy sectors that took advantage of economies of scale by standardizing a few designs and distributing the workload, they might get somewhere. That is not happening; potential customers, whether they have power cost concerns or not, are reluctant to jump in because it is not at all certain which SMRs will survive the competition. They might be willing to experiment to see if one design or another actually works — that is why the Chinese and Russian SMRs exist — but the fragmented SMR sector prevents them from trying more than one and making comparisons, at least not in a timely or financially rational manner.

I think the bubble begins to burst this coming year. The timeframe for construction to startup in most SMR pitches is four years. That’s entirely too optimistic, of course, but even if it is taken at face value, once we get a few months into 2026 without any tangible development happening, everyone will catch on that there won’t be any SMRs by 2030, and interest will turn elsewhere. It already is, among the data center sector, as was explained above.

November 6, 2025 - Posted by Christina Macpherson | business and costs, Small Modular Nuclear Reactors