10/13/2025
In the debate over the energy infrastructure required for the rapid growth of AI, small (nuclear) modular reactors (SMEs)are being touted in some quarters as a reliable, dense and zero-carbon way to supply data centers and critical networks.
A seductive idea for industries looking to justify colossal investments, it also demands rigorous and critical scrutiny. After all, a nuclear reactor is a nuclear reactor, with inherent dangers, and multiplying the number of installations also multiplies the risk vectors. An SMR on a truck: what could possibly go wrong?
The promise of “safer”, “modular”, “quick to deploy” and “low carbon” energy doesn’t hold up in the face of history, economics, or risk analysis. Modular designs have been explored before, and they faced the same obstacles: uncontrollable costs, complex engineering, difficulties in scaling and operational problems. The simple truth is that small nuclear reactors can’t compete with renewables today. Instead, the arguments are based on political, financial and institutional motivations, fueled by a mentally ill person who hates renewables.
First, the technical and operational risks. A reactor, large or small, is based on fissile materials; radioactive, very hot and requiring complex cooling systems. There are no shortcuts: the safety requirements are high and the margins for error are minimal. In the case of tens or hundreds of units, points of failure, human error, wear and tear, cybersecurity challenges, fuel transportation, institutional coordination and regulatory oversight increase exponentially. Studies already warn that SMRs are no safer than large reactors, precisely because each unit must replicate its own chain of containment, regulation, maintenance and operation.
Next, there’s the domino effect of multiple small reactors scattered across a large range of locations. More installations means more interfaces, more transportation, greater exposure to unexpected events such as earthquakes, floods, sabotage or cyberattacks. Each mini-reactor must replicate the complex safety framework: design, operation, inspection, maintenance, regulatory control. If “sh*t happens” even in large installations, and we have good proof of this, there is absolutely no guarantee that it will be avoided when the network multiplies. Modular reactors introduce challenges of inter-module dependency, increased interfaces, interrelated events and difficulties in modelling risks between multiple units.
Then there is the problem of waste and the fuel cycle. Promises of “waste-burning reactors” or “minimal waste designs” are recurring claims in nuclear marketing, but the harsh reality remains: even the most advanced prototypes generate waste that is difficult to manage. SMRs, due to their geometry and surface density, can generate more activated waste per unit of energy than larger reactors. And the storage, transportation, monitoring, and final disposal of that waste remains a gigantic challenge: expensive infrastructure, long lead times, scarce geological decisions, and social conflict. Multiplying the number of installations multiplies this challenge and the likelihood of management failures, institutional negligence or regulatory weaknesses.
But the most powerful argument is economic: producing energy through reactors, even small ones, is neither cheap nor simple. Promises of “cheap and fast” reactors turn into downgrading, cost overruns, construction delays, safety costs, decommissioning costs and optimistic unfulfilled forecasts. Critical reports stress that SMRs are still “too expensive, too slow and too risky” to play a relevant role in the energy transition. This article in CleanTechnica, entitled “The nuclear fallacy”, explains the structural constraints of SMRs: they lack economies of scale, sacrifice thermal efficiency, do not adapt well to remote sites and continue to face high safety and liability costs. And even more so when compared to increasingly mature and cheaper renewable technologies.
Let’s be clear: the buzz around SMRs in the United States is being driven by the promise of state subsidies, support from the Department of Energy, the nuclear lobby, local policies that seek to “save” declining nuclear cities, and the convergence of venture capitals and “futuristic narratives” that tend to promote the new even if their technical underpinnings are weak.
Compared to renewables, the difference is clear: levelized costs (LCOEs) for solar wind and storage have fallen dramatically. Integrating batteries, pumping, hydrogen or smart grids makes it possible to design systems with 100% renewable energy that are competitive or even cheaper than mixed schemes with nuclear. Recent studies indicate that a 100% renewable grid is possible by 2050 or earlier if massive rollout, redundancy, oversizing, efficiency and flexibility are prioritized. Some analyses indicate that an optimal share of nuclear in European decarbonized systems could be limited to 10% due to its fixed costs and minimal operational flexibility. Other models show that when decommissioning nuclear power plants, the renewable transition does not necessarily imply higher costs, and that focusing all resources on renewables accelerates the fall in emissions and deploys more robust capacities.
But perhaps most significantly, the nuclear industry and mini-reactor promoters have very powerful interests in peddling this narrative. A burgeoning market for small reactors generates contracts, licenses, spare parts, fuel, maintenance, regulation, and institutional dependence on a large-cap business. Behind the institutional push toward SMRs is the persistence of subsidies, the nuclear budget in government agencies, local politics that try to preserve declining nuclear economies, and the lure of futuristic technological discourse that seduces venture capital. Many of the supposed “advantages” of mini-reactors that are regularly trotted out, are, in reality, carefully constructed narratives rather than realities endorsed by results.
Yes, AI data centers demand colossal energy, and we cannot ignore that pressure, but this Marx Brothers “more wood, more wood” dynamic is not only unsustainable, it probably won’t solve the problem. But above all, it is not a reason to unquestioningly accept the reinstallation of nuclear technology in 50 localized versions. We can meet the demand with renewables such as solar, wind, storage, etc. combined with efficiency, demand management, smart grids, flexibility and sectoral coupling (heat, transport, industry). It is not a question of imposing SMRs as a one-shop solution, but of redesigning the energy pattern towards a distributed, resilient model that meets everybody’s needs.
SMRs are being presented as a “quick fix”, but that doesn’t turn the risks into advantages: they distribute the danger, multiply the waste, do not guarantee any cost reductions and are based on the logic of industrial dependency: if you don’t believe it, click on the links. In the 21st century, we must choose whether we want to build intelligence on a network of complicated cuckoo atomic clocks, or on a renewable foundation that is safe, flexible and above all, backed by science.
In the debate over the energy infrastructure required for the rapid growth of AI, small (nuclear) modular reactors (SMEs)are being touted…
