Ever increasing demand for energy, the “Age of Electrification”, and now data centres
We have always anticipated an increase in global energy demand, driven primarily by population growth, rapid industrialization, and the expanding electrification of our economies. Energy does more than just providing a comfortable life; It powers healthcare, communications, water and wastewater treatment, financial services and banking to name a few and to include many necessities that we take for granted. Energy abundance is crucial for a country’s security and is deeply tied to a nation's economic growth and stability.
Although we had always expected a rise in the energy demand, some might say we have entered an “Age of Electrification”. One might also argue that in our pursuit of emission-free and eco-friendly practices, we have shifted the heavy lifting to our existing power grids: Electrical Vehicles (EVs) and Electrical Buses for public transport, electrical cooktops, electrical furnaces in heavy industries and boilers to name a few utilities.
In addition to these utilities, which our existing power grids manage, we have also injected a major beast to the mix – modern day computing, or neural network computing, or artificial intelligence, which come in the form of power-hungry data centres.
This brings us to a crucial question: “Is our existing power grid ready for increasing modern-day computing demands? And what happens if a nation’s power grid collapses due to an excess in energy demand?”
To prevent our grids from collapsing, we need more energy, or a source that provides a high baseload power, complemented by sources we can turn on and off at will. Conventional thermal power plants burning coal and natural gas are reliable, but they come with a major catch: CO2 emissions that worsen the greenhouse effect, drive global warming, and cause severe air pollution. Conversely, renewables like wind and solar can support the grid, but they are intermittent and cannot deliver high, consistent baseload energy. This leaves us with a clear objective: we need a high baseload energy source that produces immense power without releasing harmful emissions into the environment.
Enter nuclear.
Nuclear Energy: the unsuspecting saviour of modern-day computing
Nuclear energy is not new. In the conventional sense, it has been around since the 1960’s and has gained sufficient operational experience (approximately 60 years) and technological maturity. Nuclear energy, obtained by splitting the nucleus of an atom (specifically Uranium-235), has a very high energy density and does not release any greenhouse gas emissions in doing so. Unlike solar and wind energy, it is not dependent on the weather and provides continuous energy on demand.
Now, the answer’s clear – we have a source of power that can provide sufficient power to our grid without polluting the environment. However, it does comes with its own challenges. First, is the obvious radioactive spent fuel that is left behind after a nuclear power plant is operated. However, one might argue that the volume of radioactive waste produced by nuclear plants is significantly lower than the pollutants released by thermal power plants. In fact, studies show the total spent nuclear fuel generated globally since the dawn of commercial nuclear power, powering millions of homes and entire cities for decades, could fit onto a single football field. Considering the massive amount of energy generated by the roughly 400 nuclear power plants operating worldwide, that is a remarkably small footprint. So, this is a bargain where the merits of operating nuclear power plants far overweigh its demerits. Then, the next obvious question is, is it safe?
Nuclear the ultimate green and safe source of energy
Because nuclear power produces no operational emissions, it is truly clean energy. The real-world consequences of abandoning it were made clear when Germany, largely due to political pressure, shut down its operating nuclear power plants. The resulting grid shock forced a reliance on fossil fuels, leading to air pollution which led to thousands of premature deaths. This reality positions nuclear power as the ultimate "green flex" of the modern era.
Another thing we should address is the bad rep that nuclear has received due to accidents such as Fukushima, Three-Mile Island, and Chernobyl. Albeit these were catastrophic accidents, these events have shaped the entire nuclear industry and has birthed something unique to the line of work of nuclear professionals – “The Nuclear Safety Culture”.
The philosophy of nuclear safety culture at its core is simple - everyone working for a nuclear power plant is responsible for safety and personnel are free (and also encouraged) to question any practices that they feel compromises nuclear safety.
Additionally, nuclear power plants go through a lot of scrutiny by nuclear regulators (each country has one), who act as watchdogs and scrutinize everything, ensuring nuclear operators maintain the highest possible safety standards.
Now, we have addressed three main concerns with nuclear energy: radioactive waste, safety, and technological maturity, then why are we not making new power plants to power our data centres?
The answer is simple: it takes a lot of time and a fair amount of space to build a nuclear power plant.
So how do we get around this?
SMRs: The pocket-sized revolution of nuclear energy
Conventional large-scale nuclear power plants take 10 to 15 years to build and are notoriously known for high capital costs, budget overruns, and construction delays. The solution lies in standardization and modularization: introducing Small Modular Reactors (SMRs).
The SMR approach shifts nuclear construction away from bespoke, complex mega-projects toward simple, repeatable installations. Where traditional nuclear projects demand massive upfront investments (often exceeding €10 billion per reactor) SMRs are projected to require significantly less capital, making them far easier to finance.
SMR components are planned to be factory-manufactured and shipped directly to the site for assembly. This ensures that SMRs will be faster and cheaper to build, significantly reducing upfront costs. Owing to their compact footprint (around 50 acres), SMRs can be installed in locations unable to support large reactors, powering smaller electrical grids, isolated areas, or sites with limited water access. They are also highly scalable, allowing capacity to be added as demand grows.
This is where the intersection with modern computing becomes critical. Data centres are facing an urgent energy crunch: AI-focused facilities now require up to 80 MW of power, i.e., more than double the 32 MW consumed by standard data centres. In the U.S. alone, data centre power demand is projected to jump from 17 GW in 2022 to 35 GW by 2030, all while grid interconnection delays stretch up to a decade. These bottlenecks have driven major tech companies such as Google, Amazon, Meta, NVIDIA to turn to SMRs for reliable, 24/7 carbon-free power.
Producing up to 300 MW of uninterrupted power per unit, SMRs offer the reliable, off-grid energy needed for high-performance computing. But despite this promise, one final hurdle remains: Is the modern supply chain actually ready for SMR deployment at scale?
Is the present nuclear supply chain ready for the nuclear expansion fuelled by modern-day computing?
Although SMRs operate on the same principle of nuclear fission as conventional large-scale reactors, shrinking their size brings in its own challenges. In addition to complex reactor physics and regulatory barriers, the main question that needs addressing is: “Does the global market have the materials required to initially launch SMRs before we can scale up the production?”
A major challenge we must address here is lack of technical maturity of SMR markets. In contrast to conventional light- and heavy-water cooled reactors, there are hundreds of SMR designs currently in development, each of them using different reactor configurations and cooling methods. This highlights a critical absence of standardization for which our current supply chain is unprepared for. Not to mention that with numerous SMR designs in development, licensing has become a nightmare because regulators are unfamiliar with new reactor types. Hence, a lack of regulatory acceptance discourages current market players to invest in materials that they are unsure about.
To provide a comparison here, recent studies have shown that procurement issues, design immaturity, and licensing delays caused severe cost overruns in First-Of-A-Kind (FOAK) Gen-III reactors, even for technologies that had been proven and operated for over 50 years. Without a well-developed, fully coordinated supply chain to reduce direct EPC costs and lower uncertainty, SMRs risk falling into the exact same financial traps as their predecessors.
Already the nuclear industry faces heavy manufacturing constraints due to a global scarcity of specialized forging presses required for reactor pressure vessels, alongside critical vulnerabilities in the availability of High-Assay Low-Enriched Uranium (HALEU) fuel required for SMRs. Compounding to these issues, there is a global shortage of specialized nuclear personnel such as nuclear engineers, certified welders, and quality inspectors.
Therefore, although SMRs are envisioned to shorten construction timelines through factory-based manufacturing, and using the concept of modularization, scaling them up requires overcoming severe supply chain bottlenecks and specialized talent pool.
SMRs undoubtedly have the potential deliver clean, reliable power to data centres without straining the existing infrastructure, ensuring grid stability. There is a bit of poetic parallelism here - data (the smallest unit of computing) has unknowingly fuelled an atomic revolution (the smallest unit of matter - well nucleus to be precise). Yet, to sustain this revolution and realize its promise, we must collaboratively as an engineering community address three critical issues: enhancing standardization of SMR design, robust regulatory frameworks, and a resilient global supply chain.
