William Theron holds a Masters in Nuclear Engineering as well as an MBA. Being a consummate problem-solver, he realized that the current power infrastructure was going to hold back AI development and devised a solution. He cleared some time in his busy schedule to discuss his project with us.
Mr Theron, please tell us where you are from and where you studied. Also, what drew you into nuclear engineering?
WT: Thank you for the opportunity, Steph. I’m delighted to be part of Libris Ignis and contribute to your mission of advancing knowledge and innovation in sustainable energy solutions.
I’m from South Africa, specifically a small town called Potchefstroom, which is home to North West University. Growing up surrounded by academic minds shaped my curiosity early on. As a child, I was constantly dismantling and reassembling objects to understand their underlying mechanisms – I was fascinated by how combinations of seemingly insignificant components could create something useful and functional.
My other passion from a young age was growing plants and maintaining trees. I developed an appreciation for how the right climate conditions and understanding each plant’s specific requirements led to successful growth and fruit-bearing.
What drew me to nuclear engineering was a natural extension of these interests. I’ve always approached things by understanding their fundamental components, combining them purposefully to achieve specific outcomes, and then allowing those systems to facilitate growth. In my view, nuclear engineering – specifically, reliable clean energy – provides the power, data centers are the soil, and AI represents the fruit of that ecosystem.
There are a lot of misconceptions about nuclear power. In your opinion, are some of them valid?
WT: There are indeed many misconceptions about nuclear power, and I believe this is a nuanced topic that deserves careful consideration.
Some misconceptions about nuclear power are based on outdated information or misunderstandings, while others contain elements of valid concern. For example:
Regarding safety, many people associate nuclear power with disasters like Chernobyl or Fukushima. While these events were serious, they’re not representative of modern nuclear operations. That said, the concerns about potential consequences of accidents are valid – when things go wrong, the impacts can be significant. Modern designs, especially small modular reactors like those we’re developing at Deep Atomic, incorporate passive safety features that address many of these concerns.
On waste management, there’s a misconception that we have no solution for nuclear waste. In reality, we have technical solutions like deep geological repositories, though implementation has been challenging due to political and social factors. The concerns about very long-lived waste requiring management for thousands of years are valid engineering challenges we must address.
Regarding costs, there’s often confusion about the economics of nuclear power. While initial capital costs are high, the long operational lifetime and low fuel costs make it competitive when considering the full lifecycle. The concern about construction delays and cost overruns has been valid historically, though standardized designs aim to address this.
Proliferation is another area with both misconceptions and valid concerns. Commercial nuclear power plants aren’t designed to produce weapons-grade material, but the technology overlap requires robust international safeguards.
What’s important is having fact-based conversations about nuclear energy that acknowledge both its significant benefits for clean energy production and the legitimate challenges that must be addressed through continuous innovation and careful implementation.
How does nuclear power generation compare to, for instance, wind, geothermal, and solar power generation? Or more specifically, why opt for nuclear?
WT: Nuclear power offers distinct advantages when compared to renewable sources like wind, geothermal, and solar, particularly for high-density applications like data centers.
The primary advantage is energy density. Nuclear power produces enormous amounts of energy from a minimal physical footprint. A single MK60 reactor (as described in our documents) can provide 60MW of electrical power plus 60MW of cooling power continuously from a very small land area. To generate equivalent power with solar or wind would require vastly more land – sometimes hundreds of times more area.
Another critical factor is reliability. Nuclear plants operate at capacity factors around 93%, providing consistent baseload power 24/7 regardless of weather conditions, time of day, or season. Wind and solar are inherently intermittent, with capacity factors typically between 25-40%, making them challenging for applications requiring uninterrupted power. Geothermal is reliable but geographically limited to specific locations with suitable geological conditions.
For data centers specifically, nuclear offers unique integration advantages. Our MK60 reactor is designed to provide both electricity and cooling in a single system, utilizing thermal energy that would otherwise be wasted. This direct integration reduces inefficiencies and dramatically improves overall system performance compared to typical grid-connected renewable installations.
Carbon emissions are comparable among these technologies when considering lifecycle assessments. All produce minimal emissions during operation, though manufacturing and construction impacts vary.
Regarding cost, while nuclear has higher upfront capital costs, its long operational lifetime (60+ years) and low fuel costs make it competitive on a levelized cost basis, especially when considering the avoided costs of grid infrastructure upgrades or extensive battery storage systems that renewables might require.
The optimal energy solution ultimately depends on specific requirements, location, and available resources. Nuclear power is particularly compelling for applications demanding high power density, reliability, and long-term cost stability – precisely what modern data centers need.
AI currently dominates a lot of global conversations, and it certainly seems like it is here to stay. What are your thoughts about the industry? Where do you think the technology will go in five or ten years?
WT: AI has rapidly transformed from a specialized research field to a mainstream technology reshaping industries across the globe. Looking at where we are today and where the technology might head in the coming years, several trends stand out.
The exponential growth in AI capabilities we’ve witnessed recently is directly tied to computational resources. As we develop more powerful AI models, the demand for computation continues to surge, creating a feedback loop driving both AI advancement and data center expansion. This relationship between AI advancement and energy consumption is precisely what we’re addressing at Deep Atomic.
In the next five years, I believe we’ll see significant progress in AI efficiency through algorithmic improvements and specialized hardware. We’ll also witness broader industrial integration, with AI becoming embedded in critical infrastructure, manufacturing, and logistics. The development of more capable multimodal models that can process and generate across different types of data will become standard.
Looking ten years ahead is more speculative, but I expect we’ll see truly domain-specific expert systems rivalling human specialists in fields like medicine, engineering, and scientific research. AI’s role in accelerating scientific discovery will likely become pivotal, potentially leading to breakthroughs in materials science, drug development, and clean energy technologies.
One of the most intriguing possibilities is AI systems becoming effective collaborators in tackling complex global challenges like climate change, where they can help model solutions, optimize resource allocation, and accelerate innovation.
Throughout this evolution, energy requirements will remain a critical constraint. AI’s future will be shaped not just by algorithmic breakthroughs but by our ability to power these systems sustainably and reliably. This intersection of clean energy, computation, and artificial intelligence represents one of the most consequential technological developments of our time.
How does your company, Deep Atomic, fit into the global picture?
WT: Deep Atomic fits into the global picture by addressing a critical intersection of three major trends: the explosive growth of AI and cloud computing, the urgent need for clean energy solutions, and the increasing challenges of data center infrastructure.
Data centers are projected to see $2 trillion in investments over the next five years, with global capacity growing at a 22% CAGR through 2030. AI workloads specifically are driving much of this growth, increasing at an even faster 39% CAGR. This rapid expansion is creating unprecedented energy demands that existing grid infrastructure simply cannot meet.
Major hyperscalers like Google face significant challenges – they’re committed to carbon-free energy but seeing their greenhouse gas emissions increase substantially (67% since 2020). They also face grid connection bottlenecks that can delay new data center projects by 5+ years.
This is where Deep Atomic creates unique value. Our MK60 small modular reactor is purpose-built for data centers, providing 60MW of electrical power and 60MW of cooling from a single integrated system with a footprint under 100m². We’ve engineered it using proven Gen III nuclear technology with established regulatory frameworks and supply chains, enabling relatively fast deployment while maintaining rigorous safety standards.
What makes our approach particularly innovative is the deep integration of power generation and cooling systems. Rather than simply generating electricity, we’ve optimized the entire system to capture and utilize thermal energy that would otherwise be wasted. Our analysis shows this integrated approach can achieve overall plant efficiencies of 45-54% depending on the cooling technology used, substantially outperforming conventional solutions.
In the broader global context, Deep Atomic helps solve multiple challenges simultaneously:
- Enabling sustainable AI growth by providing reliable, clean energy
- Reducing carbon emissions (600k tonnes of CO₂ saved annually per installation)
- Bypassing grid connection bottlenecks (saving ~5 years in deployment time)
- Advancing nuclear technology through practical commercial applications
We’re positioned at the nexus of energy, computation, and climate technology – three domains that will profoundly shape global development in the coming decades.
People tend to think of reactors as enormous structures. Please tell us more about the types of reactors you build.
WT: Our MK60 reactor represents a significant departure from the massive nuclear plants most people envision. Rather than the enormous cooling towers and sprawling facilities of traditional nuclear power plants, we’ve developed a compact, modular system specifically optimized for data center applications.
The MK60 is a small modular reactor (SMR) based on pressurized light water reactor (PLWR) technology. What makes it unique is its size and integration. The entire nuclear steam supply system fits within a modest footprint – the reactor pressure vessel itself is just 2.68 meters in diameter and about 8 meters high, including the head. This compact size means it can be transported on standard highways or railways without special arrangements.
We’ve engineered the MK60 with a thermal output of approximately 200MWth, which translates to 60MW of electrical power plus 60MW of cooling capacity – perfectly sized for modern data center modules. The full containment structure has a footprint < 100m2.
Unlike traditional nuclear plants designed for maximum power output, our reactor is optimized for reliability, manufacturability, and integration with data center cooling systems. We’ve incorporated numerous design choices that enhance this integration, including innovative approaches to use waste heat from the power generation cycle to drive cooling systems like steam absorption chillers.
What’s particularly important is that despite its modest size, the MK60 leverages proven, well-understood nuclear technology with established regulatory frameworks and supply chains. We’re not creating exotic new reactor types requiring decades of regulatory development – we’re taking mature technology and adapting it for a new application while maintaining all safety standards.
The modular nature also allows for scalability. Multiple MK60 units can be deployed together to support larger data centers, with capacities scaling to 1GW or beyond while maintaining the advantages of standardized, factory-built components.
And on a less serious note, and I know you will appreciate this being a father, what is your favourite dinosaur, and why?
WT: Velociraptors. From an engineering perspective, Velociraptors represent elegant design efficiency. They weren’t overbuilt; they were optimized for speed, agility, and cooperative hunting. Their lightweight frames, counterbalanced tails, and forward-facing eyes all contributed to their effectiveness as hunters.
There’s something compelling about how they leveraged teamwork and intelligence rather than sheer size to thrive in their ecosystem. It’s a reminder that sometimes the most effective solutions come from collaboration and smart strategy rather than overwhelming force – a principle that applies well beyond paleontology.
Visit the Deep Atomic website for more information about this exceptional project.
Did you like this article? Please subscribe to our newsletter to receive our latest interviews, articles, and learning materials. If you are a researcher (or know someone doing interesting research) who would like to increase their digital footprint through an interview or article, please contact us.
