AtkinsRéalis and Nvidia are collaborating to imagine nuclear-powered data centres that could be co-located with hyperscale AI campuses. The plan envisions on-site nuclear power sources in the range of 740 to 1000 MWe, comparable to small modular reactors.
Nvidia’s computing tools will guide the planning and integration of these plants into existing infrastructure. Proponents argue that co-locating power and data facilities could reshape campus scale and layout.
This could deliver greater resilience, efficiency, and sustainability. There are also opportunities to share electric infrastructure with local grids and recover waste heat.
This concept responds to intensified demand for secure long-term power. It signals a new chapter in how data centres might be powered in the future.
Nuclear-Powered Data Centres: A New Paradigm for Hyperscale Computing
The collaboration underscores a shift in how energy and digital infrastructure intersect on campus-scale projects. By pairing high-capacity on-site reactors with AI-driven planning tools, the project aims to streamline design processes and integration with existing systems.
Book Your Dream Vacation Today
Flights | Hotels | Vacation Rentals | Rental Cars | Experiences
The capacity targets—between 740 MWe and 1000 MWe—mirror the scale of small reactors. This makes the concept both ambitious and potentially transformative for data-centre clusters.
AtkinsRéalis emphasizes that co-location changes the typical footprint and layout. This could yield a more resilient and sustainable energy platform for hyperscale computing.
Design and Campus Implications
Co-locating nuclear power with data centres requires rethinking campus geometry, service corridors, and safety zones. It also prompts new approaches to site selection, modular construction, and lifecycle management that align with digital planning tools and simulation capabilities.
- Campus layout: Integrated power and data corridors may reduce duplication of infrastructure. This requires tighter coordination of utilities and safeguards.
- Shared utilities: Aligning with local grids could optimize peak demand management and resilience.
- Regulatory and safety frameworks: Compliance and risk-management considerations become central to every design decision.
- Delivery models: Digitisation and modular strategies could accelerate build-out timelines for both energy and data assets.
Sustainability, Heat Reuse, and Grid Services
The proposal envisions capturing waste heat through absorption chillers to contribute to cooling or other process needs. This is a key step toward improved operating efficiency.
It also foresees the plant providing services beyond simply powering AI workloads. This could strengthen grid reliability in regions with growing demand for secure power.
Techniques and Opportunities
- Waste-heat recovery: Absorption chillers could convert excess thermal energy into cooling for data-centre campuses or nearby facilities.
- Grid integration: Shared electric infrastructure with local networks may reduce stress on conventional grids during peak AI workloads.
- Resilience and security: A stable, on-site power source could offer protection against grid outages and supply disruptions.
- Emission considerations: AtkinsRéalis argues that total carbon emissions could be lower than traditional data-centre models, even with larger footprints.
Industry Context and Delivery Models
The move aligns with broader interest in co-locating nuclear power and data facilities, including discussions about Rolls-Royce’s SMR developments and earlier coverage in Dezeen.
As data-centre construction accounts for a rising share of commercial building spending, there is a clear push toward digitisation and innovative delivery models. These models aim to scale nuclear power quickly alongside digital infrastructure.
Digital Tools and Industry Impact
- Nvidia’s planning software: Advanced modelling and simulation could streamline the integration of reactors with data-centre systems.
- Delivery acceleration: Digital twins, automation, and modular construction may shorten project timelines for both energy and data assets.
- Policy and procurement: Real-world pilots could influence regulatory pathways and procurement strategies for co-located nuclear tech.
Implications for architects and engineers include rethinking safety zones and cooling strategies. Long-term maintenance within mixed-use campuses is also a key consideration.
Scalable and cost-effective deployment remains essential. As hyperscale builders seek faster and more secure energy, the AtkinsRéalis–Nvidia concept encourages industry dialogue on digital tools and sustainable energy technologies.
Here is the source article for this story: AtkinsRéalis to partner with Nvidia on nuclear-powered data centre design
Book Your Dream Vacation Today
Flights | Hotels | Vacation Rentals | Rental Cars | Experiences