Explore our data center in Glomfjord, Norway, where innovation, design, and renewable energy are transforming how compute is built and powered.
When you first arrive in Glomfjord, Norway, it’s impossible not to be struck by the stark natural beauty of the fjords and mountains. At first glance it’s an unlikely place for an advanced AI data center. Yet it’s precisely this pristine environment that makes Glomfjord the perfect backdrop for building next-generation AI infrastructure.
From the very beginning, our goal at Nscale has been clear: deliver one of the most efficient and sustainable data centers in the world, while preserving the natural environment that surrounds it. Building and operating in such a remarkable landscape comes with an inherent responsibility: not to compromise what makes it special.
The Glomfjord facility redefines what a sustainable AI factory can be. It’s powered entirely (100%) by renewable hydroelectric energy and designed to operate at the highest levels of performance with minimal environmental impact. But sustainability here goes beyond the source of our electricity. It’s about creating platforms that power AI efficiently, forming lasting partnerships with local communities, and driving innovation across the AI value chain.
At Glomfjord, we’ve focused on solving three of the biggest environmental challenges for AI infrastructure: energy consumption, waste heat, and generator emissions.
Driving efficient power consumption
Data centers are inherently power-hungry. Every model trained, every dataset processed, and every neural network deployed demands immense computational effort. This creates an ever-growing demand for electricity that can strain local utilities and challenge sustainability goals.
We chose Glomfjord as a prime location because it’s a region with abundant renewable hydroelectric power, allowing us to build an energy profile that’s both sustainable and scalable.
But access to clean power is only part of the equation. The real challenge lies in using that energy as efficiently as possible. Thanks to the region’s cold climate, the facility requires far less energy for cooling, which boosts overall efficiency. This performance is tracked through a mix of metrics that monitor power consumption and operational effectiveness.
Most traditional data centers use Power Usage Effectiveness (PUE), which is a simple calculation to determine how efficiently a data center uses energy. It is calculated by dividing the total energy used by the facility by the energy used by the equipment. Typical data centers a decade ago averaged between 1.7 and 2.0. In today’s market, even achieving 1.3 is considered inefficient. Our projections target Glomfjord operating at around 1.1.
While PUE measures the efficiency of the data center itself, it doesn’t tell the full story.
Operational efficiency also depends on how the facility is managed and that’s where standard operating procedures (SOPs) make the difference. Detailed SOPs can be developed to ensure energy efficiency across every system. For example:
- Adaptive cooling systems throttle fan speeds and power output automatically when full cooling capacity is not needed.
- Smart server management dynamically reduces power to servers during low-load periods without affecting performance.
- Optimized cooling loops like rear door coolers that adjust automatically, operating only as hard as required to maintain target temperatures.
These operational controls allow us to deliver the same performance output with significantly less power, in some cases up to 60% lower energy use while maintaining 100% uptime.
By managing both the data center and the infrastructure we can optimize every layer to achieve efficiencies that go far beyond what PUE alone can measure. Our ultimate goal is to build an AI platform that moves us closer to zero environmental impact without compromising performance.
Repurposing waste heat
If power is the input problem, heat is the output challenge. Every watt of electricity consumed by a data center becomes waste heat that must be removed to keep systems operational. Traditionally, this heat is simply vented away, which causes an energy loss that also increases cooling demands.
At Glomfjord, we saw an opportunity to turn that waste into value. We use a closed-loop water cooling system that begins with cold water drawn from the nearby fjord at just 6 to 9°C. This water absorbs heat from the data center and exits at around 34°C, a perfect temperature for local aquaculture.
Rather than discarding that heat, we channel it directly to a nearby fish farm, where it is used to warm aquatic tanks. This simple but effective collaboration saves the farm from expending its own energy on heating and transforms a byproduct of our operations into a valuable local resource.
We aim to have around 85% of the waste heat generated by the data center repurposed in this way, making Glomfjord a leading example of industrial symbiosis in the digital economy. We’re also exploring additional uses for recovered heat, such as warming local swimming pools or heating roads during the winter months to prevent ice buildup.
By capturing and redistributing our waste heat, we’ve turned one of the biggest challenges in data center design into a renewable asset that supports the surrounding community and ecosystem.
Removing diesel emissions
The third major sustainability challenge for data centers lies in emissions. Traditional facilities often rely on diesel generators for backup power, which can produce pollutants and contribute to local air quality issues.
At Glomfjord, we took a different approach. From the outset, we made the decision to eliminate diesel generators from our design. Instead, we draw on the reliability of Norway’s renewable grid and robust systems to maintain uptime.
This design choice completely removes direct emissions associated with diesel backup systems. It also simplifies operations, reduces maintenance requirements, and aligns with our goals of creating digital infrastructure that is as clean to operate as possible. Every part of our system at Glomfjord, from power intake to waste heat recovery, has been designed with sustainability in mind.
Toward sustainable AI infrastructure
The Glomfjord project is more than a showcase of engineering excellence; it’s a blueprint for how AI infrastructure can coexist with the environment where it sits. Every system, from cold-water intake to waste heat distribution, is designed to work harmoniously with the surroundings rather than against it.
As AI workloads surge in scale and complexity, the infrastructure that supports it must evolve accordingly. The next generation of AI facilities won’t just need to process trillions of operations per second — they’ll need to integrate seamlessly into sustainable ecosystems, adapt to new energy models, and contribute positively to the communities around them.
After building 54 data centers over the course of my career, this is the first time I truly feel like I’m building a legacy in a meaningful way. There’s a real sense of pride in knowing that what we’re creating here will have a lasting, positive impact. As I see the growing influence of AI, I feel I’ve played a small but important role in enabling that progress. For Nscale, it’s about driving AI forward with smart, efficient, and sustainable practices.
.png){kind=icon}
