Sweden is widely recognized as one of the most sustainable countries in the world. The nation consistently ranks among global leaders in environmental performance, maintains a strong commitment to the UN Sustainable Development Goals, and continues to advance its transition toward a low-carbon, green economy.

Yet, in the age of AI, sustainability progress can feel increasingly difficult to maintain. The inherent power demands of AI training and inference models, alongside the growth of high-performance computing, are placing greater pressure on energy resources and infrastructure.

That said, Sweden’s strong focus on sustainability puts the country in good stead to help shape the conversation around scaling digital infrastructure responsibly in the AI era. One company contributing to that effort is Conapto, a Stockholm-based data center colocation provider that is working to secure additional land and power capacity while maintaining its sustainability commitments.

In this DCD>Talks episode, Stefan Nilsson, CCO at Conapto, discusses sustainable digital infrastructure, Sweden’s evolving data center market, and how the industry is preparing to meet the demands of next-generation AI workloads.

Why is AI having such a profound effect on sustainability?

The explosive popularity of generative AI and machine learning has driven unprecedented demand for data centers, creating significant growth opportunities across the industry. However, the practical realities of meeting this demand have also introduced new challenges. Faced with pressure to scale quickly, operators and manufacturers that might previously have prioritized more sustainable options are turning to quick fixes that are not always designed with long-term sustainability in mind.

“The biggest shift is time to market in conjunction with the fact that the asks and the need for deployments are much, much bigger. That puts a lot of constraints on finding power, land permits, the ability to execute, and financing,” says Nilsson.

Sweden is not exempt

While this attitude of “more, more, more” continues to facilitate digital connectivity, the reality is that the infrastructure required to support it has to live somewhere. As a result, the data center industry has faced its fair share of regulatory changes, driven by well-warranted NIMBYism and community pushback.

Sweden, with its vast forested landscapes and areas of sparsely populated land, has been less affected by this trend than many other markets. With a nationwide ratio of roughly one person per 455,000 square feet of land, it’s easy to understand why. But the thing is, proximity matters.

Power resilience and near-zero latency sit at the pinnacle of digital demands – not only for customer satisfaction, but for the mission-critical applications that data centers serve across public and private sectors, including hospitals, banking, transport, and emergency services. And with approximately 87 percent of Swedes living in urban areas that account for just 1.5 percent of the country’s total landmass, attitudes toward large-scale digital infrastructure developments are beginning to shift.

“We are starting to see some of that pushback happening in the Nordics, as the grid constraints are also happening here. And as an industry we are challenged by other industries,” Nilsson explains, adding:

“Sweden has always been an industrialized nation, and now suddenly a new industry comes in and wants to be part of the puzzle, so it takes time for us to educate the rest of society in terms of why we need data centers, why we need to put them in Sweden, and why we need to put them in Stockholm.”

Balancing priorities

The constant pressure to execute faster and scale more quickly than ever before – while doing so sustainably – has forced industry leaders to rethink their operating models and find more effective, efficient ways to deliver compute capacity.

For Conapto and many others, this has meant a shift toward prefabricated and modular architectures, as Nilsson explains:

“It’s the only way forward when it comes to designing and building new facilities, because modular design makes it easier to grow with the needs of customers, and prefab shortens delivery times and improves the quality assurance of the actual delivery. It’s both a time to market perspective and a quality assurance perspective.”

This trend also reflects the increasingly blurred lines between the data center white space, which houses the core revenue-generating compute and networking, and the gray space, containing the supporting mechanical and electrical systems.

In the past, operators could design both gray space and white space upfront, as customer hardware was relatively standard, consisting of CPU-based servers with predictable power densities. Today, the rise of high‑density, air‑cooled, and liquid‑cooled systems means the white space must be designed in much closer collaboration with customers, with the specific hardware requirements often determining the final facility design.

But building at the pace demand requires and engaging in early-stage co-design do not always go hand-in-hand. The challenge is compounded by the rapid evolution of AI hardware, with new chip architectures emerging on an increasingly frequent basis.

“We can’t build fast enough, but at the same time we are consuming more product, more steel, more concrete – more of everything. That puts a lot of pressure on the industry to do it as sustainably and consciously as possible,” says Nilsson.

So, what’s the Swedish solution?

For Nilsson, the answer lies in a combination of heat reuse, renewable energy investment, and closer collaboration between data center operators, utilities, and technology vendors.

“It’s easier to do it in metro areas. For example, Stockholm has the largest district heating network in the world, so we have a lot of opportunities for heat reuse,” he says.

“But if you have a big AI factory out in a remote area, there are other options. You can work with PPAs, for example, to invest more in renewable power. You could also look at on-site or near-site generation with a utility company to help offload the grid.”

In other words, scaling AI and data center capacity quickly must go in tandem with enabling new, sustainable sources of power, rather than simply consuming existing capacity.

Nilsson argues that data center operators need to go beyond just purchasing electricity and take a more active role in supporting local energy production. As grid congestion and power constraints become more common – even in traditionally power‑rich markets like Sweden and Norway – the industry must help create the capacity it requires.

When it comes to supporting ever-evolving chip architectures, Nilsson points to the importance of close partnerships and industry expertise:

“The only answer to that question is to be very close to the chip manufacturers, look at what they’re doing, try to be as flexible and agile as possible with the design, and then adapt the way we do things.”

While AI is forcing data centers to become faster, denser, and more sustainable, operators like Conapto are finding that success depends on collaboration – whether through co-investing in energy production, co-designing infrastructure with OEMs and customers, or preparing for continued shifts in power and cooling architectures.

For deeper insights into scaling sustainability in the AI era, watch the full DCD>Talks episode here.