From Observer, January 30:
As hyperscalers race to build A.I. capacity, the Nordics offer a proven model for resilience, efficiency and sustainability.
The rapid rise of A.I. is reshaping global infrastructure demands, pushing data centers from around the world to scale at unprecedented speed to support increasingly compute-intensive workloads. What was once a steady expansion has become a sprint, driven by generative A.I., large language models and real-time inference applications that strain existing power, cooling and connectivity systems.
Over the past year alone, hyperscalers have announced some of the largest digital infrastructure projects on record. In the U.S., an OpenAI-Oracle-SoftBank Stargate project has outlined five new data centers, totaling roughly five gigawatts of capacity, as part of a multi-year, multibillion-dollar expansion to support next-generation A.I. models. And in India, Google is investing roughly $6 billion to develop an infrastructure hub in Visakhapatnam. But while data centers may appear to be proliferating everywhere, not all locations are equally suited to the demands of A.I.
A.I. workloads have highly specific requirements, and where infrastructure is built has a direct impact on time to market, total cost of ownership and environmental sustainability. As power constraints, permitting delays and grid congestion continue to slow new projects in major markets, the central challenge has shifted. The focus is no longer simply on building capacity, but on finding where capacity can be responsibly developed at scale.
The Nordic region, traditionally known for its mining, steel, pulp and paper production, has experienced a rebirth in recent years as an ideal location for prominent businesses such as Spotify, Nokia, Klarna and Lego, alongside a growing ecosystem of cleantech and data-driven industries. Arguably, one of its fastest-growing sectors is that of A.I.-ready digital infrastructure. A powerful combination of forward-thinking governments and favorable natural conditions has enabled the area to offer systemic lessons for scaling A.I. sustainably.
What A.I. data centers actually need
At a fundamental level, A.I.-ready data centers depend on three primary elements: land, power and connectivity. A.I. workloads require dense concentrations of compute hardware to process vast volumes of data at speed, which in turn demands large, powered sites capable of supporting both the equipment itself and the cooling systems required to keep it operational.For real-time workloads, such as generative A.I. applications or financial trading platforms, connectivity is just as critical. Ultra-low latency networks are essential to maintain performance and reliability. Even small delays introduced by long-distance data transmission can degrade user experience or undermine trust in a product. These networks must also be highly resilient, with full redundancy built in to ensure consistent service.
The combination is progressively difficult to achieve. In many developed markets, the land most readily available for large-scale development is in rural areas where high-capacity connectivity may be limited. At the same time, power availability has emerged as a primary bottleneck. According to a report by the International Energy Agency, global data center electricity consumption is projected to more than double by 2030, reaching approximately 945 terawatt-hours—slightly more than Japan’s total electricity use today. The same report warns that roughly 20 percent of planned data center projects could face significant delays due to insufficient grid capacity.
These constraints are already visible. Ireland imposed a moratorium on new data center developments in the Dublin area beginning in 2022, citing unsustainable pressure on the national grid. The ban was lifted in December 2025, with strict new rules around on-site generation and renewable energy put in place. In the U.S., a recent JLL report found that power delivery wait times now stretch two to three years in parts of the Mountain West and New York metropolitan areas, and as long as eight to ten years in the Pacific Northwest.
These pressures are playing out against tightening regulatory scrutiny: this month, the U.S. Environmental Protection Agency closed a loophole that allowed hyperscale data centers to deploy portable gas-fueled power generators without federal permits, potentially signaling a shift toward more stringent environmental oversight of A.I. infrastructure buildouts.
At a moment when A.I. adoption is widely framed as essential to economic competitiveness, such delays are more than operational inconveniences. In the U.S., A.I. investment has become a major contributor to GDP, accounting for 20 percent to 25 percent of real GDP growth, second only to consumer spending. Infrastructure bottlenecks risk becoming a limiting factor for both technology companies and the broader economy.
The Nordic model....
....MUCH MORE
