Power plays: AI data center boom hits gridlock as energy becomes the new gold

By mid-2026, power availability has unequivocally emerged as the linchpin in AI data center deal viability and expansion, with securing reliable power and energized land now considered paramount in deal diligence and execution. As one analysis put it, "Power is the first part of diligence that you do," underscoring its role as the fast track to understanding site viability and securing a competitive queue position. This criticality has elevated powered land to a distinct investment class, driven by hyperscalers' massive development pipelines demanding large-scale power increments, making sites with secured power the most coveted assets in the market.

The escalating power demands of AI data centers have fostered unprecedented collaboration between traditional power infrastructure groups and data center operators, who are jointly innovating solutions such as behind-the-meter power options to overcome grid constraints. This cooperative approach reflects the sector's urgency to meet surging consumption needs, as evidenced by BlockchAIn Digital Infrastructure's recent expansion of utility load at its CLT-01 site from 40 MW to 65 MW under a 15-year Electric Service Agreement, leveraging existing 34.5 kV infrastructure without major upgrades. Such long-term power contracts are now foundational to attracting major clients and underpinning future infrastructure growth.

The power bottleneck has become so acute that it now surpasses capital availability as the primary constraint on AI data center expansions, with operators prioritizing energized sites and grid connectivity to sidestep prolonged delays. This shift has extended project timelines, particularly where alternative energy solutions are slower to materialize, making the ability to procure land, power, and build facilities—backed by a proven operational track record—more valuable than ever. Consequently, data center operators capable of navigating these power challenges command a premium in investment viability and execution capability.

By early 2026, the strategic partnership between UGI Energy Services and Prime Data Centers underscored the critical role of natural gas infrastructure in powering sprawling AI data center campuses. Prime Data Centers anticipates natural gas demand surpassing 100,000 dekatherms per day within three to five years, leveraging Pennsylvania's abundant reserves and existing pipeline networks to ensure energy resilience. This collaboration also includes on-site gas-fueled electric generation facilities, a tactical move to secure a stable and localized energy supply while UGI retains significant underground storage and resource rights, highlighting a nuanced balance between infrastructure control and operational stability.

Looking toward the end of the decade, the data center industry is positioning itself as a pivotal anchor customer for emerging Small Modular Reactor (SMR) nuclear power, with widespread agreements signaling anticipation that SMRs will become a cornerstone stable energy source around 2029-2030. However, the path to commercial viability is tempered by regulatory hurdles and deployment challenges, prompting interim strategies such as extending the operational life of existing nuclear plants through data center partnerships. Despite nuclear's long-term promise, its slow deployment pace and persistent public resistance mean that near-term reliance remains limited, with the industry also exploring less exotic energy management tactics like curtailment and energy storage to claw back gigawatts more immediately.

By early 2026, North America's AI data center landscape has fractured into distinct regional markets, with Texas, Virginia, the Midwest, and Western Canada diverging sharply due to differences in regulatory frameworks, grid capacity, and political climates. This fragmentation has driven hyperscalers to pursue geographic diversification strategies to mitigate congestion and policy risks, fueling billion-dollar campus proposals in secondary metros like Charlotte, Tulsa, and Kansas City. Notably, Northern Virginia’s saturation and zoning resistance have redirected hyperscale demand to the Southeast, particularly Charlotte and Georgia, which now emerge as durable AI infrastructure corridors benefiting from more supportive political environments and fewer bottlenecks.

Canada’s Alberta, Quebec, and Manitoba have solidified their status as long-term AI data center hubs by leveraging lower power costs, surplus generation capacity, and strategic proximity to U.S. fiber routes, effectively serving as sovereign-aligned GPU workload spillover zones. This regional advantage contrasts with constrained U.S. grids and underscores a broader North American trend where power-secured land aggregation near substations and generation assets is becoming a prized bundled infrastructure asset, compressing permitting risks and accelerating development timelines. Concurrently, Texas remains the backbone of the continent’s AI infrastructure, anchored by multi-gigawatt projects in Amarillo and the Permian Basin, where deregulated power markets and flexible permitting continue to attract energy-first AI development.

In the Nordic region, large-scale AI data center expansion is gaining momentum through strategic financial and industrial partnerships that emphasize sustainability and technological sovereignty. atNorth’s €100 million expansion of its Espoo facility to 45 MW by 2027, alongside its launch of the 120 MW NOR01 data center in Haugaland scalable to 350 MW by 2028, exemplify this growth. Meanwhile, Nscale’s $790 million financing from a consortium of Nordic banks to expand its Narvik hub to 115 MW—with plans for 230 MW and deployment of over 30,000 Nvidia Vera Rubin GPUs—highlights the region’s regulatory and environmental advantages. These projects are further bolstered by alliances like Nordic Compass, chaired by ex-Finnish PM Katainen, which unites over 25 leading companies to drive sustainable, sovereign digital infrastructure across the Nordics.

Strategic partnerships leveraging regional energy resources are pivotal in supporting AI data center growth, as demonstrated by the collaboration between UGI Energy Services and Prime Data Centers in northern Pennsylvania. This alliance capitalizes on abundant natural gas reserves, existing pipeline infrastructure, and a skilled workforce to meet projected natural gas demands exceeding 100,000 dekatherms per day within three to five years, ensuring energy resilience for large-scale AI and high-performance computing operations. Such energy-centric collaborations underscore how regional resource endowments and infrastructure integration are critical to sustaining the escalating power needs of AI campuses.

By early 2026, a pronounced disconnect has emerged between the rapid influx of capital into AI data center projects and the slower pace of power availability and site readiness, underscoring a fundamental bottleneck in execution. Investment flows, exemplified by Nscale's $790 million financing round and Microsoft's joint decision to delay completion for optimal GPU deployment, are increasingly decoupled from traditional construction timelines and regulatory constraints. This divergence is further illustrated by OpenAI's exit from European projects due to power offtake challenges, while giants like Microsoft and Google absorb capacity through alternative leasing, highlighting that capital alone cannot overcome infrastructural and regulatory hurdles.

Success in scaling AI data centers hinges not merely on deploying vast capital but on the strategic synchronization of financing, infrastructure readiness, and tenant demand. As market participants grapple with unprecedented deal sizes—such as new entrants leasing 500 megawatts and North America surpassing 15 gigawatts of leasing in 2025—operators with proven track records in securing land, power, and phased financing structures gain a competitive edge. The complexity of structuring deals to align with phased tenant growth and evolving power solutions like fuel cells and dedicated generation underscores that execution excellence, rather than capital volume, will determine market leadership.

Nebius exemplifies a hybrid development strategy that blends ownership of expansive AI campuses with colocation partnerships to accelerate deployment while mitigating operational risks. Their Kansas City 1.2GW campus showcases advanced water-efficient cooling technologies, and their Finland sites leverage the region's cool climate and renewable energy, demonstrating a keen integration of sustainability with scale. Moreover, Nebius’s $10 billion AI factory in Finland, developed in partnership with Polarnode and utilizing Nvidia technology, highlights how strategic alliances and cutting-edge tech integration are central to their expansion blueprint.

atNorth’s involvement in Nordic Compass, a pan-Nordic alliance chaired by former Finnish Prime Minister Jyrki Katainen, underscores a collaborative approach to overcoming power and land constraints through sustainable innovation. By aligning digital infrastructure with regional energy systems and advanced computing, this consortium of over 25 companies aims to bolster Nordic and European technological sovereignty and resilience. This model reflects a broader shift toward embedding sustainability credentials at the core of AI data center growth, positioning energy-efficient operations as a competitive differentiator in the race for high-performance computing capacity.

BlockchAIn Digital Infrastructure’s recent expansion of utility load at its CLT-01 facility from 40 MW to 65 MW, achieved without major electrical infrastructure upgrades, illustrates how optimizing existing assets can enable rapid scaling of AI data center capacity. Their new 15-year Electric Service Agreement not only secures long-term power availability but also anchors commitments from major AI and financial sector clients, reinforcing the critical role that stable, extended power contracts play in attracting high-profile tenants and supporting sustainable growth trajectories in a constrained resource environment.