Enhanced Geothermal Data Centers, 7 GW US Grid Gap, $750 M Crusoe Facility, and 11 Nordic Expansions (2021 to 2026)

AI Power Demand Risk, Crusoe and 7 GW US Grid Gap

The explosive growth in Artificial Intelligence has shifted the primary constraint for data center expansion from processing power to electrical power, forcing a geographic decoupling of AI workloads from traditional, grid-constrained US markets. This is no longer a forecast but a present-day reality, with operators now prioritizing access to abundant, low-cost, and stable energy above all else. This dynamic has elevated energy-rich regions like Iceland into critical hubs for the world’s AI infrastructure.

• Between 2021 and 2024, the narrative centered on the rising energy consumption of data centers, which hit an estimated 460 TWh globally in 2022. While concerns about power costs were prevalent, the primary strategy was still centered on efficiency gains within existing markets rather than large-scale geographic relocation.
• From 2025 to today, the problem has escalated from a cost issue to a fundamental availability crisis. In the U.S., nearly half of all planned AI data centers have been canceled or delayed, creating a capacity gap of approximately 7 GW. This validates the “energy-first” deployment model pioneered by firms like Crusoe Energy.
• This shift has transformed the colocation market in Iceland from a niche for specialized High-Performance Computing (HPC) to a primary solution for mainstream AI developers. Companies like at North are now building capacity specifically to absorb this displaced demand from power-starved US and European markets.

$750 M Brookfield Credit, Crusoe Energy Infrastructure Expansion

Capital flows are increasingly directed toward business models that solve the AI industry’s energy problem, signaling investor confidence in infrastructure solutions over pure software plays. The financing structures have matured from venture capital to large-scale asset-level debt, indicating that the energy-first approach to computing is now considered a bankable and scalable strategy.

• Crusoe Energy’s ability to secure a $750 million credit facility from Brookfield in June 2025 is a landmark event. This type of large-scale financing is typically reserved for mature infrastructure assets, and its application here validates Crusoe’s strategy of co-locating data centers with stranded or low-cost energy sources as a viable, long-term business model.
• Hyperscale operators are pursuing a parallel strategy of securing energy through massive direct offtake agreements. Google’s deal for up to 150 MW of geothermal power from Ormat Technologies and Meta’s procurement of geothermal energy from XGS Energy show a corporate trend of functioning like utility-scale energy buyers.
• These two investment tracks highlight a market bifurcation. Agile specialists like Crusoe are tapping into undervalued energy resources that are inaccessible to traditional developers, while giants like Google use their balance sheets to secure vast quantities of power, confirming that control over energy is now a key competitive advantage in the AI sector.

at North 80 MW Expansion, Nordic Colocation Partnerships (2023 to 2026)

Strategic alliances between AI cloud specialists and Nordic colocation providers have emerged as the most capital-efficient and rapid method for deploying power-intensive AI infrastructure. This symbiotic model allows AI firms to bypass greenfield development risks and gain immediate access to scalable, low-cost, renewable power.

Nordic Data Center Power Capacity Sees Strong Growth

This chart shows the projected growth of the Nordic data center market, directly contextualizing the strategic importance of atNorth’s expansion. The identification of ‘atNorth’ as a major player makes it a perfect fit.

(Source: Mordor Intelligence)

• The partnership model between Crusoe Energy and at North provides a blueprint for the industry. Crusoe gains instant access to a state-of-the-art facility engineered for Iceland’s climate and powered by geothermal energy, while at North secures a high-value tenant with predictable, high-density power demand.
• This trend is market-wide, not isolated. Verne Global, another major Icelandic operator, expanded its campus to 40 MW in 2023 and has a clear roadmap to scale beyond 100 MW in response to demand from HPC and AI clients.
• In November 2024, at North announced a major expansion of its ICE 02 facility to an ultimate capacity of 80 MW. This proactive build-out is a direct response to a robust pipeline of US and European clients seeking sustainable and cost-effective alternatives to their constrained home markets.
• The partnerships are increasingly specialized, designed to support racks demanding over 50 k W and requiring advanced liquid cooling. This demonstrates that Nordic data centers are not just providing space and power but are offering highly engineered environments tailored specifically for the extreme demands of AI hardware.

Table: Key Strategic Partnerships and Expansions (2023 – 2025)

US vs Nordics, Crusoe Geographic Diversification for AI Compute

The global map of digital infrastructure is being redrawn by energy availability, triggering a significant migration of non-latency-sensitive AI training workloads from traditional US hubs to the Nordic region. This geographic diversification is a direct result of the grid-saturation and multi-year interconnection queues that now plague established American data center markets.

Mapping the AI Data Center Value Chain

This diagram provides a high-level view of the ecosystem, illustrating how partners like data center operators (atNorth) and compute providers (Crusoe) fit into the overall value chain discussed in the section’s table.

(Source: insights4vc – Substack)

• Between 2021 and 2024, US markets like Northern Virginia and Texas remained the default locations for data center development, but the first signs of stress appeared as interconnection queues lengthened. During this period, Iceland was primarily a niche market for HPC and cryptocurrency mining, valued for its low-cost power but considered remote.
• From 2025 onward, the grid bottleneck in the US became a primary driver of site selection. The documentation of a 7 GW shortfall in planned AI data center capacity transformed the Nordics from a niche to a primary expansion target for US firms.
• Iceland’s value proposition, built on 100% renewable baseload power from geothermal and hydro sources, became a decisive advantage. The country’s cool climate, which allows for free-air cooling and leads to superior PUE ratios, further strengthened its case.
• The activation of the IRIS subsea fiber optic cable in March 2023 was a critical enabler, reducing data latency concerns and making Iceland a viable location for the massive, asynchronous data transfers required for large-scale AI model training.

95% Capacity Factor, Crusoe and Geothermal Commercial Scale

Geothermal energy has transitioned from a niche renewable resource to a commercially mature, baseload power technology uniquely suited to the rigorous demands of AI data centers. Its ability to provide stable, 24/7 power at a predictable cost directly addresses the core weaknesses of both intermittent renewables and congested fossil-fuel grids, making it a cornerstone of the sustainable AI infrastructure buildout.

• During the 2021-2024 period, geothermal’s primary benefit was perceived as its low cost and renewable credentials. While its reliability was understood, it was not yet the critical differentiator it would become.
• By 2025, as the AI industry confronted the reality of powering GPU clusters at near-constant high utilization, geothermal’s high capacity factor of over 95% became its most valuable attribute. This baseload characteristic provides the operational stability that AI workloads require and that intermittent sources like solar and wind cannot guarantee without expensive energy storage.
• The technology’s application has also matured beyond simple electricity provision. Advanced concepts, such as using waste heat from a geothermal plant’s cooling towers for an absorption chiller, create a symbiotic relationship that further enhances a data center’s energy efficiency and sustainability.

Crusoe and at North SWOT, Nordic AI Hub Strategy (2021 to 2026)

The Nordic data center model for AI workloads is built on an exceptionally strong foundation of renewable energy and natural cooling efficiency. However, its long-term growth trajectory is dependent on overcoming internal infrastructure constraints and navigating the threat of new energy technologies emerging closer to primary demand centers in the US and Europe.

Large Facilities Dominate Nordic Data Center Market

This chart quantifies a key characteristic of the Nordic data center landscape discussed in the SWOT analysis, showing the prevalence of large-scale facilities which is a crucial factor in the region’s strategy.

(Source: Mordor Intelligence)

• The core strength of the model lies in Iceland’s abundant, low-cost geothermal and hydropower, which provide stable baseload power at a fraction of the cost of constrained markets. This is amplified by a cool climate that dramatically reduces cooling-related operational expenses.
• The primary weakness is the finite nature of both the energy supply and the grid’s capacity to transmit it. While resources are vast relative to the population, they are not limitless, and “system bottlenecks” in transmission could throttle growth if not addressed through continued investment.
• The most significant opportunity is to capture the immense and growing overflow of AI compute demand from the US and Europe. Iceland is positioned to become the world’s premier destination for sustainable, large-scale AI model training.
• A long-term threat is the potential development of alternative baseload, carbon-free power sources, such as small modular reactors (SMRs), located directly within major demand centers, which could eventually reduce the need for geographic energy arbitrage.

Table: SWOT Analysis for the Nordic AI Data Center Model

Scenario Modelling, Crusoe’s Next Move in the Nordic Power Market

The single most critical factor determining the future growth of the Nordic AI hub is the pace of investment in national grid infrastructure. If grid modernization keeps pace with data center demand, the region’s position will be solidified. However, if transmission becomes a bottleneck, it could throttle growth and divert investment to other emerging energy-rich regions.

AI Infrastructure Investment to Surpass $6.5 Trillion

This chart quantifies the immense market opportunity that the Nordic AI hub strategy, detailed in the SWOT analysis, aims to capture. It provides the financial context for why the region’s strategic positioning is so critical.

(Source: Medium)

• If this happens: Iceland’s grid operator, Landsvirkjun, announces a multi-year, billion-dollar plan to upgrade its high-voltage transmission network and streamline the interconnection process for new industrial users. This would be a strong positive signal that the country is committed to removing bottlenecks and supporting continued data center growth.
• Watch this: The volume and terms of new Power Purchase Agreements (PPAs). A surge in long-term, fixed-price contracts signed by companies like Crusoe Energy and at North would confirm continued market confidence in the region’s ability to deliver power at scale. Conversely, a shift to shorter-term or variable-price contracts could signal uncertainty.
• These could be happening: We could see the emergence of fully integrated “energy parks, ” where new geothermal wells are developed in direct partnership with specific data center campuses. This would represent the ultimate realization of Crusoe’s energy-first philosophy, creating a closed-loop system of power generation and consumption that bypasses the public grid entirely.

The questions your competitors are already asking

This report covers one angle of the energy-driven relocation of AI workloads to Nordic data centers. The questions that matter most depend on your work.

• Which companies are gaining or losing ground in the Nordic AI data center colocation market?
• What is the outlook for US AI workload deployment in Nordic data centers by 2026?
• Crusoe Energy investments and funding. Is the $750M Brookfield credit facility on track to support its infrastructure expansion targets?
• Which US AI developers are adopting the Nordic colocation model to overcome domestic grid constraints?

This report does not answer these. Enki Brief Pro does.

Your question, your angle, your framework. SWOT, PESTL, scenario modelling. The same niche depth, built around the decision your work actually depends on.

Run your first brief in Enki Brief Pro