Geothermal Data Centers, 1, 000 TWh Demand Surge, Meta & Equinix PPAs, and $7.8 B Nordic Market Growth (2021 to 2026)

The explosive energy requirement of artificial intelligence has rendered power availability the primary constraint for data center growth, forcing a strategic migration away from saturated hubs toward locations with abundant, low-cost energy. This shift is not theoretical; it is a material reality reshaping digital infrastructure. The hypothetical 300 MW AI data center by Nebius and Data One in Iceland, powered by geothermal and hydro, exemplifies this new paradigm where energy geography, not proximity to users, dictates site selection. This analysis examines how the immutable realities of power access are redrawing the global data center map.

AI Power Demand Forces Data Center Migration to Energy-Rich Regions

The strategic calculus for data center location has fundamentally inverted, with operators now forced to bring the data center to the power source rather than the other way around. This “energy-first” approach is a direct response to the escalating power and cooling demands of AI hardware, which have created insurmountable bottlenecks in traditional markets.

• In the period from 2021 to 2024, the growing power consumption of AI began to visibly strain grids in legacy data center hubs. Grid operators like PJM, which covers the US East Coast, issued early warnings of potential capacity shortfalls driven by data center load, while markets like Dublin and Singapore imposed moratoriums on new grid connections.
• From 2025 to today, this strain has become an acute crisis. Projections from the International Energy Agency show global data center demand will surpass 1, 000 TWh by 2026, equivalent to the entire electricity consumption of Japan. Consequently, between 30% and 50% of large data center projects scheduled for 2026 are now expected to be delayed or canceled due to power constraints.
• The core issue is power density. A traditional server rack requires 7-10 k W, but an AI rack filled with GPUs can demand over 100 k W. This concentration of power makes grid access and cooling efficiency the most critical operational and financial variables.

$650 M+ in PPAs, Meta and Equinix Secure Renewable Energy for AI

To mitigate grid instability and volatile energy prices, hyperscalers are increasingly using long-term Power Purchase Agreements (PPAs) to secure dedicated, low-cost renewable energy. These contracts directly finance the construction of new power generation, effectively creating a private energy supply chain for AI data centers and de-risking project development.

• PPAs provide the price stability and financial certainty necessary to underwrite the massive capital expenditure of gigawatt-scale AI facilities. By locking in electricity prices for 10-25 years, operators can insulate themselves from spot market volatility.
• The surge in PPA demand from data centers is transforming the energy market itself. In Texas’s ERCOT grid, for example, the intense competition for power drove the fair value of wind PPAs up by 16% in 2025, demonstrating a direct causal link between AI growth and the valuation of renewable assets.
• Recent large-scale deals show this is a global strategy. In May 2025, Meta announced a 650 MW solar PPA with AES to power its US data centers. In February 2026, Equinix secured a 15-year, 121 MW solar PPA in Japan, the largest of its kind in the country.

Table: Recent Power Purchase Agreements for AI Data Centers

Nordic Region vs. US Hubs, Data Center Growth Follows Geothermal and Hydro Power

Data center construction is rapidly shifting from power-constrained markets like Northern Virginia to energy-abundant regions, with the Nordics emerging as a prime destination. This geographic pivot is driven by the region’s unique combination of 100% renewable power, natural cooling, supportive policies, and faster grid connection times.

• Between 2021 and 2024, the limitations of traditional hubs became undeniable. Ireland’s data centers began consuming more electricity than all urban homes combined, prompting a moratorium, while Northern Virginia’s grid reached its capacity.
• By 2025, the consequences of this saturation became severe. Interconnection queues to connect to the grid in primary US markets stretched beyond four years, making rapid deployment for AI physically impossible.
• In stark contrast, the Nordic data center construction market is now projected to reach $7.83 billion by 2030. Iceland offers a nearly ideal environment, with a power grid sourced entirely from stable geothermal and hydroelectric resources and a cold climate that allows for free air cooling, reducing operational costs by up to 40%.

Energy-First Site Selection Matures From Niche Strategy to Industry Standard

The strategy of co-locating data centers with power sources has transitioned from a niche cost-saving tactic to a standard, mandatory practice for deploying large-scale AI infrastructure. The extreme power density of AI hardware has made direct access to generation not just an economic advantage but a logistical necessity.

• In the 2021–2024 period, “following the electrons” was an advantageous strategy primarily for cost optimization and ESG marketing. Early movers established a presence in the Nordics to capitalize on lower PUEs and cheaper renewable energy.
• By 2025, this approach became a survival strategy. Grid saturation and multi-year project delays in traditional markets made it a prerequisite for any new large-scale development, confirming the market had hit a structural wall.
• The maturity of this approach is validated by hyperscalers making direct investments in energy generation. Both Google and Meta have entered into partnerships with energy producers like Fervo Energy to secure geothermal for data centers, guaranteeing a dedicated and carbon-free power supply for future AI workloads.

SWOT Analysis, Geothermal Data Center Strengths and Latency Risks

While locating data centers in energy-rich regions like Iceland offers significant power cost and sustainability advantages, operators must manage the inherent trade-off of higher network latency. The strategy is optimized for AI training and other non-real-time workloads, where compute cost outweighs the need for instantaneous response.

• The primary strength is access to stable, low-cost, 100% renewable power, which dramatically lowers operational expenditures and helps companies meet aggressive corporate ESG mandates.
• A key weakness remains the physical distance from major population centers, which can introduce latency that is unacceptable for certain real-time applications like gaming or financial trading.
• The opportunity is immense, driven by the insatiable AI power demand and persistent grid failures in legacy markets, creating a clear path for growth in new energy-rich frontiers.
• The main threat is the potential for these new regions to eventually face their own grid constraints as they absorb hundreds of megawatts of new data center load, repeating the cycle seen in traditional hubs.

Table: SWOT Analysis for Geothermal-Powered Data Centers

Scenario Model, 1, 000 TWh Demand and the AI Data Center Migration

If traditional data center hubs fail to accelerate grid upgrades and new power generation over the next 12-24 months, the migration of AI compute to energy-rich frontiers will become a permanent, large-scale realignment of global digital infrastructure. The inertia of regulated utility planning cycles makes this outcome highly probable.

• If this happens: The current flow of AI data center projects to regions like the Nordics, Quebec, and parts of the US with stranded power will accelerate into a flood.
• Watch this: Continued announcements of multi-hundred-megawatt PPAs by hyperscalers are the clearest leading indicators of where the next wave of data centers will be built. Also, monitor the capital expenditure plans of utilities in Virginia, Arizona, and Texas; any delays or downward revisions will confirm further momentum for remote deployments.
• These could be happening: The success of large-scale projects will validate the economic model and encourage more operators, including specialists like Switch and Vantage Data Centers, to aggressively pursue similar “energy-first” strategies, further cementing the geographic shift in compute capacity.

The questions your competitors are already asking

This report covers one angle of how AI power demand is reshaping the global data center map. The questions that matter most depend on your work.

• Nebius and DataOne activities in Iceland. Is the 300 MW geothermal data center project progressing from announcement to construction?
• Which Nordic data center operators are gaining or losing ground in the race to capture AI workloads?
• What is the outlook for geothermal and hydro-powered data center deployment in the Nordics by 2030?
• Which hyperscalers and AI companies are adopting an ‘energy-first’ site selection strategy by moving to the Nordics?

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