Terrestrial variables influencing data centers

Our corporate site Contact us

From Arctic tundra to ocean floors, the next generation of computing infrastructure is departing from conventional geography. Here's why it matters.

Data centers are now critical infrastructure powering the global economy. These vast complexes of servers and cooling systems consume electricity on par with entire countries, process exabytes of data daily, and underpin everything from video calls to autonomous vehicles. Yet for all their importance, most data centers occupy relatively predictable real estate: industrial parks near major cities, connected to reliable power grids and fiber networks.

However, across the industry, operators and investors are reconsidering fundamental assumptions about where compute infrastructure can — and should — exist. The result is a wave of experimentation with locations that might have seemed impossible a decade ago: subsea modules cooled by ocean currents, underground facilities in decommissioned mines, Arctic installations leveraging perpetual cold, even orbital platforms in low-Earth orbit.

Exploring unusual locations is a pragmatic response to mounting constraints that are encouraging the industry to reimagine the relationship between computing, energy, climate, and geography itself.

The pressure points driving change

Energy: The exponential demand curve

Modern data centers rank among the largest consumers of electricity in most developed regions. A single hyperscale facility can draw more than 1,000 megawatts — enough to power 800,000 US homes. As artificial intelligence workloads proliferate, particularly large-scale model training, that figure is climbing rapidly. Training a frontier AI model can consume continuous power for weeks, generating heat that must be immediately dissipated or risk catastrophic hardware failure.

The traditional solution — mechanical cooling systems — often consumes 30% to 40% of a facility's total energy budget. In warmer climates, that proportion can rise further. This creates a cycle: more computing generates more heat, requiring more cooling, consuming more energy, and generating higher operational costs and carbon emissions.

Climate volatility and resource scarcity

Extreme heat events put pressure on water resources and electricity generation. Drought conditions restrict water availability for evaporative cooling towers — a critical issue in places like Arizona and parts of Europe. Grid instability, whether from extreme weather or renewable intermittency, introduces operational risk.

These current operating realities are forcing developers to incorporate resilience planning into site selection in ways that may have been unnecessary before.

Land and regulatory constraints

Urban expansion and environmental regulations are tightening the supply of suitable land near demand centers. Zoning restrictions, noise ordinances, and community opposition can complicate approval processes. Simultaneously, regulators are imposing stricter emissions reporting requirements and water-use limits on energy-intensive facilities. The effect is a growing gap between where demand exists and where facilities can feasibly be built and operated.

The new geography of computing

Against this backdrop, unconventional locations offer specific, quantifiable advantages compared to traditional sites.

Arctic and high-latitude installations

Scandinavia, Iceland, Canada, and certain Arctic regions deliver ambient temperatures that can effectively eliminate mechanical cooling requirements for much of the year. This translates directly into lower energy consumption and extended hardware lifecycles. Many of these regions also have abundant renewable energy — geothermal in Iceland, hydropower in Norway — enabling operators to meet both cost and sustainability objectives.

Subsurface and repurposed industrial sites

Decommissioned mines and underground caverns provide thermally stable environments with minimal temperature fluctuation. They are naturally shielded from surface weather extremes, require less active climate control, and repurpose existing industrial infrastructure rather than consuming greenfield land. For developers, this can accelerate permitting and reduce upfront capital expenditure.

Underwater and offshore platforms

Submersible data centers use seawater for passive cooling, eliminating the need for conventional HVAC systems. By positioning modules near coastal demand centers, operators can achieve proximity to users without competing for expensive coastal real estate. Floating platforms pursue a similar logic, combining maritime cooling with potential access to offshore wind or wave energy.

Orbital infrastructure

Low-Earth orbit (LEO) remains largely conceptual, but the physics are compelling: continuous solar exposure and the vacuum of space offer theoretically limitless power and passive cooling. Launch costs continue to decline, and satellite communication latency is improving. While likely years from commercial viability, the conversation signals how seriously the industry is taking the need for alternative infrastructure models.

The investor lens: risk, return, and strategic value

For private equity and infrastructure investors, unconventional data centers present a distinct opportunity set, with analysis suggesting that digital infrastructure could attract almost US$7 trillion in investment by 2030. However, as with most opportunities, it requires careful evaluation of risks that may not appear in traditional real estate or infrastructure deals.

The investment hypothesis

Facilities that eliminate or significantly reduce cooling costs can achieve preferred operating margins. In markets with carbon pricing or strict emissions mandates, these advantages compound over time. Early movers in novel site categories can also establish competitive barriers; there are only so many suitable Arctic sites with renewable grid access, or coastal locations with favorable subsea conditions.

Sustainability considerations add another dimension. Institutional capital increasingly flows toward assets with credible sustainability profiles. Data centers powered by renewables and cooled without water-intensive systems align well with climate commitments, potentially unlocking cheaper capital or premium valuations.

Risk considerations that matter

Many alternative concepts are transitioning from pilot projects to commercial scale. Technical challenges, such as sealing integrity for underwater modules and ventilation systems for underground facilities, may not have been validated over multi-year operating cycles. Development timelines are longer and less predictable, which can affect IRR calculations and fund deployment schedules.

Regulatory frameworks remain uncertain, particularly for maritime and orbital installations that cross jurisdictional boundaries. Permitting processes can be opaque, and the absence of precedent creates approval risk that is often difficult to model.

Exit strategy deserves particular attention. Unconventional assets lack deep secondary markets. Buyers must be comfortable with novel operating models and potentially higher technical risk. Without comparable transactions to anchor valuations, pricing becomes more subjective, which can be a challenge when fund structures require liquidity within defined timeframes.

Capital structure innovation

These challenges are spurring new financing approaches. Blended capital structures — combining institutional equity, project finance debt, and pre-commitments from anchor tenants — help distribute risk across parties with different return expectations and time horizons. Sustainability-linked debt instruments tie borrowing costs to verified performance metrics, aligning incentives and lowering capital costs for operators who meet sustainability targets. Similar to today’s environment, developing partnerships with hyperscale cloud providers can de-risk revenue assumptions by securing long-term contracts before construction begins.

The critical constraint: connectivity

While energy efficiency and climate resilience tend to dominate the narrative around unconventional locations, data connectivity often determines whether a site is viable at all.

Unlike power, which can sometimes be generated on-site, connectivity depends on fixed physical networks — fiber cables, subsea routes, terrestrial backhaul — that are often expensive and slow to build. Remote Arctic facilities may sit hundreds of kilometers from the nearest fiber point of presence. Underwater installations require dedicated subsea cables that are vulnerable to ship anchors, seismic activity, and marine hazards. Orbital platforms face physics-imposed latency that precludes real-time applications.

This creates a workload sorting problem. Unconventional sites generally work well for compute-intensive, latency-tolerant tasks, such as AI model training, genomic analysis, batch processing, and archival storage. However, they are often poorly suited for applications requiring sub-millisecond response times or direct user interaction. This narrows the addressable market and shapes revenue assumptions.

Connectivity also introduces regulatory complexity. Cross-border data flows are subject to varying legal regimes regarding privacy, sovereignty, and content regulation. Offshore and orbital facilities muddy jurisdictional questions, potentially complicating compliance and restricting certain use cases altogether. For investors, this means that connectivity risk extends beyond technical performance into legal and market-access considerations.

What the next decade may bring

The trajectory over the next five to ten years is likely to be one of selective adoption rather than wholesale transformation. Unconventional locations will find their niches. For example, Arctic facilities for AI training clusters, subsea modules near coastal megacities, and underground sites in land-constrained markets. While conventional data centers continue to dominate overall capacity.

What will likely distinguish successful deployments is rigorous attention to the fundamentals: proven technology at scale, transparent regulatory pathways, credible connectivity plans, and capital structures matched to risk profiles. Pilot projects that can demonstrate reliable multi-year operations will likely attract institutional capital.

Beyond the immediate horizon, more fundamental shifts may emerge. Flexible or modular compute infrastructure that can be repositioned in response to energy availability, climate events or demand patterns. AI-optimized site selection that dynamically balances energy costs, connectivity, cooling efficiency, and regulatory environments. Digital sovereignty frameworks that explicitly address compute infrastructure existing beyond traditional territorial boundaries.

These possibilities remain speculative, but they reflect how the relationship between computing and physical place is becoming more fluid, more contingent, and more responsive to environmental and economic realities.

Final considerations

Unconventional data center locations are neither a panacea nor a distraction. They represent rational adaptations to genuine constraints, constraints that will likely intensify as computing demand grows and climate pressures mount.

For investors, the opportunity emerges from the ability to rigorously measure and quantify risk, enabling informed and calculated decisions. Those that benefit will apply disciplined analytical frameworks to evaluate assets, while proactively managing inherent risks and optimizing planning. By embracing this approach, investors can more confidently navigate complexities, limit surprises, and unlock value through strategic decisions.

The risk landscape by site type

Grid reliability and transmission capacity.
Seasonal accessibility for maintenance.
Workforce availability.
Potential environmental and indigenous land conflicts.

Ventilation and fire suppression challenges.
Higher development costs.
Limited physical access for heavy maintenance.
Regulatory complexity around underground land use.

Corrosion, biofouling, and sealing integrity.
Marine ecosystem disturbance.
Cable and connectivity risk.
Emergency retrieval complexity.

Maritime weather exposure.
Mooring and stability challenges.
Regulation under international waters.

Launch cost and orbital debris risk.
Regulatory uncertainty (multiple sovereign regimes).
Total lifecycle cost.
Repair or hardware refresh cycles.
Each of these dimensions has implications for insurability, financing cost, contractual terms, and operational resilience.

Article

Terrestrial variables influencing data centers

From Arctic tundra to ocean floors, the next generation of computing infrastructure is departing from conventional geography. Here's why it matters.