The global race to expand AI compute is pushing data centers underwater, into orbit, and adjacent to new energy sources — reshaping critical infrastructure and geopolitical risk.

Why This Matters

AI will soon demand more power than national grids can sustainably supply.

Artificial intelligence, cloud growth, and edge compute are driving an unprecedented expansion of data infrastructure. But:

• Electricity scarcity and emissions caps constrain land-based expansion.
• Climate and physical threats highlight fragility of current infrastructure.
• Nations are rapidly reclassifying data centers as strategic assets.
• The energy mix is shifting: renewables + nuclear are emerging as core to AI scaling.

This emerging model forces executives and national-security planners to reconsider where facilities are built, how they are powered, and who controls them.

Key Points

• Subsea and orbital data centers are moving from concept to deployment; both reduce cooling costs and avoid land-use politics.
• Electricity, not chips, is becoming the primary bottleneck for AI expansion.
• Nuclear-powered data centers are entering commercial strategy roadmaps for major operators.
• Data centers are now critical infrastructure — with policy, resilience, and cyber implications.
• Growing threats: fires, grid strain, ransomware targeting physical assets, climate-driven outages.
• Nation-state competition over compute capacity is accelerating infrastructure militarization.

What Next? Maxie Reynolds Joins Us at OODAcon 2025

Toward the end of his May 2025 OODAcast conversation with Maxie Reynolds (The Attacker Mindset: Maxie Reynolds on Red Teaming, Underwater Data Centers, and Human Nature), OODA CEO Matt Devost very casually mentioned to Maxie that “should join us at OODAcon this year.”

Maxie will in fact be joining us in Reston tomorrow, on a panel that grew out of one of the many topics Maxie discussed with Matt on the OODAcast in May: the journey that led to the creation of a company that builds underwater data centers, a novel fusion of her industrial and red teaming experiences. Maxie shared the rising interest in submerged infrastructure, particularly after China’s moves into space and the demands of modern AI-driven cooling systems.

Matt was fascinated by the topic at the time – and here we are this week with a follow-up OODAcon 2025 conversation between Maxie and OODA CTO Bob Gourley – Data at the Edge: Subsea and Sustainable Next-Gen Datacenters:

“Gain new insights into how deploying subsea data centers at the network edge can dramatically improve sustainability and reduce operational costs by leveraging the ocean’s cooling properties and proximity to renewable energy sources.”

Strategic Trendlines

Watch these strategic trendlines in the next 12–18 months:

1. Orbital and subsea pilot deployments, followed by national procurement interest.
2. AI-specific power corridors and modular micro-reactor pairings with hyperscaler campuses.
3. Cyber-physical defense mandates as classification shifts from industrial facilities → national security assets.
4. Geoeconomic leverage: nations with surplus green power become AI hubs; others face dependency risks.
5. Insurance and regulatory pressure will reshape site-selection risk models.

Recommendations from the OODAcast Conversation and The Art of Attack

Adopt an attacker mindset for infrastructure strategy: Think like an adversary when planning compute sites: identify weak links in power, cooling, access, and logistics. Simulate grid failures, physical intrusions, and regulatory shocks to surface hidden risks.

Follow the power and trust, not just the compute: Shift focus from silicon to energy and logistics: choose sites with surplus off-grid power, water access, and minimal adversarial dependencies. Anchor infrastructure where resilience is harder to disrupt.

Design for layered defense and deception: Build compute infrastructure with redundant power/cooling paths, hidden fallback nodes and alternate access routes to confuse adversaries. Use deception as a strategic layer, not just obstacle.

Embed continual stress-testing and adaptive governance: Introduce regular red-team style war games and external audits focused on compute-infrastructure assumptions. Require every major deployment to include scenario-based risk assessment covering energy sovereignty, sabotage and regulation.

Prioritize the human dimension and organizational culture of resilience: Cultivate a culture where admitting near-misses is encouraged, vulnerability disclosures are rewarded, and non-technical teams engage in attacker-mindset drills. Train logistics, facilities and power teams alongside cyber to align all layers of the defenses.

Position for strategic advantage—not just defense: Use the attacker mindset to identify infrastructure opportunities where your organization can leap ahead: build compute where energy access is cheap, jurisdictional friction low, and resilience high. Turn risk mitigation into competitive differentiation.

Additional OODA Loop Resources

Underwater and Orbital: The Next Frontier for Data Centers

This piece surveys subsea and orbital data-center concepts as emerging responses to land, cooling, and resiliency constraints. It frames these unconventional siting strategies as part of a broader shift toward energy- and environment-aware infrastructure for AI-era compute.

The Future of the Energy Sector and AI Data Centers

An analysis of how electricity availability, grid stability, and fuel choices now shape the feasibility and location of AI data centers. It highlights the tightening linkage between compute growth and long-horizon energy investments, including renewables and advanced nuclear.

Constraints to Growth Across Exponential Technologies

Outlines non-silicon bottlenecks—electric power, skilled trades, and large-scale infrastructure mobilization—that limit deployment speed across emerging technologies. For data centers, these constraints translate into extended timelines and higher execution risk.

The Commercial Convergence of AI Compute and Nuclear Technology

Discusses the growing commercial logic of pairing high-density AI compute with nuclear generation. It points to small modular reactors and novel procurement structures as potential pathways to off-grid or firm, low-carbon power.

The Climate Crisis, Texas’ Bitcoin Miners, and European Farmers

Explores how rising electricity demand from digital industries collides with climate impacts and competing sectoral needs. The piece foreshadows policy and market friction over power allocation as AI loads ramp.

U

The Climate Crisis, Texas’ Bitcoin Miners, and European Farmers

Illustrates demand-side conflict over electricity between crypto mining, agriculture, and emerging AI loads — foreshadowing political battles over industrial power priorities.

Additional Sources from the Analysis

Implications of Artificial Intelligence-Related Data Center Electricity Use

A National Academies workshop assessing the grid, emissions, and sustainability risks from AI-driven compute; reinforces urgency of alternative siting and power strategies.

South Korea Data Center Fire

Case study showing cascading physical → cyber risk: a facility fire triggered national cyber-threat escalation, emphasizing data centers as critical national targets.

Google $9B Virginia Data Center Project Investment

Signals hyperscaler concentration near key U.S. government and network backbones — raising resiliency and zoning complexity.

New Era of Nuclear Energy for AI

A WSJ update on commercial enthusiasm for nuclear as the only viable path to multi-terawatt AI compute. A new era of nuclear energy seems to be emerging, and Google looks to power AI data centers.

UK Data Centers as Critical Infrastructure

Policy signal: reclassification drives enhanced regulation, security protections, and government oversight.