Submersible Datacenter

DeepData

Shipyard-built compute modules submerged in lakes. Where water is the heatsink.

~1.03 Target PUE
90% Less Cooling
75% Carbon Cut
2-3yr ROI Payback
The Crisis

AI Is Breaking Datacenters

Traditional cooling can't keep up. The industry is drowning in heat, cost, and carbon.

🔥

Cooling Energy

30-40% of datacenter power goes to thermal management.

40%
💧

Water Waste

Billions of liters consumed annually per facility.

5-10 L/kWh
⚡

AI Density

NVIDIA H100 racks demand 40+ kW. Air fails.

40 kW/rack
🌍

Carbon Crisis

Datacenter emissions exceed many countries.

200+ Mt/yr
The Innovation

Surrender to Thermodynamics

DeepData eliminates mechanical cooling by making the environment the heatsink.

★ CORE CONCEPT
🌊

Submersible

Shipyard-built hulls submerged in lakes. Water conducts heat from metallic walls.

Passive heat sink fins on hull
Year-round 4-15°C water
Tow-to-site deployment
DeepCore nuclear synergy
Concept Architecture

Wall-Integrated Design

Racks embedded in hull walls. Heat transfers directly to surrounding lake water via conductive fins.

DeepData concept sketch showing submersible hull in lake water — cross-section and longitudinal views
Cross-Section & Longitudinal Views Technicians walk inside the central corridor (~1.2m). Server racks surround them on all walls, ceiling, and floor — each thermally coupled to the hull and external lake water via conductive fins.
🌊 Lake water passive cooling 🔲 60cm rack depth 👷 Human-scale corridor 📐 Scalable 6m/12m modules
🔩

Heat Sink Fins

Exterior lamellae maximize passive heat rejection to water/air.

💨

Future: Active Flow

Low-energy water circulation for peak loads (R&D phase).

Thermal Engineering

Head-to-Head

DeepData vs. traditional datacenter performance.

Performance Comparison
Traditional
DeepData
PUE (Power Usage Effectiveness)
1.58
1.58
1.03
1.03
Energy Used for Cooling
35%
35%
3%
Water Consumption
7 L/kWh
7 L/kWh
0 L/kWh
Environmental Impact

Decarbonizing Compute

Eliminating mechanical cooling slashes emissions.

Metric Traditional DeepData
CO₂/MW/year ~3,500 t ~875 t
Energy overhead 58% 3%
Water use 5-10 L/kWh 0
Refrigerants High GWP None
2,600+
tonnes CO₂ saved per MW/year

= 550 cars off roads, or 43,000 trees planted

Financial Model

The Business Case

Higher upfront. Dramatically lower opex. Math wins.

10-Year TCO per MW
Traditional
DeepData
Yr 1
$8M / $10M
Yr 3
$13M / $11.5M
Yr 5
$18M / $13M
Yr 10
$35M / $20M
43%
TCO savings over 10 years
Ecosystem

DeepData + DeepCore

Same artificial lake powers both: nuclear energy and datacenter cooling.

☢️

DeepCore

SMR Nuclear Plant

↓
💧

Artificial Lake

Shared thermal reservoir

↓
🖥️

DeepData

Submersible Datacenter

⚡

Carbon-Free Power

24/7 nuclear baseload with zero emissions.

🏗️

Shared Infrastructure

Security, grid, roads — amortized across both.

Modularity

Ship More. Scale More.

Each hull is self-contained. Scale by towing more units.

Hull

200U
150 kW
×

Cluster

10
1.5 MW
×

Mega

100
15 MW
🔧

Module Independence

Each hull autonomous. Failures don't cascade.

🚢

Relocatable

Tow hulls to different lakes. Follow demand.

📦

Factory Built

Shipyard lines. Consistent. Fast.

Deployment

Any Lake. Any Climate.

Natural lakes, reservoirs, or purpose-built basins.

🇳🇴

Norwegian Fjords

Natural • Deep Cold
4-10°C
🇫🇮

Finnish Lakeland

Natural • 188k lakes
4-15°C
🇨🇦

Canadian Shield

Natural • Glacial
4-12°C
🇨🇭

Alpine Reservoirs

Artificial • Hydro
6-14°C
🌐

DeepCore Lakes

Artificial • Nuclear
Engineered
🇮🇸

Iceland

Natural • 100% green
5-10°C
Next Steps

Submerge the Future

DeepData seeks partners for pilot deployment, shipyard collaboration, and investment.

"In water environments, datacenters should not fight thermodynamics—they should surrender to it. Let the lake do the cooling."

— DeepData Core Principle

1 / 11