Don't understand why they need fresh water when sea water works better.
Grok.
**Yes, the sea can effectively cool data centers using heat exchangers, and this approach is already in use in several real-world deployments.** It leverages the ocean's vast thermal capacity as a natural heat sink, often reducing energy use for cooling (which can account for 30-40%+ of data center power) and eliminating or minimizing freshwater consumption.
### How It Works
Seawater cooling typically uses **closed-loop heat exchangers** to avoid direct contact between corrosive saltwater and IT equipment:
- **Cold seawater** is pumped from the ocean (often from deeper, cooler layers) through sealed pipes or heat exchangers.
- Heat from the data center's internal cooling fluid (or air) transfers to the seawater via the exchangers.
- Warmed seawater is discharged back to the sea at a controlled temperature (e.g., with limits like no more than a 3°C rise in some regulations) to minimize environmental impact.
**Key methods include**:
- **Wet Surface Air Cooling (WSAC)**: Seawater is sprayed over closed pipes carrying facility coolant; salt buildup is managed with periodic fresh-water flushes. Google uses this at its Hamina, Finland facility on the Gulf of Finland.
- **Closed-loop direct or air-exchanger cooling**: Seawater circulates in sealed systems for direct-to-chip liquid cooling or to chill air. Microsoft's Project Natick tested underwater server pods using ambient seawater.
- **Deep Seawater Cooling (DSWC)**: Pipes draw from cold deep ocean layers (e.g., ~2,770m depth at stable ~18°C). Studies (e.g., IIT Bombay) show up to **79% energy savings** vs. traditional chillers for a 100 MW facility, with payback in ~8 months.
### Real-World Examples
- **Google Hamina (Finland)**: Repurposed a paper mill with seawater cooling via tunnels; achieves very low Power Usage Effectiveness (PUE ~1.09) without mechanical chillers.
- **Start Campus Sines (Portugal)**: Uses massive seawater flows (~1,000 m³/min) from a deep ocean basin in a closed circuit with titanium heat exchangers; operational since ~2024 for AI workloads.
- **China's underwater projects**: Wind-powered submerged pods off Shanghai (24 MW) and others use seawater directly as a passive coolant, cutting cooling energy dramatically (PUE ~1.15) and eliminating freshwater/land use.
- **Floating/offshore concepts**: Projects like Samsung's planned floating data centers (target 2028) and others use seawater in closed loops.
Similar lake-based systems (e.g., Toronto's Deep Lake Water Cooling) demonstrate the principle.
### Benefits
- **Energy efficiency**: Ocean water's high thermal capacity (thousands of times that of air) enables "free cooling" with minimal or no chillers.
- **Water savings**: Near-zero freshwater use, addressing scarcity issues in data center growth.
- **Sustainability**: Lower carbon emissions; integrates well with renewables (e.g., offshore wind).
- **Scalability for AI**: Supports high-density racks where air cooling struggles.
### Challenges and Considerations
- **Corrosion and biofouling**: Salt, marine life, and sediments require corrosion-resistant materials (e.g., titanium, HDPE pipes), filtration, and maintenance.
- **Environmental impact**: Discharge temperature must be regulated to protect marine life; permitting and environmental assessments are key hurdles.
- **Infrastructure**: Pipelines, pumps, and coastal access add upfront costs and engineering complexity (e.g., insulation for long pipes).
- **Location limits**: Best for coastal or offshore sites; not universal.
- **Regulatory/operational**: Maintenance is harder underwater/floating; local ecology and permitting can delay projects.
Overall, seawater cooling via heat exchangers is proven, viable, and increasingly attractive amid AI-driven demand and water constraints. It's not a one-size-fits-all solution but excels where geography allows. Ongoing innovations (e.g., hybrid systems, better materials) are expanding its use.