🚀 Space vs. Sea: Where Should We Build the Next Generation of Data Centres?
Jeff Bezos recently made a bold prediction: within the next 10–20 years, the world’s largest AI data centers will no longer be on Earth. As generative AI models grow into gigawatt-scale training clusters, he argues that their immense power needs would be better served in space, where solar energy flows continuously and land is unlimited.
It’s a thrilling idea, but is space really the best place for our digital infrastructure? Or could the answer lie beneath the waves, where the ocean’s natural cooling power is already transforming data center design?
Let’s unpack both possibilities.
🌌 Data Centres in Space – The Solar Dream
The vision of orbital data centers sounds like science fiction, but its logic is compelling. Space offers one critical advantage that Earth can’t match: unlimited, uninterrupted solar power. There are no nights, no clouds, and no atmosphere to filter the sun’s rays. In theory, gigawatt-scale AI workloads could run 24/7 on clean, constant energy.
Advantages
- ☀️ Constant Solar Energy: Space-based solar panels can harvest sunlight continuously, providing a stable and predictable power source perfect for energy-hungry AI.
- 🌍 No Land Footprint: Orbital data centers wouldn’t compete for valuable land or radiate waste heat into our cities and ecosystems.
- 🌐 Potential Global Reach: When linked with satellite constellations like Starlink, they could deliver high-speed computing access to any point on the globe.
Disadvantages
- 💸 Extreme Cost: Launching hardware into orbit is prohibitively expensive. Operating and maintaining servers in such a hostile environment would add astronomical costs.
- 🧑🔧 A Maintenance Nightmare: You can’t just send a technician to replace a failed hard drive or power supply. Repairs would require complex, costly, and slow robotic missions.
- 🔥 The Heat Dissipation Problem: This is the biggest hurdle. In the vacuum of space, there is no air or water to carry heat away through convection. The only way to cool equipment is through thermal radiation, which is far less efficient. Managing the immense waste heat from thousands of servers would be a monumental engineering challenge.
- 🛰️ Data Latency: Even in low-Earth orbit, the thousands of kilometers a signal must travel create a noticeable delay. For real-time applications like financial trading or cloud gaming, this latency is unacceptable.
- ☄️ Space Debris Risks: The growing cloud of space debris and the threat of micrometeoroids pose a constant danger. A single high-velocity impact could destroy critical infrastructure in an instant.
Despite these challenges, Bezos’ argument taps into a long-term vision: offloading heavy industry from Earth to preserve our planet while leveraging the unique resources of space.
🌊 Data Centres Under the Sea – The Cooling Revolution
While space remains a distant dream, undersea data centers are already a proven reality. In 2020, Microsoft successfully concluded Project Natick, a two-year experiment that submerged a data center off the coast of Scotland. The result? It ran flawlessly with near-perfect reliability. The concept is simple yet brilliant: use the ocean’s natural, deep cold to cool servers efficiently, slashing energy use and boosting reliability.
Advantages
- ❄️ Efficient Natural Cooling: Seawater is an excellent heat sink. It absorbs server heat far more effectively than air, potentially cutting the energy required for cooling, one of a data center’s biggest expenses, by up to 90%.
- ⚡ Proximity to Renewable Energy: Coastal sites are ideal for pairing with offshore wind, tidal, or wave power generation, creating a truly sustainable energy ecosystem.
- 🌍 Less Land Impact: Building on the seafloor frees up land and eliminates the need for vast surface facilities and noisy, water-intensive cooling towers.
- 🧱 Proven Reliability: The Project Natick container was filled with inert nitrogen and sealed, eliminating corrosion and human interference. This controlled environment resulted in a server failure rate that was one-eighth of its land-based equivalent.
Disadvantages
- 🧰 Maintenance Limits: Like in space, physical access is difficult. However, the current model involves a modular design where entire containers can be retrieved, serviced, and redeployed, which is far more practical than an orbital repair mission.
- 🐠 Environmental Care Needed: Deployment must be done with extreme care to avoid disrupting sensitive marine ecosystems. Site selection and ongoing monitoring are crucial.
- 💰 Upfront Cost: The specialized hardware and marine deployment are expensive, but these costs are still significantly lower than launching infrastructure into space.
⚖️ Verdict: Sea Now, Space Later
Both frontiers offer compelling solutions to the data center energy crisis. Yet, when weighed on practicality, undersea data centers are the clear winner for the short to medium term. They are:
- Technically feasible today, with proven success from major industry players.
- Cost-effective and environmentally efficient, leveraging natural cooling to dramatically reduce power consumption.
- Naturally aligned with existing global subsea fiber optic cables and emerging offshore renewable energy grids.
Space-based data centers, though visionary, remain decades away. They are fundamentally limited by the physics of heat dissipation, the dangers of orbit, and astronomical costs. However, as AI systems continue to scale exponentially and our planet’s energy and land resources become more strained, Bezos’ prediction could still hold true in the long arc of human innovation.
We may one day look back and see the ocean as a crucial stepping stone, a liquid rehearsal for humanity’s eventual leap into space-based computing.
🌍 Final Thoughts
Whether floating in orbit or submerged in the sea, the next era of data centers will be defined by energy efficiency, autonomy, and sustainability. The AI revolution has ignited an arms race for computing power, but the future may depend less on the processors inside the servers and more on where, and how, we choose to power them.
For now, the sea wins on efficiency, but in the future, space may win on scale.