The race to power artificial intelligence is heating up, and China is diving headfirst—literally—into a bold new frontier: underwater data centers. In an ambitious effort to tackle both energy demands and environmental challenges, China is building offshore facilities that will use seawater for cooling and draw 97% of their power from wind energy.

One of the most eye-catching projects in this initiative is a $223 million facility being constructed six miles off the coast of Shanghai. Scheduled to launch in September, the center is designed to house up to 792 high-performance AI-capable servers. The goal? To create a sustainable yet powerful hub capable of training advanced AI models—think GPT-3.5—in as little as a single day.

Cooling With the Ocean Itself

Cooling is one of the biggest cost and efficiency headaches for data centers worldwide. Traditional land-based facilities rely on massive air-conditioning systems, which eat up enormous amounts of electricity. But under the sea, things work differently.

By placing the servers in a sealed, pressure-resistant chamber on the ocean floor, engineers can use the naturally cold seawater to dissipate heat. This reduces the need for energy-intensive cooling infrastructure and cuts electricity consumption by an estimated 30% compared to land-based centers.

The constant, stable temperatures underwater not only make cooling easier but also help prolong the lifespan of the hardware. Less wear and tear means lower maintenance costs and fewer downtime risks—a big win in the high-stakes world of AI computing.

Mostly Green Energy, Massive Computing Power

China says these underwater data centers will run on 97% renewable wind power, harnessed from offshore wind farms. Combined with the cooling benefits of the ocean, the facilities promise to be both greener and more cost-effective than many of today’s massive inland server farms.

And they won’t just be green—they’ll be fast. With the right configuration, these servers could train massive language models like GPT-3.5 in just 24 hours, compared to the days or even weeks required by older, less efficient data centers.

This combination of speed and sustainability could make underwater facilities a key player in the AI boom—especially as nations race to secure computing capacity for everything from advanced language models to complex climate simulations.

Not the First Attempt, But a Bigger Push

This isn’t the first time the world has flirted with the idea of putting data centers underwater. Back in 2018, Microsoft tested the concept with its Project Natick—a capsule-shaped data center deployed off the coast of Scotland. The project ran for two years and showed promising results: the underwater servers were eight times more reliable than their land-based counterparts.

But despite the success, Microsoft shelved the idea, citing scaling challenges and unclear commercial benefits.

China, however, seems determined to push forward where others paused. With AI demand surging and energy efficiency becoming an urgent concern, the country is scaling the concept aggressively.

Environmental Concerns Still Linger

Not everyone is cheering this development. Environmental experts warn that while using renewable energy is a step in the right direction, placing large amounts of technology underwater comes with potential risks.

There are concerns about the impact of heat discharge into marine ecosystems, possible contamination if a unit leaks, and the logistical challenges of repairing equipment deep under the sea. While these facilities are designed to be sealed and low-impact, large-scale deployment could bring unforeseen consequences.

China has said it is conducting environmental assessments and will monitor the impact closely. Still, the tension between technological progress and ecological responsibility remains unresolved.

A Regional Trend Emerges

China isn’t alone in exploring this approach. South Korea and Japan are already researching similar offshore solutions, driven by the same factors: high energy costs, limited land space, and an urgent need for sustainable cooling methods.

If these projects succeed, we could see a regional shift toward underwater data centers in the Asia-Pacific, potentially setting a global example for how to expand AI capacity without pushing the planet’s energy systems to the brink.

The Bottom Line

Underwater data centers might sound like something out of a sci-fi movie, but they’re quickly becoming a real contender in the fight to balance computing power with environmental responsibility.

China’s $223 million facility off Shanghai is more than just a technical experiment—it’s a statement. A declaration that the future of AI infrastructure may lie not in the clouds above us, but in the seas below.

Whether this vision proves to be a sustainable breakthrough or just a bold experiment remains to be seen. But one thing’s certain: in the race to power the next generation of AI, the world is now looking beneath the waves.

Thanks for reading!