Data economy

China kicks off the race for underwater (and more sustainable) data centres

The growing demand for data storage and management is driving a parallel search for energy-sustainable solutions: offshore solutions are being developed that combine infrastructure with wind power. But Italia, too, could have a role to play in this arena

 (Adobe Stock)

3' min read

Translated by AI
Versione italiana

3' min read

Translated by AI
Versione italiana

Artificial intelligence is energy-hungry, and meeting that demand in a sustainable way is becoming an environmental as well as a technological challenge. In 2025, AI-dedicated data centres consumed nearly 450 TWh globally, a figure comparable to Germany’s entire electricity demand. According to estimates, this demand could double within five years.

Against this backdrop, China has taken a step that could reshape the sector: in June, it opened the world’s first commercial undersea data centre powered by offshore wind energy.

Loading...

The project, developed by HiCloud Technology in collaboration with the Government of the Lingang Special Area in Shanghai, involves a total investment of 1.6 billion yuan, approximately 223 million dollars. The facility uses seawater for cooling, completely eliminating the need for cooling towers and energy-intensive air conditioning units. The result is a Power Usage Effectiveness (PUE) of less than 1.15, where 1.0 represents perfect theoretical efficiency.

The concept is not entirely new. Microsoft had explored this technology with Project Natick, submerging server capsules off the coast of Scotland between 2018 and 2020. The results had been promising: a lower incidence of failures compared with land-based data centres, thanks to the sealed environment free from corrosive agents. But the project had remained at the experimental stage.

China has stepped up the pace: in less than two and a half years, HiCloud has progressed from a pilot scheme in Hainan to a commercial roll-out, and is now aiming to expand the model on a large scale.

On the other side of the Pacific, several American start-ups are following similar paths. Panthalassa, a start-up backed by Peter Thiel, amongst others, is trialling floating modules powered by wave energy. The estimated cost of these solutions remains high — between 25 and 30 million euros per megawatt — with significant challenges in terms of maintenance and stability in a marine environment.

Aikido Technologies, based in San Francisco, has developed a hybrid system that integrates a wind turbine, energy storage and digital infrastructure into a single floating structure. The data centres are housed in the structure’s submerged legs, at a depth of around 20 metres, utilising seawater for passive cooling. The company is developing a 100 kW prototype in Norway, with commercial deployment planned off the British coast by 2028.

Pressure from demand and growing local opposition to the construction of onshore data centres – as has also been seen recently in the Milan area – are fuelling interest in offshore alternatives.

From this perspective, the Nordic countries’ competitive advantage in this sector is structural. As Fiorenzo Manganiello, founder of Polar Dc, explains, the annual cost per megawatt in Norway is around 300,000 euros, compared with around 1.5 million in Italia. Low average temperatures drastically reduce the need for active cooling, bringing the PUE down to 1.1 compared with the 1.3 typical of southern Europe. Added to this is the abundance of renewable hydroelectric power, a requirement increasingly sought after by AI clients for ESG reasons.

But southern Europe is not out of the running. The Iberian Peninsula is emerging as an alternative hub thanks to competitively priced solar power, comparable to Norwegian hydroelectric power. And Italia, according to Manganiello, has a specific opportunity in the field of AI inference – the phase of model execution that requires low latency. Southern Italia lies at the crossroads of the main undersea cable routes linking Europe, North Africa and the Middle East: regions such as Puglia could host data centres serving the entire Mediterranean.

The model for Italia is a distributed network of small data centres, each with a capacity of 2–5 MW, rather than large centralised facilities. This approach improves resilience, reduces dependence on concentrated infrastructure and could make use of disused industrial sites with existing electricity connections, such as the former Ilva site in Taranto, thereby avoiding long waits for new connections to the national grid. With this in mind, Manganiello is working on a spin-off focused on installing small centres dedicated to AI inference in the south of the peninsula.

The obstacles remain significant: a national electricity grid that is often outdated, authorisation processes that are still slow despite recent improvements, and, above all, the lack of a clear policy strategy on digital infrastructure, unlike in France or the United Kingdom.

But the path forward is clear: whether underwater, on floating platforms or in small, distributed onshore facilities, the future of data centres will be increasingly shaped by the ability to combine computing power with energy sustainability.

Copyright reserved ©
Loading...

Brand connect

Loading...

Newsletter

Notizie e approfondimenti sugli avvenimenti politici, economici e finanziari.

Iscriviti