Samsung plans a floating AI data centre by 2028



As data centres run into unhappy neighbours and water limits on land, Samsung wants to float them offshore. Samsung Heavy Industries plans to launch its first floating data centre by 2028, Seoul Economic Daily reported. It would be a purpose-built barge parked near the coast.

The design is specific. Rather than convert an old ship, Samsung is building a new 50MW barge with a server hall, onboard power generation and liquefied natural gas fuel tanks. The first version is “nearshore”, sitting close to land and drawing some power from the grid, a hedged first step before anything ventures far out to sea.

Why put a data centre on water

The pitch answers a growing problem on land. Data centres need vast plots, cheap power and huge volumes of water for cooling, and communities increasingly refuse to give up all three. Floating offshore sidesteps land shortages and slow planning fights, while the sea offers ready cooling.

The economics stay unproven. Saltwater corrodes, storms threaten, and running fibre and power to a barge adds cost and risk. Samsung is betting the trade is worth it as AI demand outruns what many grids and towns will bear.

A shipbuilder’s pivot

The move also reflects where shipbuilding is heading. Samsung’s yards are hungry for new work. “Data centres on the sea” turn spare hull-building capacity into AI infrastructure. “Floating datacentres represent a major new opportunity for the shipbuilding and offshore industries,” said Samsung Heavy Industries chief executive Sung-an Choi.

Rivals are circling the same idea. A Japanese pair, Mitsui OSK Lines and Hitachi, is fitting data-centre kit onto existing ships, aiming to go live in 2027. China has gone further and put a data centre under the sea.

Samsung is not going it alone either. At the Posidonia maritime fair it signed partners including Greece’s Capital Clean Energy Carriers and the classification society Lloyd’s Register. It also struck a deal with Supermicro to test AI servers at sea, The Register reported. The American Bureau of Shipping and Lloyd’s Register have granted approval in principle.

Why it matters

Floating data centres remain a bet, not a fact. But a shipbuilder drawing up plans shows how far the industry will go to escape the backlash on land. If neighbours will not host AI’s power-hungry sheds, the next stop may be just off the coast.



Source link

Leave a Reply

Subscribe to Our Newsletter

Get our latest articles delivered straight to your inbox. No spam, we promise.

Recent Reviews


YouTube has an AI slop problem, and its crackdown is catching legitimate creators in the crossfire. Faceless channels, where no human host ever appears on screen, have existed for years and are not inherently AI-generated.

Many are run by solo creators who simply prefer to stay anonymous. The problem is that AI tools made it easy to flood the platform with low-effort faceless content at scale, and YouTube’s algorithm is now penalizing the format as a whole.

How bad is the AI slop problem on YouTube?

A Kapwing study found that roughly 21% of the first 500 videos recommended to a new YouTube account were classified as AI slop, while 33% fell into a broader brainrot category. The problem extends to children, too, as more than 40% of YouTube Shorts recommended to kids in a 15-minute session contained low-quality AI content.

YouTube’s response has been to tweak its algorithm to favor videos with real human faces on camera, which is hitting faceless creators even when their content is entirely human-made.

How is YouTube tackling its AI slop problem?

YouTube is now testing a new pop-up on mobile that asks viewers to rate whether a video feels like AI slop, on a scale from “not at all” to “extremely.” The idea sounds reasonable, but crowdsourcing AI detection has real problems. People are bad at spotting AI content, and they are getting worse at it as AI capabilities continue to improve.

There are also legitimate concerns that YouTube could use this viewer feedback as training data for its own AI models, potentially making future AI-generated content even harder to spot.

🚨 Did you just see what YouTube did?

YouTube isn’t banning AI slop.. They’re making you label it so they can train their next model to not look like slop.

Read that again…

You flag the bad AI content. YouTube collects it. Google feeds it into Veo 4… Then next year their… https://t.co/8UC2J3mjjv pic.twitter.com/mIrTChqC1b

— Tuki (@TukiFromKL) March 17, 2026

Meanwhile, faceless creators are scrambling to adapt. According to The Hollywood Reporter, some are hiring cheap on-camera hosts through platforms like Fiverr and Upwork. Others are doubling down on niche educational content, which has held up better than broad content farms.

The AI text-to-video space is still valued at enormous sums, with Higgsfield AI alone sitting at $1 billion, but on YouTube, the math for faceless creators is getting harder to work out every month.



Source link