The Ocean, the Orbit, and the Grid: How AI's Hunger Is Reshaping the Planet's Infrastructure

A Server Farm in the Middle of the Ocean

Picture a data center rising and falling with the swells of the open Pacific — no power grid in sight, no cooling towers, no land. Just waves, satellite uplinks, and servers humming beneath a salt-spray sky.

That's not science fiction. It's the $140 million bet that investors just placed on Panthalassa, a US-based ocean technology company now valued near $1 billion. As Singularity Hub reports, the company's vision is to house server farms far from shore, powered by ocean wave energy, cooled by seawater, and networked by satellite. With tech companies planning to spend roughly $750 billion on data centers in 2026 alone, the push to find new ways to power AI is no longer abstract. It's urgent — and it's getting strange.

Why the Ocean? Why Now?

The math is simple. AI needs computing power. Computing power needs electricity. Electricity is running short.

The floating data center idea, as TechRadar notes, would tap into "tens of terawatts of new capacity potential in the power of the open ocean" — an almost incomprehensibly large reservoir of renewable energy that land-based infrastructure simply cannot access. Yes, the waves are harsh. Yes, corrosion is real. But so is the alternative: a global AI buildout stalling for want of a wall socket.

The pressure on infrastructure is rippling through every corner of the tech economy. In Asia, as Reuters reports, Samsung's first-quarter profit increased eightfold this year, with chips responsible for 94% of a record 57.2 trillion won total. Its stock has more than doubled, crossing the $1 trillion market cap threshold — only the second Asian company to do so after TSMC. Andy Wong of Pictet Asset Management calls the region "a shrimp among whales": compact but indispensable. In certain tech segments, he says, "Asia has the best companies in the world."

That AI bull run needs somewhere to run to. Which is exactly the problem researchers and engineers are racing to solve.

Teaching Machines to Think Faster — and Smarter

Speed is everything when systems grow complex. A satellite attitude control system with discrete actuators — the kind that has to make thousands of calculations just to orient itself in orbit — used to require thousands of seconds of computational planning time. Researchers have now developed a "myopic" AI framework, as detailed in a new paper on arXiv, that teaches controllers to learn from expert demonstrations and dramatically shorten their decision horizon, enabling real-time control. The satellite can now think on its feet.

The same principle is being applied to power grids. A researcher at Delft University is proposing what he calls "Foundation Twins" — a new generation of digital twins for power systems that combine foundation AI models with reinforcement learning. As he outlines in a position paper, power grids operate across wildly different timescales and geographies simultaneously, and no existing digital twin has managed to model that complexity well enough to be genuinely useful. His framework, he argues, could finally close that gap, accelerating decision-making across the entire grid in ways that could make renewable energy integration dramatically more efficient.

If floating data centers are the brute-force answer to AI's energy hunger, Foundation Twins are the elegant one — making the existing grid smarter rather than simply building more of it.

From Infrastructure to the Human Body

Not all of these frontiers are measured in terawatts. Some are measured in heartbeats.

Researchers at The City University of New York have been running a pilot study that combines smartwatches, GPS devices, and real-time health surveys to track how environmental exposures — extreme heat, air pollution — affect people's bodies and moods in the moment. As Medical Xpress reports, the goal is to understand health impacts in real time, not in retrospect, using the same satellite and sensor infrastructure that underpins everything from navigation to AI training data. Climate change is making the need urgent. Wearables may be the way in.

The thread connecting a floating server farm and a smartwatch tracking a New Yorker's heart rate in a heat wave is the same: real-time data, gathered from the physical world, processed at speed, and used to make better decisions.

The Governance Question No One Can Ignore

All of this infrastructure — oceanic, orbital, biological — raises a question that steel and algorithms alone cannot answer: who makes the rules?

Tech Week Shanghai 2026 brought together technology pioneers, enterprise leaders, and public sector stakeholders to grapple with exactly that, positioning Shanghai as a global testbed for AI governance and cross-border data collaboration. The challenge, as PR Newswire reports, is that AI is being deployed across "fragmented regulatory" environments — different countries, different rules, different definitions of what's acceptable.

Meanwhile, Flipkart Foundation is hosting webinars exploring technology for social impact, with speakers from Lenovo Foundation and Sattva Consulting mapping how tech can serve development goals rather than simply shareholder returns. It's a quieter conversation than the billion-dollar ocean bets, but no less important.

The World Being Built Right Now

The frontier of AI in 2026 is not one place. It's a floating platform in the Pacific. It's a satellite adjusting its orientation in milliseconds. It's a power grid in Europe being modeled in real time. It's a researcher in New York watching a smartwatch measure a person's stress response to a summer heatwave.

What these stories share is a refusal to accept the limits of existing infrastructure — physical, computational, institutional — as permanent. The solutions are sometimes audacious, sometimes incremental, always interconnected. And they are being built right now, in the open ocean and the open source, for a world that is going to need all of them.

The question for the rest of us is not whether this transformation is coming. It's whether the rules, the benefits, and the risks will be shared as widely as the satellites that make it all talk to each other.