The Untapped Energy Source Cheaper Than Solar and Natural Gas
TL;DR
Panthalassa wants to turn the deep ocean into a power plant — CEO Gar Sheldon Coulson says the company’s “nodes” harvest wave energy far from shore, where a red band of high wind and wave intensity covers roughly 25% of Earth’s surface, especially in the southern hemisphere.
The core bet is simplicity, not clever mechanics — after years of failed wave-energy attempts near shore, Panthalassa converged on a hollow steel, hydro-dam-like device with essentially one moving part—the turbine—because flappers, articulating joints, and coastal systems kept breaking or costing too much.
Their first real market is AI inference, not the grid — instead of laying subsea cables, Panthalassa wants to put compute directly inside the node, using wave power plus ocean cooling, and send results back over Starlink/Kuiper where low data-rate, high-energy workloads like inference make the economics work.
The cost target is aggressive: 2.5¢/kWh, maybe 1.5¢ later — Coulson claims commercial-scale nodes could undercut solar and natural gas in many places while avoiding major data-center costs like land, chillers, and skilled on-site construction labor.
The timeline is ambitious enough to sound absurd — Panthalassa plans an Ocean 3 mini-fleet off Washington/Oregon this year, larger commercial nodes by late 2027, and Coulson argues that if manufacturing works, this could be the fastest energy technology ever to reach a gigawatt—faster than solar, gas turbines, or nuclear.
This is also a geopolitical manufacturing story — Coulson says the system is mostly “steel and paint,” with far less dependence on Chinese-heavy supply chains than solar, and he sees Australia, New Zealand, or Chile as likely southern-hemisphere factory sites close to the best wave resource.
The Breakdown
The pitch: a giant floating machine that drinks wave power
Ashley Vance opens by framing Panthalassa as a quasi-stealth hard-tech company staffed with SpaceX-style engineers who’ve spent a decade quietly building ocean energy systems in Portland. Gar Sheldon Coulson immediately widens the lens: they’re not just building a machine, but a full system for harvesting energy from the middle of the ocean and either bringing it home or using it on-site.
Why almost every old wave-energy idea went wrong
Coulson draws a sharp contrast with the graveyard of coastal wave and tidal startups Ashley remembers from the dot-com era. The problem, he says, was too many moving parts near shore, too little energy, too much infrastructure—so Panthalassa went in the opposite direction: deep ocean, minimal mechanics, and a hydro-dam-like water cycle inside a hollow “lollipop” structure roughly 20 meters across and 80 meters long.
The first-principles origin story: only three big energy sources matter
Coulson says he and co-founder Brian Moffett came at the problem by asking where humanity could actually get tens of terawatts of new energy. Their conclusion: only solar, nuclear, and the middle of the ocean make the cut, and they were stunned that almost nobody was trying to build a deep-ocean tech stack to access that third category.
YouTube, Chick-fil-A, and the garage-lab years
The company started in a way that feels very internet-era hard tech: Coulson found Moffett through a YouTube talk on wave energy, flew to LA, had a four-hour Chick-fil-A meeting, and both agreed to quit their jobs. What followed was years of garage prototyping near Moorpark, early sea trials with a tube-and-turbine device dangled off workboats, and lots of what Coulson describes as “depressive long walks” imagining contraptions failing in huge seas.
The weirdest part: the nodes propel themselves like bunny rabbits
One of the most memorable sections is Coulson explaining that the vertical nodes can be towed out horizontally, flipped upright, and then self-propel to their destination using asymmetry in their shape. Ashley is visibly delighted by the image: the devices move at about 1 to 1.5 knots, “sort of like a bunny rabbit bouncing along,” eventually roaming 50, 200, even 1,000 miles to sit in high-energy waters.
Why the first serious business is floating AI infrastructure
Instead of trying to solve grid interconnection first, Panthalassa now wants to build ocean data centers. Each node produces roughly 200 kilowatts to 1 megawatt depending on size, and Coulson argues the best fit is AI inference—where the energy per bit of input/output is huge, but the actual data moving over satellite is tiny compared with model training.
The economics: cheap steel, free cooling, no giant buildout
Coulson’s case is that a node is fundamentally a low-precision hollow steel object with a small turbine, generator, and electronics—order-of-magnitude around $1 million per unit. That leads to claimed energy costs of 2.5 cents per kilowatt-hour at scale, maybe 1.5 cents eventually, plus another advantage Ashley keeps circling back to: the ocean is a giant cold heat sink, so the system avoids expensive land, buildings, chillers, and armies of electricians and plumbers.
Where they are now, and why they’re finally talking
Panthalassa is developing two things in parallel: an Ocean 3 mini-fleet of sub-10-meter nodes for deployment off Washington and Oregon this year, and the factory system to mass-produce larger commercial versions by late 2027. Coulson says if manufacturing ramps the way they expect, Panthalassa could reach a gigawatt faster than any energy technology in history, and that’s why the company is stepping out of the shadows now—with commercial partnerships, hiring, and likely future factory discussions in places like Australia, New Zealand, or Chile.