Fifteen years ago, a crop of startups promised to make wave energy a mainstream renewable resource. None succeeded, and today the technology remains niche. But now, as deep-pocketed investors seek novel ways to power data centers, the idea is getting renewed attention.
The basic technology is nothing new. The first tide wheel was built back in 1844, and ever since inventors have been iterating, trying to harness waves. They are energy-dense, predictable, and estimated to hold the potential to generate up to as much as 2.64 trillion kilowatt-hours just off the coast of the U.S., according to the EIA.
But the hurdles are many: Developing technologies capable of surviving the harsh ocean conditions and the infrastructure needed to transmit the energy produced have historically driven costs too high for wave energy to compete with cheaper forms of clean power.
Many of the wave energy startups that have been founded since the 1990s have since ceased operations or pivoted away from energy production, mostly because of a lack of funding, and because of the fact that the ocean is a difficult place in which to put expensive tech. Just last year, for example, the Finnish company AW-Energy Oy, which was developing an oscillating wave surge converter to be anchored near the shore, filed for bankruptcy; AquaHarmonics, a U.S. company developing an offshore buoy system, also ceased operations in 2025.
So when wave energy startup Panthalassa announced it had raised a $140-million Series B earlier this month — in a round led by Peter Thiel with participation from John Doerr and Marc Benioff’s Time Ventures, among others — it turned heads for a few reasons. First, it’s a large raise for a startup in a niche and technologically complicated field.
Also, the pitch is very much primed for 2026 and the data center power push: putting artificial intelligence data centers on its large, offshore, buoy-like devices and letting them float, unmoored in the deep ocean. It’s a twist on wave energy that adds a science-fiction bent, as well as logistical and technical complexity.
But regardless, Bryson Robertson, director of the Institute for Integrated Energy Systems at the University of Victoria and former director of the Pacific Marine Energy Center at Oregon State University, said that the round gives wave energy something it’s never had before: the time and space to iterate. While official numbers are hard to come by, the entire global market was estimated to have a value of approximately $1 billion in 2025, according to a report by intelligence provider the Business Research Company.
“Of all the renewables, wave energy has never had the level of investment that others have had, so [building] scale has been very hard,” Robertson said, noting that most companies have only had the chance to work with one or two prototypes. “If they fail with any prototype, they fundamentally fail as a company, because they don’t have the capital bankrolling to allow them to build one, fail, learn some lessons, build another one, fail, learn some lessons. They just haven’t had the financial vacuum to do that iterative loop effectively.”
What’s changed, of course, is the data center use case. The concept, which is still at the pilot stage, has lured a who’s who of Big Tech, who are increasingly funding even unproven clean firm power options for AI, like fusion.
The list includes Mike Schroepfer, the founder of Gigascale Capital and a returning investor in Panthalassa’s latest round, and former chief technology officer at Meta. “I scaled gigawatts of data center infrastructure at Meta, and I know the constraints on building the physical systems,” he told Latitude Media via email. “An additive energy source co-located with effectively unlimited cold seawater, that doesn’t require waiting five to seven years for a grid hookup, is uniquely fit for this moment.”
He added that “compute demand may be the catalyst that helps to scale this new energy technology.”
‘Go where the energy is’
The fact that Panthalassa’s compute capacity could bypass grid constraints entirely is an integral part of its pitch. The company was incorporated in 2016, the same week AlphaGo defeated Lee Sedol at the board game Go, a milestone widely seen as an AI turning point. Garth Sheldon-Coulson, the company’s co-founder and CEO, and his co-founder Brian Moffat realized at that point that if the exponential growth in computing demand continued, “the grid would likely not be able to meet demand by around 2030,” he told Latitude Media.
“So in 2017, we filed patents on wave-energy-powered computing clusters at sea and self-propelled, self-organizing computing grids made up of these converters,” he said. “We summarized the approach in a mantra: ‘Go where the energy is.'”
To do that, Panthalassa builds what it calls “nodes” — self-propelled buoys, made of a hollow tube extending 50 to 80 meters vertically down into the water, and a top section that’s 15 to 30 meters across sitting at the surface. Inside, there’s a small water turbine. “It looks a bit like an upright lollipop,” Sheldon-Coulson said.
The mechanics resemble those of a self-filling hydroelectric dam. As the node bobs up and down in the waves, it forces water into a pressurized reservoir, where it spins a turbine, generating electricity. That electricity is then used to power AI chips, which are located in a sealed enclosure and cooled by the ocean water.
A single node can generate up to a megawatt of power, which is used to process AI workloads. Data is then transmitted via low-Earth-orbit satellites, such as Starlink.
Due to their shape, Sheldon-Coulson says these nodes are able to mostly maintain their position, and they are equipped with a self-navigation system in case they stray off-course. “The best deployment zones are far offshore, where waves are powerful and consistent,” he said. “Our Ocean-3 pilot series, like Ocean-2 and Ocean-1 before it, will deploy in the North Pacific off Oregon and Washington. Future fleets will primarily operate in the Southern Hemisphere.” Ocean-2 and Ocean-1 are the two prototypes the company deployed in 2021 and 2024, which focused on the energy production mechanics and did not include data center components.
The nodes are designed to sidestep some of the problems that have historically defeated offshore wave energy technologies, the biggest one being transmission. Getting electricity from the open ocean back to shore requires expensive undersea cables, which Robertson estimates run in the hundreds of millions of dollars for just a handful of miles. By putting the chips directly on the floating device, Panthalassa doesn’t need to transmit any power, just compute. The company is also exploring using the nodes for synthetic renewable fuels production, which would be collected by ship and delivered to shore.
As for the hazards of having delicate technologies floating around in the sea? Well, Robertson pointed out that we have been building complex machines to traverse the ocean for a very long time. “We built millions of ships that survive the worst ocean conditions,” he said. “So can you build the structure that can survive in the ocean wherever they want to deploy them? 100%.”
Perhaps the more pressing question, however, is whether Panthalassa can make its technology cost-effectively: a hurdle other companies have so far failed to clear.
On this point, Sheldon-Coulson is optimistic. Panthalassa estimates each full-scale node costs about $1 million to $1.5 million to manufacture, which is below the $10.7 million per MW average data center cost in 2025, according to JLL. Panthalassa’s estimate doesn’t include logistics and maintenance.
‘Energy infrastructure is not software’
Skeptics remain. Climate tech consultant Michael Barnard, who’s known for his occasionally scathing critiques of hyped companies and energy technologies, wrote in Clean Technica that including data centers in wave energy does not automatically make the latter easier.
“Sadly, the ocean does not care about clean narratives,” he wrote, adding that the nodes are not a single technology but rather “a stack of hard technologies and harder operations,” including “wave-energy conversion, floating offshore structure design, autonomous marine stationkeeping, and corrosion control,” among others.
Still, Sheldon-Coulson maintains that the company is planning for streamlined maintenance, with coatings designed to last at least 15 years, very occasional repair needed, and systems that can drive themselves home when needed. Whether that plan holds up if and when the company manages to deploy many of them, however, remains to be seen.
As streamlined as the operations can get, however, the logistical complexity remains. So the Panthalassa approach might not be worth it if you’re a country like the U.S. or Australia, where there’s plenty of land for commercially proven solar and wind, which are increasingly being deployed with storage — at least not in the short term.
“We can’t ignore the fact that the ocean has a massive renewable energy flow,” Robertson said. And if and when siting renewables on land becomes even more challenging, especially in smaller countries with less land to spare, we will have to look for other places to extract clean energy, he added. Given the demands of load growth related to the AI boom, that day may come sooner than later.
But even in the best case scenario of Panthalassa “using AI software money to do energy projects,” Robertson said, it’s still going to take time. “Energy infrastructure is not software, so if you’re expecting a disruptive solution within months that’s going to upend the way we do things, that’s not how it works,” he said.
