Panthalassa raises $140M offshore AI

AI infrastructure has a power problem. The easy sites are getting taken, grid hookups take years, and cooling big clusters burns through water and money. Panthalassa’s pitch is to stop fighting for land altogether and move some of the problem into the ocean. On May 4, the company said it raised $140 million in a Series B led by Peter Thiel to build floating, wave-powered compute nodes for AI workloads. (prnewswire.com) ### What is Panthalassa actually building? Not a conventional offshore wind farm, and not a ship full of servers. Panthalassa is building autonomous offshore platforms — its Ocean-3 series — that harvest energy from wave motion, use that power on site, and run modular AI compute without plugging into the terrestrial gr(prnewswire.com)wer back to shore if you can run the compute at the source instead. (panthalassa.com) ### Why put compute in the ocean? Because the bottlenecks on land are getting ugly. New AI data centers need huge power loads, lots of cooling, and long transmission and permitting timelines. Panthalassa’s model tries to sidestep that by using open-ocean wave energy, seawater cooling, and offshore space that is effectively unconstrained compared with industrial land near major grids. Basically, it(panthalassa.com)ce — electricity, cooling capacity, and buildable sites. (prnewswire.com) ### How does the wave-power part work? The company says wave motion forces water through an internal turbine as the platform rises and falls. That electricity then powers onboard systems and compute hardware. The clever bit is not just generation — it is colocation. If you generate offshore and compute offshore, you av(prnewswire.com)ter has to survive the open ocean like a piece of marine equipment, not a building. (dataconomy.com) ### What changed this week? Money, mostly — but meaningful money. Panthalassa said the $140 million round will fund its pilot manufacturing facility near Portland and support first deployments. The investor list is also a signal. Alongside Thiel, the round included names like John Doerr, TIME Ventures, SciFi Ventures, Hanwha Group, Fo(dataconomy.com)infrastructure play. (prnewswire.com) ### When is this supposed to become real? The company says 2026 for initial Ocean-3 deployments in the northern Pacific. That matters because this is no longer being pitched as distant science-project hardware. Panthalassa is moving from prototypes toward pilot manufacturing and first fielded systems. But “pilot” is the important word — nobody should confuse an early deployment with proven hyperscale economics. (convergedigest.com) ### What’s the hard part? Marine reliability. Saltwater, corrosion, storms, maintenance access, satellite connectivity, and hardware replacement all get harder offshore. AI servers are not usually designed to live inside autonomous ocean platforms. So even if the energy math works, Panthalassa still has to prove uptime, se(convergedigest.com)ssel into one machine — every subsystem now inherits the others’ failure modes. (geekwire.com) ### Why does this matter beyond one startup? Because AI’s next constraint may be infrastructure, not chips. If land-based buildouts keep running into power and cooling limits, capital will keep chasing weird-sounding alternatives that would have seemed fringe a few years ago. Floating compute may fail. But (geekwire.com)out of slack. (prnewswire.com) ### Bottom line Panthalassa has not solved offshore AI yet. What it has done is turn a wild-sounding idea into a funded test of whether compute can follow energy into the open ocean — and whether that is cheaper than dragging more power, water, and land into the AI buildout onshore. (prnewswire.com)