Atomarine, a Cambridge, Massachusetts startup, has raised $19 million in seed funding to develop floating data centers powered by nuclear reactors on ships, DCD reports. Y Combinator led the round. Pantera Capital, Soma Capital, Seven Seven Six, Liquid 2 Ventures, Audacious Ventures and E14 Fund also took part.
The company was founded by MIT students Dimitris Koutentakis and Emile Germonpré. Its plan is to put small modular reactors (SMRs) on vessels that supply power to platforms housing modular data centers. At full build-out, Atomarine expects a nuclear ship to deliver 150MW, which it says would be enough for two platforms.
The nuclear element depends on hardware that does not yet exist commercially. According to the report, no SMR vendor has released a commercial model or won regulatory approval in the US or any other market. Atomarine therefore intends to start with natural gas generators mounted on its ships.
Koutentakis told DCD earlier this year that the company is talking with potential partners about access to an SMR. He cited the minor engineering changes required and the regulatory timelines, and said the target for a first deployment is 2031 or 2032.
In a LinkedIn post after the announcement, Koutentakis argued that moving offshore means the company does not compete with communities for water, power or land, and does not push up local electricity bills or rents. He also said the approach would let the US stay competitive in AI while scaling compute on nuclear rather than fossil fuels. He said the company has already built a functional proof of concept and is working toward its next milestone.
The seed money is going partly toward hiring. Atomarine is recruiting engineers with backgrounds in naval architecture, nuclear engineering or data center design.
DCD notes that Atomarine is one of many companies in the US and elsewhere pursuing AI compute at sea, a segment it covered in issue 62 of its magazine.
Why it matters: Siting compute offshore is one proposed answer to local limits on power, water and land, but Atomarine's nuclear case rests on reactors that are not yet commercially available or approved. The gas-first path and a 2031 or 2032 target suggest the nuclear claims should be treated as long-dated, and the company's proof of concept has yet to be shown at commercial scale.
TwinStack automated analysis, based on reporting by DCD. Check the original for full details.