AI reinforcement learning energy demand probably already exceeds traditional pre-training
Garth Sheldon-Coulson · Building inference data centers on the high seas · May 28, 2026 · at 43:57
Garth Sheldon-Coulson, CEO of ocean compute startup Panthalassa, compares the energy requirements of reinforcement learning versus traditional model pre-training on Catalyst.
0:00 / 0:12exact quote · 12.1s
720p mp4 · rendered on demand · StarZero watermark
“And this is an area that is becoming huge in terms of energy demand. It'll, it will, The probably already is bigger than what we have historically considered training, you know, pre-training”
quote is from the automated transcript, cleaned for reading: filler sounds and stutters are removed, nothing is rephrased. names can be misheard (the analysis reads context, assessments check outside sources). how →
More from Garth Sheldon-Coulson
Panthalassa wave data centers achieve 99.8 percent availability with minimal batteries
“In all of our optimizations, we can be achieving for payloads, you know, we can be achieving with very little battery, like, 99.5, 99.8% power availability with far less battery than you would need for an equivalent solar installation, for example.”
Panthalassa designs deliver wave power at 3.5 to 4 cents per kWh
“We have designs that are two cent per kilowatt hour, On the power. And we, the optimum tends to be, you know, given everything I was just describing, the optimum tends to be in the four cent, three and a half to four cent per kilowatt hour range. But keep in m…”
Panthalassa expects lower AI chip failure rates at sea than on land
“In many cases, we believe that the reliability of the chips, the failure rates will actually be lower on our platform because we can provide much colder cooling temperatures than is typical on land, and we have no oxygen. We eliminate the oxygen from the paylo…”
Bulk of AI energy consumption will power long-running background processes
“There's a whole class of like super latency sensitive applications where you wouldn't want to use us. But that's not where the bulk of energy will be going. The bulk of energy will be going to very long running processes that are churning, churning, churning t…”
Ocean convective cooling eliminates traditional data center cooling infrastructure costs for Panthalassa
“We are actually in a resource that gives us free, extremely good convective cooling, and that's huge, because it essentially lets you eliminate the entire cost structure of the data center. And our object is actually replacing both power plant and data center,…”