Lotte Innovate is partnering with Ajou University and other research institutions to develop next generation thermal management for data centers. The project aims to replace traditional air cooling with AI driven systems and fiber optic sensing to handle the high power density of modern GPUs. The goal is to reduce Power Usage Effectiveness (PUE) to 1.2 or lower. If successful, applying these efficiency gains to data centers in Seoul and Yongin could cut annual carbon emissions by 22,000 tons.
Log in to join the discussion
PUE of 1.2 is a solid target, but the real question is the cost of the retrofit. Moving from air cooling to in rack liquid or hybrid systems usually requires a massive overhaul of the physical infrastructure. Do they have a timeline for when the first commercial site actually switches over?
Exactly. The hardware is one thing, but the maintenance on these high efficiency systems is where the budget usually leaks. I want to see if they are accounting for the long term upkeep of fiber optic sensing in a high heat environment.
The scale here is interesting. If the broader industry moves from 1.5 to 1.2 PUE, 684,000 tons of CO2 is a significant chunk. It would be worth seeing if the Korean government is tying these R&D grants to mandatory reporting standards for data center operators.
Basically Lotte is trying to fix the heat problem caused by AI GPUs. They are using a consortium to test AI thermal management and new cooling racks. If it works, it cuts a lot of carbon, but it is still in the R&D phase.
