Vertiv Unveils Full-Stack AI Cooling Orchestration for High-Density Data Centers

Vertiv has introduced a portfolio of liquid and immersive cooling technologies designed for AI workloads in high-density data centers. These innovations aim to address the rapidly growing thermal and power challenges created by AI model training and inference, particularly in facilities now exceeding 100kW per rack.

Announced during Vertiv’s “Cooling Innovation Day” on July 15, 2025, the new systems include immersive cooling, direct-to-chip cold plate solutions, and integrated thermal orchestration software. A key feature is the CoolLoop Trim Cooler, designed to handle wide water temperature ranges while supporting chip-level cooling at up to 45°C.

Vertiv claims these systems can reduce annual cooling energy usage by up to 70% when used in hybrid free-cooling and mechanical modes. In high-demand environments, this could significantly improve operational efficiency and sustainability. These systems are also space-efficient, an increasingly critical factor in APAC’s urban data center markets.

As AI compute requirements surge, data center operators across Asia are facing intensified pressure to reduce Power Usage Effectiveness (PUE) and maintain thermal reliability. Vertiv’s full-stack offering is positioned as a scalable solution for AI-ready retrofits and new builds, especially in markets like Singapore, South Korea, and India.

The company has ongoing partnerships with GPU and CPU manufacturers, including NVIDIA, AMD, and Intel, aligning its thermal roadmap with the high-power demands of next-generation AI accelerators. With these solutions, Vertiv is targeting operators transitioning to rack densities of 50-100kW and aiming to stay ahead of liquid cooling adoption curves across the region.

About Author

Rex Sayson

See author's posts