THERMAL MANAGEMENT
also bring redundancy, serviceability and reliability to the forefront. With centralised control modules and front-service access, they minimise maintenance downtime and simplify operations in dense compute environments where accessibility can be challenging. For operators managing hundreds or thousands of racks, such design considerations translate directly into resilience and uptime.
What does a large multi-site liquid cooling orders reveal about AI data centre trends?
Recent multi-site liquid cooling deployments (such as LiquidStack’s large-scale order from an AI factory in the southern US), reflect more than market growth; they mark a turning point in data centre design. Large, distributed orders suggest that operators are no longer testing liquid cooling but standardising it across sites as core infrastructure.
This shift aligns with a broader trend towards AI-ready campus design: modular, standardised and scalable. Hyperscalers and colocation providers alike are now building “cities of compute,” with power capacities stretching into the gigawatts. For
these facilities, cooling is not a supporting system, it is the architectural foundation that dictates how quickly capacity can be added, how efficiently it runs and how sustainably it operates.
At the same time, liquid cooling’s scalability suits both greenfield mega-sites and adaptive reuse projects, where existing industrial buildings are being retrofitted into AI-ready data centres. In both cases, modular CDUs provide a consistent cooling platform that can be replicated, maintained and expanded globally.
How do CDUs function within direct-to-chip and hybrid liquid cooling architectures to provide effective thermal management and operational efficiency?
Within direct-to-chip (DTC) systems, CDUs serve as the circulatory system-regulating coolant flow to and from cold plates that sit directly atop processors. The liquid absorbs heat at the chip interface and transfers it to a secondary loop for rejection via heat exchangers or dry coolers. This closed-loop architecture drastically reduces energy waste and nearly eliminates evaporative water loss.
In hybrid architectures, CDUs play a bridging role, combining DTC with legacy air systems or immersion setups to support mixed workloads. This approach allows operators to migrate to full liquid cooling gradually, aligning thermal management with capital expenditure and compute evolution.
Advanced CDUs incorporate
instrumentation kits for real-time monitoring of pressure, flow and temperature, enabling predictive maintenance and smart orchestration across the facility. As AI and analytics become embedded in facility operations, CDUs themselves are evolving into intelligent platforms, capable of adjusting flow dynamically based on workload intensity, environmental conditions, or energy pricing signals.
Cooling as the cornerstone of the AI era
The lesson for engineers is clear. The race to scale AI isn’t just about chips, racks, or power—it’s about thermals. Cooling has become compute’s equal partner, and in the age of intelligent machines, that partnership will define how far, and how responsibly, we can go.
FEBRUARY 2026 | ELECTRONICS FOR ENGINEERS
23
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50
