Smarter Power and Cooling Strategies for Modern Data Centers
(Source: Nomad_Soul/stock.adobe.com)
As data centers push to pack in more computing capacity per unit area to meet increasing workload demands, powering all of the IT equipment, cooling technology, and other building systems becomes a limiting step. McKinsey projects that global data center capacity demand will grow at 19–22 percent annually from 2023 to 2030, reaching between 171 and 219 gigawatts.[1]
This surge is driven by AI workloads that consume significantly more power than traditional applications and it accelerates the need to mitigate thermal effects resulting from fitting more power into smaller spaces. Recent developments in power distribution, backup systems, and thermal management demonstrate how emerging engineering solutions address these increasing power demands. In this blog, we cover some of the latest approaches to improving power and thermal efficiency in modern data centers, highlighting efficient and dependable solutions from Molex.
Rethinking Power Distribution with Open Rack V3
The Open Rack V3 (ORV3) specification developed by the Open Compute Project (OCP) changes how data centers approach power distribution. The specification defines scalable, adaptable rack power distribution with the ongoing transition from 12VDC to 48VDC power. Higher voltages increase efficiency while increasing data center power per unit area (power density), subsequently mitigating spatial constraints that often plague data center designs.
With this transition, power is distributed through a 48V busbar in each rack, to which servers connect directly. By eliminating the 12V conversion step, facilities improve overall efficiency and reduce energy losses during transmission. The architecture saves space by removing individual power supplies for each server, as servers run directly off the 48V busbar. This consolidation reduces component count, simplifies maintenance, and improves power density within each rack.
Moreover, the shift to 48VDC represents more than incremental energy efficiency improvement. Higher voltage enables lower current draw for the same power delivery, which translates to reduced conductor sizing requirements and decreased resistive losses. For enterprise-sized facilities, these efficiency gains compound across thousands of racks. The ORV3 specification provides a standardized framework that allows interoperability between vendors while giving operators flexibility in system configuration.[2] Molex's PowerPlane Bus Bar Connectors and PowerPlane OCP ORV3 cable assemblies are at the forefront of component innovation designed to meet these architectural requirements.
Designing Reliable Backup Power Systems
Getting power from the grid to servers presents one set of challenges, but building redundancy against grid failures introduces an additional set of engineering opportunities. An uninterruptible power supply (UPS) must draw on battery banks and generators to fill gaps if grid power fails, and the transition must occur seamlessly to prevent service interruption. In data centers, continuous operation is the top priority.
Online double conversion UPS systems are currently the preferred topology because they can provide instant, clean, and consistent power.[3] These systems continuously convert incoming AC power to DC for battery charging and system operation, then invert DC back to AC for output. This double conversion isolates the load from grid disturbances and ensures zero transfer time during power interruptions.
UPS performance depends heavily on how much resistance exists at each connection. Lower contact resistance reduces heat and voltage loss, improving both reliability and energy efficiency. Molex's 2023 State of Power survey indicates that 40 percent of data center engineers cite power management, including circuit protection, switchgear, and battery systems, as their most significant design challenge (Figure 1).[4]
Figure 1: Responses to a 2023 survey on what industrial application is most challenging for designing or implementing power systems. (Source: Molex)
Components engineered for high current density and low contact resistance address this concern by ensuring reliable power delivery under varying load conditions. Molex's innovative line of power management solutions are designed with these electrical parameters as primary specifications, enabling data centers to achieve peak efficiency and reliability.
Tackling the Cooling Challenge
Approximately 40 percent of data center electricity consumption goes toward cooling,[5] making thermal management one of the largest operational expenses for facilities. With power consumption at this magnitude, engineers and fluid chemists have intensified focus on cooling efficiency to reduce both energy costs and environmental impact.
Immersion cooling has emerged as a promising technique that improves energy efficiency compared to traditional air cooling systems, while also reducing the physical footprint of cooling equipment.[6] In immersion cooling, servers are submerged in dielectric fluid that directly absorbs heat from components. This eliminates the need for air handlers, so the system doesn't have to work as hard to move heat, translating to lower energy use and more efficient heat removal within the same footprint.
There are also single- and two-phase immersion cooling approaches, as well as direct-to-chip (DTC) cold plate cooling. These liquid cooling methods improve data center Power Usage Effectiveness (PUE) between 1.03 and 1.2, whereas air-cooled systems operate between PUEs of 1.2-2.0.[7] PUE compares total facility power against IT equipment power, meaning lower PUEs are more efficient. A PUE of 1.0 is theoretically perfect, but impossible to achieve due to losses. Reducing PUE from 2.0 to 1.2 represents a 40 percent facility power reduction, for example.
Regardless of the cooling method deployed—air, DTC, or immersion—connectors for cooling hardware must meet strict environmental requirements. These connectors must be rated for harsh conditions and sealed against moisture to prevent corrosion and electrical failures. High power and heat make reliability a constant concern. If connectors can't handle temperature or humidity, even a small failure can spread quickly through the system.
Engineering for Efficiency at Scale
Data center engineers are finding new efficiency gains not by tweaking existing systems, but by rethinking how power, backup, and cooling are built from the ground up. Part of this new approach involves moving to 48VDC power distribution to lessen conversion losses, online double conversion UPS systems with low contact resistance connectors to maintain power quality during transitions, and immersion cooling to reduce the energy overhead of thermal management.
Engineers who apply these design principles now will be ready to meet today's efficiency goals and tomorrow's growth demands. Each advancement increases computing capacity within tighter power and thermal envelopes. As workload demands continue to grow, especially with the expansion of artificial intelligence applications, these solutions increase the limit for what’s possible.
[1]https://www.mckinsey.com/industries/technology-media-and-telecommunications/our-insights/ai-power-expanding-data-center-capacity-to-meet-growing-demand
[2]https://www.molex.com/en-us/industries-applications/servers-storage/open-compute-project/ocp-rack-and-power-orv3
[3]https://www.molex.com/en-us/blog/ups-key-data-center-continuity
[4]Molex, "2023 State of Power Survey Report," May 2023
[5]https://www.energy.ca.gov/publications/2024/demonstration-low-cost-data-center-liquid-cooling
[6]https://www.molex.com/en-us/blog/orv-immersion-cooling-insights
[7]https://www.thegreengrid.org/
Author
Adam Kimmel has nearly 20 years as a practicing engineer, R&D manager, and engineering content writer. He creates white papers, website copy, case studies, and blog posts in vertical markets including automotive, industrial/manufacturing, technology, and electronics. Adam has degrees in chemical and mechanical engineering and is the founder and principal at ASK Consulting Solutions, LLC, an engineering and technology content writing firm.