What AI Power and Cooling Solutions Should Enterprises Consider?
Having a robust power infrastructure is crucial for most businesses trying to maintain uptime. Having a backup generator is the first step to reinforcing a building’s power source.
Depending on the location of the business, a battery energy storage system could also be helpful. A BESS device charges at night and can be used during the day to augment a building’s power and lower electricity costs.
Whether the company uses a generator, a BESS, an uninterruptible power supply (UPS) or a combination of those options, it must ensure it has enough extended runtime to support its power load.
On the cooling side, organizations with AI-intensive workflows will likely need to upgrade to liquid cooling from traditional, air-based cooling systems. Many organizations currently use a hybrid system in which liquid cools the processor and sometimes memory, while the remaining components are cooled via a fan-based system. However, the next phase of AI will likely require 100% liquid cooling, according to Steve Gillum, solutions manager for power and cooling at CDW.
“When upgrading systems for AI, it’s critical to look at systems holistically,” says Loeb. “The GPU technology powering AI is advancing rapidly, and there are multiple interdependencies that should be considered among servers, racks, power, cooling and the entire system.”
Eaton recently announced a collaboration with NVIDIA to accelerate the shift to 800 volts direct current (VDC) power.
“These next-generation systems can help deliver megawatt-scale rack power safely, efficiently, and with minimal material and infrastructure costs,” Loeb adds.
READ MORE: How are data centers adapting for artificial intelligence?
When Should Businesses Upgrade Their Power and Cooling Systems?
Even organizations that are not yet deploying AI workloads should prepare for the future and have a plan for making IT systems AI-ready, according to Loeb. He recommends that businesses thoroughly assess existing infrastructure and work closely with equipment manufacturers to determine where upgrades are needed to support the potential demand for AI applications.
“Examples of such equipment can include medium- and low-voltage power distribution and protection equipment; power quality systems; UPSs; rack power distribution; racks, cabinets and enclosures; and cooling solutions (including liquid cooling) to manage the thermal needs of computing infrastructure,” Loeb explains.
Avoid Common Power and Cooling Mistakes
Upgrading power and cooling is crucial for organizations that currently have a traditional approach to their data center and are looking to adopt AI. However, it’s equally important to ensure new systems are implemented with cybersecurity in mind.
“Cybersecurity is a must-have and is foundational to building trusted, secure environments. As energy systems become more complex, it opens vulnerabilities for malicious actors. At Eaton, we’re committed to making trusted connections work across a more connected, digital and electrified world,” says Loeb. “Eaton’s Product Cybersecurity Center of Excellence and secure-by-design philosophy help ensure that our products meet rigorous cybersecurity and safety design standards. Our cybersecurity experts integrate cybersecurity processes and a secure development lifecycle into every stage of product development.”
EXPLORE: Modernizing your data center will prepare you for the next phase of AI.
Preparing for AI hinges on clarity, as IT leaders need everything in one place to know what’s coming and act before problems happen, according to Loeb. “They need confidence that their infrastructure is secure, efficient and always on. Software plays a vital role in protecting uptime, simplifying operations and providing value to the business.”
Having a unified dashboard such as Eaton’s Brightlayer portfolio brings power, cooling and IT systems together. Loeb says this helps businesses stay ahead of outages, optimize performance and meet their commitments.
“Brightlayer is built to scale with customers, integrates seamlessly and keeps teams focused on what matters,” he adds. “It helps move from reactive firefighting to proactive control — helping customers sleep better at night knowing that their data center is covered.”
