Erick Soares, Director of Product Management – Stationary, C&D Technologies, says as AI and edge computing continue to impact power requirements, UPS cannot be considered at the end of the design process either.
Unprecedented growth in AI and edge computing is changing how data center power demands must be met. While AI and edge computing both have a profound impact on data center requirements, they have distinctly different power needs.
When determining backup power for modern data centers, three factors are often overlooked but must be considered to ensure uninterrupted operation – and prevent potential loss of data. They are power quality and availability, sustainability and total cost of ownership (TCO).
Before taking a closer look at these factors, let’s first discuss the main differences between typical AI and edge backup systems.
Understanding AI Power Requirements
AI is becoming an even bigger part of our daily lives. With the additional benefits it brings, it also comes with higher complexity and requirements. For example, a ChatGPT search requires 10x more electricity than a conventional Google search. Compounding the issue is all that additional computational capacity is expected to be supported within the existing data center footprint. The extra capacity also places considerable added pressure on temperature, cooling, and grid resiliency.
Only a few years ago, data centers often had power densities of 30kW per rack with many now having 50kW per rack. The fast adoption of AI is quickly making that a thing of the past with densities as high as 150kW per rack already being planned, with 500kW on the horizon. All told, AI systems will require backup systems with well over 1mW of power while having to deal with up to 150% power peaks. For these reasons, AI requires a massive singular, and ideally purposely built, data center.
Edge Computing: Decentralized Computing
Edge computing has very different power requirements than AI. It brings computational power closer to the user to reduce latency. With cost and availability of both land and power closer to urban areas being a huge challenge, one solution is smaller data centers with lower power requirements.
While AI backup systems might be around 1mW and larger, edge can, in general, be anywhere from 250kW to 500kW. Given their smaller footprint and more difficult access to resources, backup time is often approximately 15 minutes or longer.
Designing UPS Based on Application
Because of these differences, data center decision makers must design a backup power system that addresses the specific requirements of AI or edge computing. The three main considerations, however, remain consistent.
Power Quality and Availability: Quite simply, without power there is no data, making power availability a main consideration. Is the grid powering the data center or will an alternative source, such as wind or solar, be used? Will the source be efficient enough to support the data center or will it be necessary to use the backup system regularly? The answers to these questions will help determine the best battery technology to use.
For example, float batteries are very power efficient but must be charged at all times so they can be used when necessary. If a float battery is used too frequently, its life expectancy will be shorter, meaning the return on investment (ROI) will be lower due to more frequent battery changeouts. Cycling batteries are better for on-off applications and can better cope with constant use. They are not as power dense as float batteries, though, leading to the need of oversizing the system.
Another consideration is the capacity of the data center on Day 1. If it is at 100% there must be enough power to support it. Future power needs – and the source – must also be factored into the system design.
Sustainability: Everyone agreesthat sustainability is important. True sustainability must account for the full life battery cycle with all three factors – sourcing, operating, and disposing.
Selecting a battery from a manufacturer who acts responsibly, ethically, and with integrity will ensure a battery is properly sourced. Manufacturers should follow sustainable practices throughout the design and manufacturing processes.
Different technologies will have different recycling programs. Lead acid, for example, has a well-established ecosystem that will pay you to recycle your batteries. Newer technologies, such as Lithium-ion, on the other hand, are still developing those ecosystems and might cost money to be recycled.
At the end of the day, every customer must have a very clear understanding of what it means to be sustainable. It will ensure that all its products, and suppliers, are aligned with that vision.
TCO: So much more goes into the lifetime cost of a battery than the initial price. Decision makers must take a holistic approach. When thinking of a battery backup system, there are four factors – installation (which is a CAPEX), maintenance (OPEX), replacement (OPEX), and disposal – to accurately determine TCO.
Installation includes Day 1 costs, such as the batteries, along with cabinets or racks, and any labor required to connect the system. For maintenance, frequency is the biggest factor with potential requirements for higher skilled labor, depending on the technology that is chosen, also playing a factor.
Every asset – regardless of technology – is finite and all batteries will need to be replaced eventually. For replacements, cost will, of course, be an important factor, but it will also be heavily influenced by frequency, supply chain resiliency, and lead times.
When comparing different technologies, it can be very hard to determine when a battery should be replaced. While there isn’t one approach, it is highly recommended to start a TCO calculation using the standard warranty provided for each technology. A warranty is a clear sign of how much a company is willing to stand behind its products, making the only fair way to compare alternatives. However, you must be careful with special one-off warranties, as well, and use the standard warranty as the true measure.
Let’s use an example to illustrate this. If a product has a standard 5-year warranty, and you are being offered 7, is that a good thing? Are you really that special or is it just more risk they are willing to take to earn your business? If a battery was made to have a standard 5-year warranty, can it even reach 7 years?
Conclusion
There are many factors to consider when designing a battery backup system, especially in the age of AI and edge computing. It goes beyond the battery itself and extends to finding a power partner who will sit down and discuss specific requirements to develop a solution that best fits the application.
As AI and edge computing continue to impact power requirements, UPS cannot be considered at the end of the design process either. It must be done at the outset, so power demands are consistently met.