The demand for data center services is growing rapidly and evolving in ways that are challenging to anticipate. Traditionally, forecasting for business growth, developments and investments would involve the study of the overall regional and country landscape, as well as the technological changes and needs happening for the different industries immediately affected.
However, it is unlikely that today's data centers were designed to support increasing consumer demands for the services and applications being strongly influenced and driven by artificial intelligence (AI), machine learning (ML), or high-performance computing (HPC). These technologies are revolutionizing how data is consumed and processed, requiring a more holistic, robust, and adaptable critical infrastructure.
Industry leaders must plan, approach, design, and manage a data center that is future-friendly. Among the complexities brought in navigating AI-driven developments, we pose three main questions that data centers can use as guiding queries for their strategy in the coming years:
- With the changes brought by HPC, AI/ML, how to anticipate growing business needs vis-a-vis data center services?
- What do data centers need to support these new technologies?
- How can we reduce the risk of downtime with the continued increase in densities?
Aspects to consider when building the data center
To address these challenges, it is imperative to consider a phased doubling of the energy matrix in response to the unprecedented pressure on energy consumption. It is also essential to increase the capacity of data center optimize services and increase CAPEX investments to take a holistic approach. Ignoring all of these factors can result in a data center that is obsolete by the time it is operational. So, how do you build a data center for a future that is expected to change so much?
Preparing for the unknown requires a combination of flexibility, innovation, and a long-term vision that takes into account several aspects:
- Manage increasing rack densities: Power density is predicted to increase rapidly to 500 kW per rack. This increase in density also means supporting hardware weights over 5,000 pounds per rack. With the typical lifespan of a data center approaching two decades and the design cycle for AI chips less than two years, today's critical digital infrastructure must be able to support these increasing power and weight requirements.
- Innovation in cooling: Advanced cooling technologies, such as hybrid or direct-to-chip liquid cooling, are emerging as energy-efficient solutions to maintain optimal temperatures and prevent overheating at the chip level. Since most currently existing facilities were not designed with liquid cooling in mind, upgrades are needed to accommodate this technology. Whether the operator is redesigning racks or rooms at a time, he must consider where to place a new busway, how to plan for redundancy, and where to route power, among many other technical and business considerations.
- Scalable data centers: Enterprises must be prepared to meet the demands of AI/HPC to remain competitive in the marketplace. Prefabricated modular critical infrastructure solutions are ideal for achieving the necessary scalability and building future data center expansions. Benefits include fast deployment compared to new build, energy-efficient design, reduced risk, cost certainty, and high quality with the ability for integrated factory control.
- Ongoing training: HPC data centers require current and upcoming professionals to acquire new skills while remembering their adaptable foundational knowledge. More than ever, keeping this environment running efficiently requires a skilled technical team to maintain it. To help train professionals prepare for this new era, we have launched the Vertiv Academy Latin America, replicating the evolving critical digital infrastructure in a data center environment by providing hands-on training in a safe and controlled testing environment.
- Efficient resource management and optimization: The development of new AI servers generates thermal loads of 6 kW and 10 kW per server, and the thermal densities of chips, servers, and racks continue to increase. Facility energy expenditures can represent over 50% of data centers’ total operational costs. Considering rising energy costs worldwide and the fact that a percentage of the said operational costs are passed to the customer to enable the energy industry to consistently provide electricity, efficient energy management becomes even more imperative in the face of these changes.
Moreover, these must be integrated with equally effective site management technologies. Efficiency is not limited to the equipment; the manager must deeply understand the available operational modes, costs, and tariffs to make informed decisions that reduce the facility’s environmental footprint.
- Work holistically: When selecting a data center provider, it is critical to evaluate whether the provider covers all aspects of the data center infrastructure—from site planning to selection, to deployment and maintenance. At Vertiv, we have found that having a local presence allows us to efficiently meet our customers' distinct needs. Our work goes beyond the purchase of hardware, software, or services: it is a holistic view critical to successful operations from planning to equipment end-of-life cycles, and avoiding environmental, logistical, and material issues.
Being ready for the future means maintaining data center efficiency despite an increasingly demanding marketplace, which requires a proactive and comprehensive management approach based on responsible business practices, innovation, and staff training. At Vertiv, our collaboration and research with our local partners, combined with a deeper understanding of the drivers behind technology evolution, enables us to design and deliver solutions that prepare both data centers and enterprises for their future needs. With our expertise and a broad and diverse range of solutions, we enable data centers to be more competitive and adapt to the demanding applications of the future, such as AI, machine learning or high-performance computing.
This means we not only react to what happens today, but also help organizations plan for growth by synchronizing our expertise with the business needs and technologies of tomorrow. This foresight allows us to empower our customers to anticipate and respond to market demands as quickly as possible to maintain their competitive edge through an optimal balance of performance and efficiency.
At Vertiv, we are your technology partner. We understand that each of these issues takes work and time, and we have developed benchmark designs for revamped and new data centers to help you find integrated solutions for power and cooling HPC and AI applications. To know more about these solutions, find our comprehensive portfolio here.
