As the industry drives to build even more data centers, thanks to the rise of artificial intelligence (AI) and Edge rollout, not to mention the insatiable demand for data by the global population, streamlining construction processes for the new era is a vital consideration. It is no longer enough to simply copy an existing blueprint that worked last time.

Black Box is a digital integration company that works with hyperscalers as they build out their data center infrastructure. Following their DCD>Talk at our Virginia event in November , DCD spoke to Tashbeeb Shahid, senior director of global data center solutions at Black Box, to look at their best practices when building facilities of the future.

Shahid points to three of the biggest factors affecting data center design in the AI age. The first is the need for faster chips with more compute to process data at the necessary throughput. The second is the power draw that this more advanced equipment requires, and the third, which we focus on here, is the accelerated speed to market required to keep pace with demand.

Harnessing AI for proactive and predictive data center management

He tells us: “All the apps that people are deploying, the various level servers, the networking equipment which primarily supports the AI flow – how fast you deploy, collect data, and process that data – everything is dependent upon speed to market.”

There is little doubt that, whoever you are, even outside our industry, AI is the future, Shahid reminds us: “But we who are living this dream, so to speak, see it as a present phenomenon – one that has been unfolding over the past decade or even longer. It has just accelerated, particularly with the media hype over the past few years.”

The starting pistol for the wider adoption of AI is generally agreed to have been ChatGPT, which brought the possibilities into the home. Shahid points to previous examples of AI use cases, such as Industrial IoT, which has not had the same level of “cut through” with the general public, but has seen wider adoption as the potential of AI becomes part of the public consciousness. Meanwhile, within the data center, the use of AI as a tool to optimize and regulate operations is one of Black Box’s tenets for the future of data center design:

“The use of Data Center Infrastructure Management (DCIM) for infrastructure management is going to be the key,” says Shahid, “It is always the starting point for how issues are resolved. Instead of manual troubleshooting and analysis, which has been the case for years, you can now really benefit from automating these tasks. It’s going to be game-changing.”

What is often forgotten is that a lot of the infrastructure that will enable intelligent DCIM systems is already being installed. Industrial IoT sensors track almost every facet of the data center environment, but the potential of the information they collect is not being realized. AI can collect, collate, compute, cross-reference, and diagnose issues at a speed that would be impossible for a human to do in anything near real-time.

“Whether it be network sensors, systems sensors, or peripheral sensors – the data they have been gathering has mostly just sat there. We have not been able to process it dynamically and in real-time until now,” says Shahid.

“With an AI automation tool processing that data, you can have proactive troubleshooting or live healing networks. You can create a proactive predictive analysis of what component is going to fail next, and how to heal it. Self-healing networks, as they're known, are already being tested and deployed in data centers. They can reroute traffic so it's a seamless migration to the end user, while the nodes which have failed either undergo troubleshooting, are replaced or changed.”

With every new piece of data and correct diagnosis, an AI tool will learn and improve, meaning that it becomes more valuable and more useful as time goes on, and can soon reach the point where it will predict problems before they occur.

Doomsayers may suggest that this is a step down the path to unstaffed data centers and massive industry layoffs, but we have the opportunity to use AI to primarily advise humans, freeing us from menial, repetitive tasks in favor of more esoteric ones that must be done manually, or even creatively.

Shahid believes that the AI rollout in the data center will go in phases: “It’s not a bleak sci-fi with some AI taking over, it’s not ‘AI gone bad’,” he quips, “The first phase has been allowing the AI to learn correlations, with human supervision. The second phase, which we’re beginning to see already, is semi-autonomous data centers capable of self-healing, basic provisioning, and automation to alleviate some of those mundane tasks.”

‘So the third phase isn’t Skynet?’ we inquire, equally lightheartedly, and after swapping a few fictional AI tropes, he assures us: “Wearing my pragmatic hat, there’s some time before that is even possible. The biggest danger would be deploying technology without stress testing, lab testing, and quality assurance. That could decrease efficiency, or even do harm, which in turn will reduce confidence, so those scenarios are a bit in the future.”

The rise of modular design: Streamlining construction and sustainability

Another key tenet in the Black Box view of future data centers is a move toward modular design. Integration with existing infrastructure: when implementing and deploying a modular data center solution, it is imperative to plan how the solution will connect with the current systems in place.

Similarly, the scalability through the modular data center is equally important: How would data center operators be allowed to scale and replace or augment their systems, instead of additional overhead and expensive reconfiguration projects. Shahid believes that this is the way that many data centers are moving, and it is an exciting opportunity.

“There is a huge growth potential for this industry – a large number of players in the market are moving toward a modular ecosystem. Everyone from hyperscalers and tech giants, to the solution vendors, the service vendors, the system integrators – everyone is trying to see how they can package their products in enclosures and modules that can be bundled and moved on-site, or augmented within existing data centers.”

This approach has seen an increasing alliance between software and hardware businesses that have traditionally been quite insular, now working toward a common aim. The cause is a just one because modular and prefabricated data center construction can shorten lead times for equipment, and build times for facilities. There are other benefits too, as Shahid explains:

“Having those parts in stock, units already pre-built and available for immediate shipment means they can be tested in an offsite facility, and certified before shipment. It shortens the commissioning taking place on-site, perhaps as much as 30-50 percent. Then, because they're built in a very focused environment, the manufacturing, assembling, and integration are optimized and quality improves. You can also look at sustainability elements – a potential reduction in Scope 3 emissions, for example.”

A modular approach also opens up the data center market to areas where it may have been impractical to locate in the past, perhaps due to the lack of infrastructure and skilled labor force, or a limit on the amount of clement weather in which to perform the build. With more elements being built offsite, the time to deploy is reduced.

“The key is, ultimately, we're still dependent upon humans,” explains Shahid, “And you need a specialized, skilled workforce to construct power, electrical infrastructure, and networking. In remote areas, these modular data centers bring fantastic efficiency because you can just build and ship them. It's a repeatable process in a very controlled environment, and you need a smaller workforce to do that.”

This has even led to Black Box working on hyperscale facilities closer to the Arctic Circle, with the bulk of assembly and testing happening in a climate-controlled environment, before being put in data halls. Although building this far North has obvious logistical challenges, these can be offset by prefabricating infrastructure to reduce deployment time and complexity, having colder climate generally helps with energy efficiency – making such an investment pay off over the lifetime of the data center. However, this is not a casual undertaking, whilst the project has become a success for Black Box, many other factors at play would make it a prohibitive undertaking in most circumstances.

Edge computing and sustainability: The path to smarter, greener data centers

A third priority for the Black Box construction philosophy is Edge computing. This is another area where a huge explosion in take-up has led to more companies working together to ensure best practices and interoperability.

“The key drivers of the Edge market include augmented reality, virtual reality, video streaming, and industrial IoT applications – all the biggest tech trends that have proliferated over recent years. Companies have realized the significance of proximity, being the key to the end user base, because it reduces the distance and therefore the latency between them and the data center.”

5G has been a huge driver in the Edge market. “Low cost, low latency, high bandwidth,” says Shahid. “The perfect example is a smart city, where you get IoT devices, only now you've got low latency on the Edge to process the mountain of data. There's a huge amount of data that is collected from Edge devices. So it’s ‘smarter’ to perform distributed data processing at the edge to manage traffic more efficiently and improve energy consumption and various other related KPIs.”

This combination, particularly when used holistically, can also help on the path to a sustainable future. Building modular components invites closer scrutiny from a more controlled environment. Building at the Edge spreads out the load on local power grids. Machine learning (ML) or more specifically ‘Predictive Analytics’ important facets of AI, can reduce malfunctions that could otherwise lead to increased emissions, or even accidents that could cause environmental crises. It can allow for data centers to be positioned nearer to sustainable, natural power sources, and make use of an individual projects’ circumstances and environments

“When and as AI processes data locally, it reduces the need for routing traffic through high latency networks across geo-diverse data centers,” adds Shahid. “It cuts down on the carbon footprint associated with long-distance data transmission to centralized data centers. When you deploy closer to the Edge, you're not only enhancing productivity but also taking steps to be a little bit more sustainable. Network congestion is reduced, for example, which reduces energy consumption in itself.”

We finish by asking Shahid what needs to happen to encourage this approach. He explains that we need to continue down the path of cooperation between stakeholders, a process that, as we’ve discussed, has already begun:

“The communication path between stakeholders in the ecosystem needs to be transparent, broader, and more inclusive. That will increase the efficiency, so we will all benefit,” he says. “We just all need to cooperate a little bit more.”

Black Box is at the forefront of innovation, bringing modular design, AI-driven efficiencies, and Edge computing to data center construction. Whether you're planning hyperscale projects or Edge deployments, our tailored solutions ensure speed, quality, and sustainability.

Contact Black Box today to discuss how we can help build your next-generation infrastructure. Visit this link to start the conversation or learn more about Black Box data center solutions here.