What a difference a year makes. Just 12 months ago, we highlighted how the soaring demand for AI compute in data centers was set to drive the need for more efficient processes, faster builds, and innovative problem-solving, especially in the face of a skilled IT talent shortage. It’s safe to say that this has played out even more dramatically than anticipated.

AI technologies, especially generative AI, demand considerably more power compared to traditional data center workloads. This is due to the intensive computational processes involved in training and running AI models, leading to higher energy consumption. Indeed, The International Energy Agency (IEA) forecasts that AI-driven web searches could cause a tenfold increase in electricity demand. The IEA also predicts that by 2026, total data center electricity consumption will surge from around 460 TWh in 2022 to more than 1,000 TWh.

If recent reports about DeepSeek are accurate - and it does deliver equal output but with lesser power demands - then those lofty energy projections may need reassessing. However, efficiency improvements don’t always lead to reduced energy consumption. In AI’s case, increased efficiency could accelerate adoption and expand capabilities while keeping energy usage in line with initial projections.

In the meantime, we’re seeing major developments and ambitious plans from the industry, with shifts in data center infrastructure designed to elevate cloud computing to new heights.

The Shift From Big Compute To Small

The expansion of AI into nearly every aspect of our personal and professional lives has been remarkable. The closest comparison might be the early days of the World Wide Web – when the internet first emerged as a global force in the mid-to-late 1990s. Initially viewed with a mix of excitement and skepticism, the internet soon became indispensable. History shows that it took about 50 years for the telephone to become a household staple, while the internet achieved widespread adoption in less than two decades. AI tools like ChatGPT appear to be accelerating at an even faster pace, particularly in the enterprise space, where its adoption is largely supported by data centers.

The range of enterprise applications for AI is already huge. Yet, while it is proving the most transformative technology in decades, its influence is often subtle and it is regularly put to work behind the scenes, driving efficiency and innovation across the enterprise landscape.

The Surge In Data Center Construction

The world’s largest tech companies are ramping up infrastructure investments, driving long-term capital expenditures higher as the race for AI computing power accelerates. But it’s not just the AI network that’s evolving, its delivery is transforming as well. AI-as-a-Service is streamlining enterprise adoption, particularly for generative AI, which is now enhancing functions from customer service to financial planning. Even data centers are leveraging GenAI to monitor, manage, and support lean IT teams more efficiently.

Given the concern around energy supplies for AI data centers, operators in many regions are seeking new locations nearer to affordable power sources, or securing dedicated grid capacity to ensure future needs are met. In some cases, this has even led to partnerships with nuclear power providers.
It might seem like a strange connection, and yet nuclear energy offers a practical solution; data centers need clean and reliable power; nuclear can provide that. Furthermore, data centers are also hoping to optimize energy consumption by making use of tools like water-cooling systems, which are far more effective than traditional air cooling. As GPU-powered AI workloads continue to scale, such efficiency gains will be critical, not only for reducing costs and environmental impact, but for preventing overheating, which is a major contributor to outages and premature hardware failures.

Streamlining Infrastructure For The Future

AI workloads require every GPU in an AI array to be fully interconnected, significantly increasing network complexity and making cooling even more challenging. To manage the bulk of this fiber infrastructure, data centers are turning to high-density fiber systems, maximizing the number of fibers and connectors within existing spaces to support AI-driven networks.

Another strategy for optimizing energy use is consolidating more compute power into fewer racks, improving energy efficiency and simplifying cooling requirements. Looking ahead, as hyperscale data centers transition from 2×400G (aggregate 800G) to native 800G, advanced fiber infrastructure will play a crucial role in providing the capacity needed to support future energy demands.

Balancing Standardization And Flexibility In Multitenant Data Centers

Experts have extensively analyzed the role of hyperscale data centers and their licensed AI-as-a-Service models in enterprise environments. However, another crucial aspect to watch in 2025 is how multi-tenant data centers (MTDCs) will evolve to meet the rapidly shifting demands of their enterprise customers. Regardless of industry, businesses require increasingly adaptable solutions, and MTDCs must stay agile to accommodate these needs.

A standardized approach to denser fiber infrastructure plays a critical role in this evolution, helping to reduce IT staff demands and streamline configuration changes. Several top fiber infrastructure manufacturers are actively developing and refining simpler plug-and-play technologies. These advancements are particularly beneficial for MTDCs, enabling them to be more agile and responsive, maintain service-level agreements, and operate effectively with leaner IT teams.

What To Expect In The Year Ahead

The rise of AI is driving significant, foundational changes in data centers, reshaping how they are built, powered, and scaled. In the coming year, MTDCs will begin to scale up their fiber capabilities while reducing their fiber's physical profile, integrate advanced cooling solutions and rethink their approach to energy procurement and consumption. Although a shortage of skilled IT professionals will remain a challenge, AI itself is already proving to be a valuable tool, with GenAI-powered monitoring and management helping to fill gaps.

As AI becomes more deeply embedded in enterprise operations, data centers will be critical in providing the vast computing power necessary to translate its potential into real business benefits. In line with AI’s evolution, data centers will have to innovate, adapting to industry demands and delivering the cutting-edge solutions needed to support this changing sector.

Data Management

How AI Will Shape The Future Of Data Center Design