Perspectives | 23.07.2025 | 5 min read
In this Perspectives article, Aleksandar Grbić, Global Data Center Solution Manager at Hitachi Energy, explores the evolution of data centers as critical components of the power grid, and the necessity for their flexibility.
Data centers are no longer on the periphery of energy planning - they are at the center of it. In the span of just a few years, they have evolved from large but predictable IT loads to dynamic, power-hungry AI hubs shaping how entire grid systems are planned and operated.
This is a structural shift. We are seeing the emergence of data center campuses that require gigawatt-scale connections, where just five years ago, a 100 MW facility would have been considered a giant. The scale and complexity of AI demand in particular is changing the game, not just for operators and developers, but for utilities, regulators, and energy infrastructure providers.
In short: Data centers are now critical grid players, and the systems that power them must be built on flexibility from the start.
The demand surge we are seeing is unlike anything in the sector’s history.
In some customer conversations, we are talking less about “data centers” and more about “AI factories”, which gives a sense of the scale of the shift.
What makes this different from past growth cycles is not just the volume of power required, but the volatility of it. AI workloads can behave very differently from traditional cloud applications. The compute intensity is higher. The cooling requirements are steeper. And perhaps most critically, the power load patterns are far more unpredictable.
In this context, a power system built for predictability and baseload planning is no longer fit for purpose. Instead, we must plan around fast-moving, high-density, intermittently peaking loads that may behave more like industrial machines than traditional server farms.
Data centers are now critical grid players, and the systems that power them must be built on flexibility from the start.
It’s clear from every customer conversation that energy availability is the number one constraint.
In many leading data center markets, developers have access to capital, land, fiber, and skilled labor. What they don’t always have is a viable grid connection within the required timeframe. That’s a problem, and it’s getting worse.
Being a load on the grid is not enough. It’s necessary to reimagine the role of the data center to become a flexible and versatile player in a much more dynamic power system.
This means data center operators need to work together with utilities and infrastructure partners earlier in the design cycle. Additionally, it means building in capabilities, such as storage, on-site generation, load shaping and participation in ancillary markets, which were previously considered out of scope even a few years ago.
To accelerate capacity delivery and reduce land use, modular, prefabricated substations are being deployed, with integrated gas-insulated switchgear and compact transformers. These solutions are not only faster to deliver, but inherently more flexible and scalable.
Consider Castlebagot in Ireland. In one of Europe’s most grid-constrained regions, a compact substation was delivered to support a hyperscale campus enabling grid connection despite land constraints and tight deployment timelines. This project represented a key milestone, necessitating a complete rethink of the grid design toolkit. And it demonstrated what is possible when grid and site planning are considered in parallel from the outset.
This approach is now being adapted across projects in the US, Europe, and Asia, especially in urbanized regions or where traditional grid buildouts are too slow or costly to deliver.
It’s clear from every customer conversation that energy availability is the number one constraint.
AI and data centers are often seen as consumers of power. But they can also provide flexibility back to the grid, if designed to do so.
This is especially true when data centers combine behind-the-meter generation such as solar, wind, or hydrogen-ready gas engines, battery energy storage systems (BESS), and intelligent energy management to provide reliability of power supply.
Transforming data centers with this approach to energy enables operators to reduce their reliance on the grid and enhance operational continuity by providing seamless power during grid outages.
Historically, the energy sector and the IT sector operated in isolation. But this method is inadequate to address the current scale of challenges.
In Finland, the recaptured waste heat from a data center’s cooling process is redirected into the local district heating system, cutting emissions and lowering local heating bills. This kind of whole-system integration is increasingly becoming the norm, not the exception.
Just as AI is revolutionizing what is possible in software and services, it is also reshaping the way we operate infrastructure.
Intelligent systems are not just delivering power - they are optimizing it. From smart transformers to digitally monitored switchgear and predictive analytics for maintenance and operations, real-time data and decision-making is being embedded into every part of the energy value chain.
We have reached the moment where the effective collection, analysis, and application of data is as important in energy as it has been in IT.
Digitalization is what enables flexibility. It helps data centers shift loads, align compute with renewable energy availability, and coordinate with grid conditions in real time.
Digitalization is what enables flexibility.
We need to be honest - no one actor can solve this challenge alone. The complexity and urgency of these changes require deep collaboration across sectors, that is, between operators, utilities, equipment manufacturers, regulators, and governments.
We have seen this work. In projects where energy and IT stakeholders come together early, we have delivered faster, cleaner, and more resilient systems than anyone thought possible.
The trajectory of AI, data centers, and power systems remains uncertain. We don’t yet know whether tomorrow’s AI workloads will achieve even greater optimization and efficiency through the development of new language models or ultra-high intensity computer engines. We don’t know exactly where demand will peak next, or how regulation will evolve.
But we do know this: If we build with flexibility, we can adapt to any situation. Combining availability, sustainability, and reliability, a flexible strategy is a "no-regrets" strategy.
In an era of transformation, building flexible infrastructure is how we stay resilient. It’s how we protect the grid. And it’s how we ensure that AI and digital services can continue to grow - without leaving sustainability behind.
A flexible strategy is a "no-regrets" strategy.
