As companies race to embed artificial intelligence (AI) across everything from customer service to product development, they’re also driving an unprecedented surge in energy consumption. Behind every slick chatbot or predictive dashboard is a data center working overtime, with many at capacity. It’s therefore no wonder that British real estate services company Savills reported that, at the start of 2025, 2,870 megawatts of new data center capacity is under construction in EMEA, with nearly 64% of this capacity pre-let. To put 2,870 megawatts into perspective, this could power 287,000 UK homes for a day, based on calculations that an average household uses around 10kWh a day. As we continue to feed our daily routines into AI systems, both at home and at work, how data centers respond will be critical.
According to Cadence, the leading EDA and system solutions tools provider, in its Innovation Imperative report, nearly three-quarters of operators (74%) report rising demand from AI workloads, while 70% say national power grids are already being stretched to their limits. There’s no getting around it. AI is here to stay, and rightly so, given its potential to boost business and everyday life. However, unless data centers evolve, and quickly, we risk stalling the very technology we now depend on.
That’s where digital twins — virtual replicas of physical systems — can help. Long used in industries like automotive and manufacturing, they’re now being hailed as a game-changer for data center operators seeking to innovate without compromising critical AI infrastructure.
The Hidden Cost of Progress
Data centers are the beating heart of AI. However, they’re also incredibly complex ecosystems where energy, cooling, and space must be managed with pinpoint accuracy. Any inefficiency comes at a cost, and right now, those inefficiencies are rife.
Take stranded capacity. 29% of operators admit their facilities aren’t fully utilizing the energy or space they’ve invested in. It’s like playing five games of Tetris at once. Racks are partially filled, energy is underused, and no one has complete visibility of what’s going on where. Once you add in the tendency to overprovision resources, which 60% of operators admit to doing, you have a recipe for bloated energy use and unnecessary strain on the grid.
This is precisely the type of challenge digital twins are built to solve. Even better, when powered with AI, simulations give operators a real-time, bird’s-eye view of their infrastructure. They can identify underused servers, model workload distribution, and test changes in a virtual space before deploying them in the real world. Not only does this save power, it also provides data center operators with a much clearer, holistic view of the facility.
AI’s Heat Problem
Of course, energy isn’t the only challenge. AI doesn’t just consume power; it generates immense heat. As high-density servers become standard, traditional air-cooling systems are struggling to keep up. As a result, operators are increasingly turning to liquid cooling, which can handle much higher thermal loads with greater efficiency. For AI data centers, where powerful processors drive both performance and heat, liquid cooling is fast becoming essential, enabling greater density without compromising stability.
Nevertheless, introducing a new cooling method isn’t as simple as flipping a switch. This is where digital twins come to life. By simulating liquid cooling systems in advance, operators can explore and compare options for retrofitting, building new, and/or finding colocation availability. With 45% of decision-makers already adopting liquid cooling and a further 19% planning to follow suit within the next year, digital twins are quickly becoming the go-to blueprint for smarter, safer cooling strategies.
This kind of foresight is critical as AI workloads continue to push thermal limits. Without it, operators face costly trial-and-error deployments, system outages, or worse, complete failure. And yet, a quarter of decision-makers (26%) still say they’d never consider liquid cooling at all.
What’s Stopping Progress?
The benefits of digital twins go far beyond cooling. 73% of decision-makers believe they are a game changer for driving technological innovation in data centers — a figure that rises to 81% among those already using them. Put simply, the more operators engage with digital twins, the more indispensable they become.
So, if the benefits of this technology are clear, why aren’t all data centers getting on board? The answer, frustratingly, is caution.
Despite 86% of operators acknowledging that failure to innovate could have serious consequences, many are still held back by concerns around cost, outdated systems, and uncertain ROI. It’s understandable, but failure to innovate with new tools is risky. Those who delay will find themselves left behind. Digital twins are exactly the kind of solution that helps operators make smarter, safer decisions. They reduce guesswork, unlock hidden capacity, and extend the lifespan of existing infrastructure, often delaying or even avoiding the need for costly new builds. Digital twins aren’t a nice-to-have anymore — they are fast becoming essential.
The Only Way Forward
AI may be driving the current energy crunch, but equally, it’s the only way forward. When used inside digital twins, AI enables better forecasting, fault detection, and capacity planning, offering data center operators not just more visibility but predictive power.
In a landscape where demand is surging and scrutiny is intensifying, digital twins offer a rare combination of performance, precision, and preparedness. They let us test what’s possible before committing resources, help pinpoint inefficiencies before they become problems, and, crucially, offer a smarter, more adaptive way to support AI without overstretching resources.
As we continue to embrace AI, whether that be in our homes, hospitals, and high streets, the pressure behind the curtain will only grow. If we want the benefits of AI to continue, data centers can’t afford to stand still. Digital twins will not just keep the lights on, they’ll help keep the AI future on track.
