Across the globe, 70% of data center facility leaders say their national power grid is being stretched to its limits. Now, the sustainability warning bells aren’t just ringing, they’re deafening.
Behind much of this growing concern is the surging energy demand driven by artificial intelligence (AI) in data centers. To quantify the use of AI, McKinsey’s latest survey highlights that 78% of respondents say their organizations already use AI tools in at least one business function. From workplace productivity gains to life-saving capabilities such as detecting illnesses, AI innovations are second to none.
Data centers, which house the computing power for AI, must now focus on supporting its growth sustainably, ensuring national grids are protected. If not, we might run out of power, leading to data center outages that affect communities and livelihoods. However, according to Cadence’s Innovation Imperative, while 88% of data center operators say they’re actively working to enhance energy efficiency, only three in ten (31%) believe that they’re doing enough.
The good news is that data centers can reduce their energy impact by harnessing AI in smarter ways. AI-powered digital twins—virtual replicas of facilities—help operators shrink their environmental footprint, prevent costly outages, and finally, boost sustainability.
Ultimately, the research shows that data center operators want to make a difference. The challenge is knowing where to start. The first step is uncovering where the real problems lie and what’s truly driving excess energy use.
Product Engineering Director at Cadence.
Apprehension on Energy
Before beginning to invest in efficiency tactics, data centers must assess the challenges in their facilities.
Our latest research shows that almost two-thirds (60%) of facility leaders overprovision, which is allocating more resources to a system than necessary. This is due to concerns that scaling back will cause outages. While it’s understandable that facility leaders want reliable systems, it also wastes energy, drives up their footprint, and increases operational costs.
As energy needs rise to power AI, they lead to excessive overprovisioning, with immense energy waste.
Another challenge that Cadence uncovered is that many data centers struggle with stranded capacity. This is another unsustainable practice, where installed capacity in the data center cannot be used, and 29% of leaders reported stranded capacity as a constraint.
Picture stranded capacity like a game of Tetris, where data centers are playing five levels at the same time, trying to fit all the systems (blocks) into the data center. Operators are often unaware of doing this and can’t spot the available capacity. Thus, the facility fails to meet its design goals and has a costly impact on the planet.
Furthermore, while high-density servers are great for holding immense power, their high energy requirements can create several challenges for data centers. Currently, 59% of data center operators are using high-density servers, so it is important to make sure they run properly and effectively, with as little stranded capacity and over-provisioning as possible.
This is especially true with rack densities exceeding 100kW, and as high as 600kW, with the latest Rubin architecture that Nvidia presented at GTC.
Addressing These Matters is Not a “Nice-to-Have”
Tackling data center energy challenges is now critical, especially as regulatory factors come into play. This includes stricter reporting requirements, such as the EU’s Energy Efficiency Directive, which requires carbon emission reporting.
Local communities are increasingly opposing data center facilities. This is primarily due to claims that such facilities consume large quantities of energy, competing for energy resources and water with the population. Recently, this has become a concern in Virginia, where some residents will soon be neighbors to a 466,000-square-foot data center.
Addressing these issues requires a nuanced, multi-faceted approach. From energy reporting and thermal modeling to capacity planning and workload optimization, digital twins will play a critical role in tackling stranded capacity, reducing excessive energy use in data centers, and allowing them to trial renewable energies.
Reducing Inefficient Resource Allocation
By simulating real-time operations, digital twins enable operators to fully utilize available capacity, optimize energy consumption, and minimize the environmental impact on surrounding areas.
Minimizing overprovisioning is a good place to start to reduce energy consumption. Digital twins, enhanced by AI, offer a powerful solution. Through real-time data and historical trends, operators can create a virtual environment that mirrors the physical facility. This allows them to test different scenarios, evaluate the impact of resource allocation decisions, and identify potential areas of over-provisioning.
It also provides an excellent stranded capacity solution. By integrating sensors and data collection mechanisms, operators can continuously monitor the performance of various components, such as power consumption, cooling efficiency, and overprovisioning. This data can then be analyzed using predictive analytics to identify potential bottlenecks or areas of underutilization.
By proactively addressing these issues, operators can optimize resource allocation and reduce stranded capacity.
Reduce, Reuse, Recycle
The transformative capabilities of digital twins do not end here. Using these tools, data centers can capture and repurpose waste heat from cooling systems for other applications, such as heating buildings or industrial processes. To do so, digital twins can replicate the physical facility and help manage the implementation of technology. This will reduce energy waste and lower overall carbon emissions.
Repurposing wasted heat is important because the EU Energy Efficiency Directive mandates that data centers with a high level of energy input utilize waste heat or implement other waste heat recovery measures.
In addition, the increasing heat caused by growing server density puts cooling systems under significant strain. Digital twins allow operators to model the effectiveness of alternative cooling methods and explore how these systems interact with the entire infrastructure.
Evaluating Cooling Effectiveness
Re-evaluating data center cooling is an important strategy for reducing energy consumption. Cooling is one of the most energy-intensive elements of data center operations, particularly as AI workloads increase power demands. Digital twins are making it more feasible for data centers to adopt liquid cooling, which is gathering momentum.
At present, 45% of decision-makers use liquid cooling, and a further 19% plan to introduce it in the next year. This is largely because high-density server racks, intensive workloads, and increasing power densities are surpassing the capabilities of traditional air cooling. While air cooling can manage heat loads up to 20kW per rack, loads beyond 20–25kW are more efficiently and cost-effectively handled by a mix of liquid cooling and precision air cooling.
Using digital twins to implement liquid cooling, data center operators can examine factors that are otherwise difficult to detect or measure, such as overall cooling efficiency. They can assess the pros and cons of various liquid cooling options before investing in technology. The result is a customized solution tailored to the facility’s specific heat load requirements.
Transforming the Data Center Trajectory
Clearly, data centers are serious about improving their environmental impact. However, implementation remains the biggest hurdle. Digital twins are proving to be the sustainability game-changer the industry needs, helping operators move from ambition to action.
Even the process of deploying digital twins drives immediate value, forcing facilities to gather their data, surface blind spots, and build a clear picture of their operations. This alone creates the foundation for smarter, more sustainable decision-making.
Those that turn to digital twins won’t just optimize data center performance, they’ll unlock a roadmap to a greener, more efficient, and future-proofed data center industry.
We rate the best web hosting services.
This article was produced as part of TechRadarPro’s Expert Insights channel where we feature the best and brightest minds in the technology industry today. The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/news/submit-your-story-to-techradar-pro
