Skip to content Skip to footer

How data centres can make AI greener

Image: Adobe Stock / Irina Ukrainets

Mark Seymour, Distinguished Engineer at Cadence, discusses the role digital twin technology can play in how data centres adapt to the challenges of AI and the renewable energy transition.

The data centre industry is at a turning point: environmental regulations are insisting the industry re-examine its practices, while AI is drastically increasing demand. For instance, even a single ChatGPT query consumes nearly 10 times the electricity of a Google search. Unsurprisingly, key industry leaders, such as John Pettigrew, the CEO of National Grid, are calling for change to ease pressure on the UK’s electricity grid.

In response to AI power demands and regulations, we’re seeing data centres explore the possibilities of renewable energy to reduce grid load and improve environmental impact. Establishing renewable energy sources, such as wind energy or solar, is complicated and requires careful planning. Data centres must conduct trials in a testbed before making significant changes to their infrastructure and processes.

One strategy that can help is implementing digital twin technology into design and operations. These virtual replicas of the physical data centre can simulate designs based on new energy solutions to ensure their suitability and efficiency before implementing them in the actual facility. Let’s unpack why digital twin technology is necessary in a world that’s becoming dominated by AI.

Capacity catastrophe

The rapid advancement of AI is causing a capacity shortage, putting pressure on already stretched data centres. According to a report from CBRE, European data centre demand (511 MW) outstripped supply (467 MW) across the five largest European markets in 2023. Though AI offers enormous promise for businesses with its potential to unlock huge productivity gains, data centres face a huge obstacle because of the technology.

The massive amounts of storage and processing power required by AI algorithms are already an overwhelming challenge for data centre leaders. Facilities need a lot of energy to keep computers operating at peak efficiency. By 2026, that need is expected to soar from 49 GW to 96 GW, with 40 GW of that increase coming from AI. These numbers will only rise as AI’s exponential growth continues.

Power usage effectiveness (PUE) – a measure of how effectively a data centre utilises energy – is stagnating. Currently, scores remain stuck at 1.5, with best-in-class data centres achieving 1.1. This translates to approximately a whopping 5x energy waste in most data centres. That’s a massive inefficiency that needs to be resolved if facilities are to supply AI’s power needs without driving power blackouts.

Maintaining power in data centres is extremely difficult. People everywhere have witnessed the effects of capacity demands on the grid. California, for instance, experienced blackouts in 2017 and 2019. More recently, parts of the US have been warned that they are at risk of power outages with artificial intelligence data centres and cryptocurrency mines pushing energy demand. This brings further attention to the difficulties of preserving a dependable energy supply and power grid. However, the challenges don’t end there.

AI vs green

From networking hardware to servers and storage, data centres need a lot of energy to function effectively. To maintain cool temperatures, the average data centre uses 3-5 million gallons of water each day, the same amount of water as a city of 30,000 – 50,000 people. This is creating increasing public tensions, especially when combined with the rising energy demands. For example, plans for new data centres are being openly rejected due to mounting opposition from local governments, residents, and councils.

Furthermore, sustainability is a pressing matter for regulators and policymakers across industries. Businesses that rely on data centres can no longer shift the source of their emissions down to the supply chain. Regulations like the Energy Efficiency Directive (EED) and Corporate Sustainability Reporting Directive (CSRD) of the European Union, which require big businesses to report their carbon usage both directly and indirectly, are driving change.

To make matters more complex, AI poses a threat to data centre sustainability. Facilities want to power AI but know that relying on traditional energy sources will impede sustainability metrics. Therefore, data centre leaders are urgently searching for solutions that enable them to meet ESG targets and AI demands simultaneously.

One standout answer is renewable energy. Ireland, Singapore, and Amsterdam are already investigating alternative energy sources to power their data centres. Using renewable energy sources to power data centres will lessen the chance of blackouts and lower their carbon footprint. Doing this could also lessen the opposition to future data centre projects from the public and local governments, leading to easier project approvals.

However, the top goal for data centre managers should be ensuring that all electricity – traditional or alternative – is used as efficiently as possible in operational data centres.

Digital twins: AI and green synergy

Data centre managers want to implement efficiencies to satisfy the increasing power demands of AI while avoiding compliance issues and meeting sustainability criteria. This is where digital twins come in.

These virtual replicas of the data centre serve as a testing ground for new strategies and technologies without disrupting real operations. These higher power densities from AI technologies are more challenging to cool effectively and efficiently because the power densities require lower cooling temperatures to ensure junction temperatures are managed.

Digital twins can be used not only to evaluate the efficiency of cooling delivery in the white space but also the impact of the environmental conditions heat rejection on the environment. This is critical to avoid rising power density completely undermining free cooling strategies and resulting in spiraling energy use. Consequently, leaders possess the analysis and data to recognise obstacles and opportunities before implementing renewable energy plans.

Another advantage is that digital twins are the key to unlocking the full potential of renewables across wider operations in the data centre. Through digital twins, data centre managers can pinpoint the most effective configurations for their unique needs. These simulations can account for both cooling strategies and resource allocation, dynamically adapting to the fluctuations of renewable energy sources. Such an approach ensures that facilities maximise utilisation of clean energy while reducing their environmental footprint.

Digital twins are more than just virtual replicas. This powerful solution enhances the ability of leaders to make data-driven decisions. Digital twins offer the ability to report on key sustainability metrics, helping facility managers illustrate possible energy savings and efficiency improvements. Using real-time data, data centres can see valuable insights into resource utilisation, optimise capacity planning, and greatly reduce over-provisioning. For instance, the data digital twins provide can help spot issues like stranded capacity. By analysing these insights, operators can understand where inefficiencies lie and create a plan of action to address them.

These benefits enable more sustainable operations and can lead to an improved PUE score or other metrics that may be developed that are more appropriate for liquid- and hybrid-cooled systems. With this relentless focus on efficiency across the entire data centre, operators can more confidently esplore integrating renewable energy sources.

Sailing the AI wave

If AI is to be fuelled successfully and responsibly, purely relying on traditional energy won’t be enough. Digital twins, however, act as a beacon of light. By using this sophisticated technology to understand the performance ahead of implementation, data centres can ensure responsible resource utilisation across the entire data centre and achieve significant energy savings. They can do so, all while providing a clear picture of the facility’s environmental impact. The data centre teams that embrace digital twins will be able to face the hurdles of AI while paving the path for a more sustainable digital landscape.

Picture of Mark Seymour
Mark Seymour
Distinguished Engineer at Cadence

You may also like

Stay In The Know

Get the Data Centre Review Newsletter direct to your inbox.