Could AI-enabled data centres save Europe’s ski season?

Assessing the impact of artificial intelligence on the carbon footprint of data centres.

Key takeaways:

• Europe’s Christmas ski season was abysmal.
• Climate change has real-world impacts; data centres have an impact on the climate change discussion.
• The artificial intelligence space is growing faster than ever.
• AI could be used to make thermal models 40% more efficient while also boosting process efficiency.
• High efficiency operations could result in less power consumption, reducing the overall carbon footprint of a data centre.

Raise your hand if you tried to go skiing over Christmas and were faced with a mountain of mushy greenery? Sights like the picture below were not uncommon across this winter, with a vast amount of Europe’s go-to resorts not being able to accommodate skiers until late January.

A slippery start to the year.

The position of the data centre industry in the climate discussion is not insignificant, and with more and more advancements in internet accessibility the conversation will only continue to grow. The industry has proven time and again to be conscious and innovative when it comes to sustainability.

Nowadays, data centre clients and tenants are putting more significance on sustainability requirements in the sites that they occupy, while data centre operators are also exemplifying circularity and sustainability because of their investors and the desire to differentiate themselves from the competition.

Most of us have heard of ESG; by now, most of us have also heard of ChatGPT. In fact, some of us may have already integrated AI chat-bots into our daily workflows as personal assistants, coders, or financial planners, as a way of smoothening our daily lives and making ourselves more efficient. Could the same be applied to data centres?

Let’s have a look.

On thermal management and efficiency

Studies on thermal management of data centres have shown that proper application of artificial intelligence could result in up to 40% higher energy efficiency compared to standard consumption.

In some instances, the application of artificial intelligence was made in the conception of the thermal management model, running various simulations to identify the best method of AC/AH unit placement, RMCU placement, and application of load-balancing software to ensure the reduction of hot zones which typically leech cooling power and reduce efficiency (see here: Research Advances on AI-Powered Thermal Management for Data Centers). These algorithmic conclusions boost cooling efficiency and avoid unnecessary workload strains.

In comparison to AI-led thermal design, active AI-enabled thermal scheduling has also been seen to improve efficiency by constantly monitoring thermal schedules alongside IT load to efficiently and proactively balance thermal management. The proper application of DCIM, sensors and AI/ML could enable a reduction of power consumption, improved workload distribution, reduction of outage risks, and ultimately an improved carbon footprint. Google’s DeepMind AI is a great example of achieving operational improvements, having improved their PUE by 15% upon application of AI/ML.

Clients, especially MSP and cloud providers, also bear responsibility to maximise the efficiency of their compute applications. Although this is arguably a topic in itself.

On digital twins

All of this is tied in through a cleverly designed system of ‘digital twins’. A digital twin is essentially a digital copy of the data centre, running real-time calculations and providing analyses of operations and making real time adjustments to improve efficiency. Not only does this have cost advantages, but also contributes majorly towards sustainability by ensuring the proactive adjustment of assets to deploy workload in areas that are really in need; water consumption and WUE have also seen improvements.

Although quite a fresh concept, advancements in digital twins could also improve Scope 2 CO2e emissions by reducing the travel needs of technicians and engineers. The real- time virtual models of data centres essentially enable maintenance, adjustments, and operational excellence works to be done remotely, or with remote advice. An indirect impact on carbon footprint, but an impact nevertheless.

Where are we now?

The data centre industry is notoriously risk-averse. New technologies, such as liquid immersion cooling, take time to be adopted on a wide scale. Operators of data centres wish not to risk their reputation and their clients’ data by implementing unproven technologies. However, some of the leaders of the industry have been rolling out AI/ML models since the better half of the last decade with good results.

New data centres are at an obvious advantage: without the need to retrofit AI into a well lubricated mission-critical system, there is more room for innovation and freedom of choice when implementing relatively new technologies.

More has to be done in the data centre centric AI space to determine whether or not it can help us save Europe’s melting ski season, but with OpenAI’s ChatGPT taking over global headlines we can assume that the AI discussion will only grow over the coming years.

In short, AI-enabled data centres can:

• Improve thermal management design,
• Improve active thermal management,
• Improve efficiency and PUE,
• Improve allocation of resources, such as water or personnel,
• Have direct and indirect impact on the carbon footprint of a data centre,

But can’t save the Swiss Alps just yet.

CIRKLA’s data centre circularity model

At CIRKLA, we enable data centre professionals around the world to contribute to circular economy by providing a wide range of checked 2nd Life infrastructure assets; enabling CapEx savings, slashed lead times, and a reduction of your carbon footprint.

Our unique Data Centre Circularity auditing and certification program consolidates your circularity metrics against international standards and builds trust between you and your clients. Click here to find out more.