How AI impacts data centre operations

The rising popularity of artificial intelligence (AI), as led by generative AI (GenAI), should give pause to CIOs, infrastructure leaders, and data centre operators. In an Andreessen Horowitz post titled, Navigating the High Cost of AI Compute, organisations may consider using the cloud as the right place for their AI infrastructure, to take advantage of less up-front costs, the ability to scale up and down, regional availability and less distraction from building own data centre.

The few exceptions to the cloud include where the company needs to operate at a very large scale, requires very specific hardware that cannot be obtained from cloud providers, or no cloud provider exists that is acceptable for geopolitical reasons.

Then there is the energy requirement that data centres need to support AI loads. Before December 2023, the threat of much greater sustainability legislation has largely been a background worry with legislators more concerned about disrupting an important new industry. Andy Lawrence, executive director of research with the Uptime Institute warns that the “green honeymoon” is coming to an end and for some.

“From 2024, new reporting laws and a toughening of requirements will enforce stricter carbon reporting in many countries. These will attempt to ensure that corporate promises are both realistic and evidence-based (see the end of the report for details), and their effects will not be confined to countries or states where there is primary legislation,” he wrote.

Asked how organisations will shift their sustainability initiatives post COP28, Abhijit Sunil, senior analyst for Forrester, believes that since data centres are a central and important part of overall corporate carbon footprint reduction, sustainability will continue to be an important aspect.

“After COP28, there was a lot of emphasis on the energy transition. There will be a lot more efforts we will see in three major aspects: procurement (energy, water and other resources), operations (efficiency improvements – another important declaration from COP28), and end-of-life (circular economy, e-waste management),” he elaborated.

For his part, Danny Elmarji, vice president for presales, Asia Pacific & Japan at Dell Technologies, says CIOs have to remain even more focused on embedding sustainability into DC operations post-COP28, where historic compromises such as the fossil fuel phase-out agreement were adopted.

Liking sustainability initiatives to a marathon, he suggests that organisations must continually adapt and advance these practices to secure long-term viability and contribute to a greener, more efficient, and sustainable future.

He goes on to add that moving to a green data centre is not just a sustainability initiative – it is a smart business decision. “When we reduce electricity consumption, we also save on operational costs,” he opined.

He added that the recommendation for companies that have yet to begin this transition is to do so immediately.

“Organisations should embrace a data-driven approach and invest in the appropriate infrastructure and power consumption solutions to enhance performance. This innovation in data centre cooling techniques is now available to all clients, as such, it is not just for the large-scale enterprises and public clouds.”

Danny Elmarji

The two faces of AI in sustainability goals

According to Forrester, AI has the potential to aid in sustainability, as well as be detrimental. Sunil cites that example where AI can be helpful in energy modelling, for analysing supply side energy data, as well as being included in tools for building management.

He concedes that running AI workloads is computationally intensive and therefore may add to how much energy data centres consume globally. He posits that even though data centre capacity and workloads have increased manyfold in the last decade, total energy consumption has remained at a plateau.

“However, we may see this trend unlikely in the future as data centres may consume proportionately more energy to run computationally intensive workloads the most prominent of which are AI workloads,” he opined.

Conceding that most organisations will continue to use the cloud to support their computing needs, Dell’s Elmarji suggests CIOs query their cloud providers about how they are addressing sustainability commitments.

Specifically, how cloud providers are optimising their infrastructure, whether their current infrastructure is modernised for energy efficiency, and how these operators are measuring their success towards meeting their sustainability goals.

AI trends in the years to come

Elmarji acknowledges that large amounts of computing power will be needed as data processing demands grow exponentially with AI adoption. “To offset this environmental impact, organisations should prioritise having greater control over data centre energy consumption via right-sizing AI workloads, compute infrastructure and data centre economics,” he elaborated.

He suggests that flexible “pay-as-you-go” spending models help organisations save data centre costs while supporting sustainable IT infrastructure. 

“AI can also address environmental challenges. We’re helping customers build smarter buildings with AI, managing building performance issues like predicting energy demand and identifying energy leaks in real-time. When combined with emerging cooling techniques, we can offset the power requirements involved,” he continued.

Navigating an uncertain future

Asked how CIOs should focus their data centre strategies given the uncertainties, Elmarji opines that transitioning to a more sustainable data centre will involve:

Tracking data centre metrics: By using real-time data and control at the device and data centre levels, businesses can identify opportunities to save energy and reduce carbon emissions.

Understanding how much power your hardware is consuming: Pay attention to thermals and cooling, optimise your equipment to reduce heat and avoid wasting energy in the cooling process.

Dropping old and inefficient technologies: Key areas to consider are power management, thermals, and processor improvements. By reducing the amount of hardware required, businesses can save costs and energy while reducing e-waste.

For his part, Forrester’s Sunil says data centres’ sustainability can be very impactful in the overall carbon footprint reduction of a company.

“Exploring the right automation possibilities including investing in building management solutions and DCIM software that can feed data into corporate carbon accounting software platforms will be critical.”

Abhijit Sunil