Infosys: How AI Can Cut Data Centres’ Energy Consumption

Data centres have become the backbone of the digital economy, quietly powering everything from streaming services to AI applications. 

But as their role expands, so does their appetite for electricity. 

That’s putting pressure on organisations and governments trying to meet climate targets.

Joseph Alenchery is SVP and Business Head of Energy Next at Infosys, the technology services and consulting firm. 

His portfolio spans strategic clients across Europe, the Middle East and India and he’s spent years working on ways to make digital infrastructure more energy efficient.

Joseph shares his expertise on data centres in this Q&A.

How much energy are data centres really consuming?

The International Energy Agency estimates data centres currently account for about 1.5% of global energy usage.

While this might sound modest, it’s important to recognise that this figure is growing rapidly, with the exponential growth of data and increasing reliance on digital services, the demand on data centres, from hyperscalers to edge computing nodes, is scaling up at an unprecedented rate.

Much of this energy is consumed not just in computing, but in cooling. Cooling systems alone can account for up to 40–50% of a data centre’s total energy consumption.

What technologies can make an impact on this?

There is a growing suite of technologies that can significantly improve the energy efficiency and sustainability of data centres.

Immersion cooling is one of the most promising advances. This technique involves submerging processors in dielectric fluids that are far more efficient at heat transfer than traditional air or water-based systems.

In fact, Infosys has collaborated with Shell to implement a synthetic, single-phase immersion cooling solution in its Texas data centres that significantly reduces energy and saves an impressive 91% in water consumption.

We’re also seeing how the move toward modular or containerised data centres brings notable energy benefits.

These modern setups can be up to six times more powerful and energy-dense than legacy infrastructures, and they offer better scalability and faster deployment. Retrofit-based energy efficiency technologies are also gaining traction.

Deploying IoT-enabled systems can also enable data centres to intelligently monitor and manage power and cooling in real time, reducing waste and increasing responsiveness to varying load conditions.

In terms of site selection and infrastructure design, leveraging naturally cooled environments can make a substantial difference too.

Through our collaboration with Daimler, we supported the organisation to move its high-performance computing (HPC) cluster to the Lefdal Mine Datacenter in Norway. Located in a former mineral mine and powered entirely by renewable hydroelectric and wind energy, this facility benefits from natural cooling and stable underground temperatures, which substantially lowers the energy footprint of the data centre.

Finally, integrating renewable energy sources such as solar, wind or hydro into the data centre’s grid helps reduce Scope 1 and Scope 2 emissions, bringing operations closer to long-term carbon neutrality goals.

How can AI be leveraged to support data centre energy use?

AI plays a growing role in making data centres more intelligent and adaptive and there are several high-impact applications that stand out. Firstly, AI-driven predictive analytics can forecast data load trends and help optimise the allocation of resources, which reduces both overprovisioning and unnecessary energy use.

Secondly, through dynamic cooling control, AI systems can monitor processor loads and adjust cooling mechanisms in real time, making operations more energy efficient.

AI can also aid in infrastructure planning by helping organisations assess when and how to scale their data infrastructure.

For example, in a healthcare setting, AI might project increased data from IoT-enabled devices and recommend capacity expansions accordingly. With the ability to tailor AI for specific industries organisations can also model data flow from various systems, calculate efficiency and forecast when a data centre might need expansion, say within 12 to 18 months.

This can help organisations determine the adequacy of their existing infrastructure.

Plus, when integrated with renewable energy sources, AI can help manage the demand–supply balance effectively, ensuring uninterrupted operations while maximising sustainability.

Why should businesses want more sustainable data centres?

AI adoption is skyrocketing and is showing no signs of slowing down, increasing data centre utilisation which could double the power consumption in data centres by 2026 in the worst-case scenario.

Not only would this mean higher costs but also a roadblock to climate change action. Meanwhile, data centre reporting rules are set to change in the EU, as part of its larger goal to achieve a 11.7% decrease in energy consumption in the decade of 2020 to 2030.

From a capex perspective, adopting solutions that I’ve mentioned such as efficient immersion cooling can deliver overall savings ranging from US$0.50 to US$1.10 per watt with potential for even higher savings.

Assuming a data centre load of 2 MW, savings can potentially range from US$1-2m at the capex stage, and US$1m per year in energy savings. At an operating level, the average power usage effectiveness (PUE) in data centres can be reduced from 1.4 to 1.2.

From a brand and trust standpoint, greener infrastructure elevates corporate reputation and strengthens relationships with customers, investors and regulators.

These developments call for significant efforts and prompt action towards making data centres sustainable and cost effective. Future data centres must be set up from the point of view of green energy supply, energy-efficient cooling and energy-efficient operations.

In short, sustainable data centres aren’t just good for the planet, they’re good for business.