Insights: Taking AI-driven data centres into the future

The pace of technological innovations in the last decade has seen the game-changing impact of artificial intelligence (AI) and machine learning (ML) permeate all aspects of our daily lives.

This democratisation of AI in an ever-growing digital economy has also led to explosive growth in the data centre industry as the demand for cloud services and the expanding use of web-enabled devices globally are spiralling upward.

The growing need for more and larger energy-intensive data centres brings with it an exponential growth in power requirements. A 2024 report by the International Energy Agency (IEA) estimates that data centres have the potential to double their energy usage by 2026. These critical infrastructures are an important new factor of higher electricity consumption, and this accelerated demand is expected to reach more than 1,000 terawatt-hours (TWh) — roughly equivalent to the entire electricity consumption of Japan.

Data centres today consume around 1.5 per cent of the world’s electricity, which accounts for around 1 per cent of energy-related greenhouse gas emissions.

Electricity demand in data centres primarily comes from two key processes. Computing activities account for around 40 per cent of the total energy consumption, while cooling systems that are essential for maintaining stable processing efficiency, contribute another 40 per cent. The remaining 20 per cent is used by other associated IT equipment.

As of March 2024, there are more than 10,000 data centres globally. As new facilities are commissioned, growth will be driven by increased adoption of 5G networks and cloud-based services. The rapid growth of AI-related services in the past year alone and the race to build more powerful generative AI models have significantly escalated usage, as the latest chatbots and AI models are computationally intensive.

On average, a ChatGPT query needs nearly 10 times as much electricity to process as a Google search. New AI capabilities like audio and video generation are expected to further increase energy demand.

Growing energy needs of data centres

This growing appetite for power imposes an increased challenge on utilities and the energy sources they manage. They see this as a challenge to their climate pledges (when faced with the challenge of expanding operations and accessing more power while remaining sustainable and reducing their carbon footprint).

To meet the growing energy needs of both data centres and traditional businesses and industries while maintaining the agreed pathways to net zero for their respective country, utilities have been moving from carbon-intensive conventional sources such as coal and liquid fuels to lower carbon power generation such as natural gas — ideally with hydrogen capabilities — and nuclear, and deploying cutting-edge renewables and energy storage technologies, especially battery energy storage systems (BESS), known for its capability to provide short-term storage.

While these technologies are engineered to mitigate the carbon footprint and address the power density challenges significantly, an often-overlooked aspect is that of AI’s ‘water footprint’. For instance, even a short conversation comprising roughly 20 to 50 questions and responses with ChatGPT requires 500ml of water to cool down the servers. This amount could vary depending on the type of power plant and the region, as water usage is influenced by local weather conditions and the cooling technology employed.

This significant water demand is driven by the cooling processes of both power generation and data centre operations. To offset the intense heat generated, most data centres are air-cooled in a process known as swamp cooling, which uses large amounts of water to reduce the temperature of the surrounding air.

Data centres are therefore turning to novel techniques such as liquid cooling which chills the equipment directly, and immersion cooling, a new liquid bath technique where servers are submerged into tubs of coolant to improve their performance and energy efficiency while reducing water usage. Rear-door heat exchangers, on the other hand, encourage better running conditions in dense server environments.

Energy efficiency and greenhouse gas emissions have become mainstream in the sustainability strategies of data centres, but to truly build toward sustainability, water must be part of the equation too, especially as a mid-sized centre in the US uses about 300,000 gallons of water a day, equal to the water consumption of 100,000 homes.

In 2027, the accelerated global AI demand is expected to account for 4.2 to 6.6 billion cubic metres of water withdrawal.

Solving the ‘trilemma’

Solving the power, carbon, and cooling trilemma in data centres will be key going forward, as providers work towards fulfilling the demand reliably and more sustainably. A promising, more modern and sustainable technology to power even the largest data centres with a significantly smaller carbon footprint than reciprocating engines, all the while complementing renewables, are aero-derivative gas turbines, like GE Vernova’s LM series technology, and even heavy-duty gas turbines like GE Vernova’s H-Class. Their fuel diversity attribute allows these turbines to run on various alternate fuel sources, including hydrogen, blends of natural gas, and associated gas, supporting data centres on their sustainability journey, including the path towards net zero carbon emissions and low water consumption.

In addition to advanced power generation, grid technology also plays a critical role in ensuring the reliability of power supply in data centres, which is crucial for their continuous operation. By integrating advanced grid systems, data centres can maintain a stable and uninterrupted power flow, even in the face of fluctuations or disruptions in energy supply. This reliability is vital for preventing downtime and data loss, which can have significant financial and operational consequences.

Furthermore, grid technology can enable seamless integration with BESS and renewables, supporting the decarbonisation efforts.

The sustainability challenges associated with data centres will have implications in the Gulf countries too with the UAE and Saudi Arabia on their path to becoming regional AI superpowers. With a 15.3 per cent increase in live IT capacity, reaching 235.3 MW, the UAE is the top data centre hub in the Middle East and North Africa (MENA) while Saudi Arabia’s 29.7 per cent rise in live IT capacity (109MW) positions the kingdom as the fastest-growing data centre market in the region.

The future of data centres is inextricably linked to the future of energy, and in a water-scarce region like the Middle East, there is a greater urgency to shape a more sustainable digital landscape by integrating sustainable practices into data centre operations.

Prioritising modular and energy-efficient principles, and improving resource utilisation through zero water cooling, recycling, and low-carbon energy sources will power the growth of data centres in a way that aligns with national and global sustainability goals.

The writer is the CTO of EMEA, Gas Power, GE Vernova.

GE Vernova has recently published a whitepaper focused on data centres, that can be downloaded here.