Do data centres simply increase electricity consumption or can they help make electricity systems more flexible as part of the energy transition?
© Canva
Cloud services, streaming, digital administration, data-intensive research and applications based on artificial intelligence are increasing the demand for computing power. As a result, the electricity consumption of data centres is ever-increasing. Their electricity consumption is expected to double worldwide between 2025 and 2030. This poses a challenge both to electricity generation from renewable sources and to the grid infrastructure. Are data centres primarily an additional burden – or can they help to make a renewable electricity system more flexible?
Why a flexible electricity system is becoming more important
As the electricity system becomes more reliant on renewable energy, flexibility becomes increasingly important due to greater fluctuations in electricity generation. When there is plenty of wind or solar power available, flexible consumption can absorb the extra energy. When there is a shortage of electricity or the grids are overloaded, flexible consumption can be reduced or postponed.
Flexibility is not achieved solely through storage, flexible power plants and grid expansion. Large sources of electricity consumption can also contribute by adjusting their demand in terms of time or location. This is precisely where data centres come into play. They can contribute to flexibility in three ways. These approaches are, respectively,
Market-orientated: Data centres adjust their electricity consumption in line with supply and prices.
Grid-orientated: Data centres help to prevent or reduce grid bottlenecks.
System-orientated: Data centres or their infrastructure contribute to the short-term stability of the electricity system, e.g. through balancing energy.
Not all data centres are equal
Data centres vary greatly in terms of how flexibly they can be utilised. For instance, enterprise data centres are operated by individual organisations for their own IT services and often use older infrastructure. In colocation data centres, several customers or tenants share server capacity; here, contracts and responsibilities often limit the degree of control. Hyperscale data centres are very large data centres offering cloud services; they are equipped with state-of-the-art technology, are centrally controlled and operate across multiple sites. Data centres for artificial intelligence and high-performance computing have particularly high performance and expensive hardware, which should be utilised to full capacity wherever possible.
Flexibility therefore depends not only on the technology, but also on the business model, the applications in use and the contractual framework. It is particularly important to distinguish between what is technically possible and what can actually be utilised in real-world operations.
Three approaches to flexible data centres
There are three ways to make the electricity consumption of data centres more flexible:
1. Postponing or adjusting computing tasks
The greatest potential lies in areas where computing tasks do not need to be carried out immediately. They can be timed to match the availability and prices of electricity. In this context, it makes sense to
shift tasks such as certain batch data processing tasks, backups or AI training to a later time,
relocate them to other data centres, or
scale them back or adjust them dynamically (server power scaling).
By combining various measures, considerable load reductions of up to 30 percent can be temporarily achieved. However, these figures are context-dependent and often originate from environments where operators had full control over workloads and control systems. Services such as streaming and real-time systems are not well suited to this as they require high availability and low latency.
Analysis: The potential is high if operators have access to the control systems. Obstacles include service requirements—such as speed—contractual commitments and potential revenue losses if expensive computing capacity is not fully utilised. This strategy is particularly suitable for market-driven flexibility.
2. Utilising the supporting infrastructure
Data centres have infrastructure in place to safeguard their operations, such as uninterruptible power supplies (UPS systems), backup generators and cooling systems. This infrastructure can also be utilised for a flexible electricity market. UPS systems can react very quickly and are therefore relevant for short-term system services. Cooling systems can also be temporarily adjusted, for example by raising temperature settings, which can lead to a temporary load reduction of around 5 to 10 percent to avoid grid congestion. While it is technically possible to use backup generators, this is problematic due to emissions, noise, fuel consumption and their primary role as an emergency backup.
Analysis: The potential depends heavily on the available security and cooling infrastructure. The main hurdle is operational reliability, as the availability and protection of IT systems must not be compromised. This strategy is particularly relevant for grid and system services, for example in the event of short-term load relief requirements or for balancing energy.
3. Integrating additional storage and generation capacities
Data centres can install additional capacities that provide targeted flexibility. These are the most versatile and can support functions that benefit the market, the grid and the system. These include:
battery storage,
thermal storage,
on-site generation capacities, such as generators or renewable energy systems.
This additional infrastructure does not directly interfere with IT operations, but has the advantage of being very quick to deploy. Thermal storage is particularly well-suited to data centres with modern cooling technology such as liquid cooling. This enables a data centre to respond to electricity prices, generation peaks or grid bottlenecks without immediately having to adjust its actual computing power.
Analysis: The potential is medium to high, as additional storage and generation capacities can provide flexibility independently of the actual IT operations. The main hurdles are investment costs, space requirements and the question of whether installing additional capacity is economically viable. This strategy is the most versatile as it can support functions that benefit the market, the grid and the system.
What framework can policymakers establish?
Technical possibilities alone are not enough to motivate operators to invest in flexible data centres. High investment costs in infrastructure and contractual guarantees – for example, regarding speed – act as a deterrent. This is where policymakers must step in. They can create conditions and incentives to ensure that flexibility is economically attractive and technically feasible. They can do this by
designing electricity markets in such a way that demand can participate on an equal footing with generation and storage flexibility,
introducing dynamic electricity tariffs and appropriate charges for grid access.
enabling flexible grid connection agreements so that flexible data centres can connect to the grid more quickly or at lower cost when they provide flexibility,
converting grid bottlenecks more effectively into incentives, without compromising operational reliability,
aligning the procurement of renewable energies more closely with time-based factors, e.g. through hourly matching rather than purely annual balancing.
reviewing information requirements, standards or minimum requirements for data centres,
incorporating flexibility into the planning and design of new data centres from the outset, rather than retrofitting them at great expense later on.
Without political incentives, data centres remain a drain on the electricity grid. With incentives, however, they can be a valuable building block for a future, flexible electricity system.
Prof. Dr Dierk Bauknecht is an expert on the electricity sector and Head of the Energy System Analysis & Regulation Group within Oeko-Institut’s Energy & Climate Division. He works in the Freiburg office.
