Image: 123RF
Europe Economics has developed a new full delivery chain methodology for estimating the energy consumption of data centres relative to credible alternatives.
The study, produced for Britain’s Department for Energy Security and Net Zero, is aimed to examine how the growth of digital services, and the data centres that support them, affects the overall energy consumption – thus considering not only the energy required to deliver those digital services, but also the energy avoided by no longer delivering them through physical means.
The approach adopted focuses on energy consumption across the full delivery chain – unlike earlier analyses, which have focused on carbon emissions or assessed individual system components in isolation.
In the digital case, this includes the energy used by data centres, both for the IT equipment and supporting infrastructure, the internet transmission networks and the end user devices.
Have you read?
Digitalisation and AI in Europe’s energy sector open for consultation
AI’s energy-infrastructure trade-off: how photonics offers a more efficient path
For the physical alternatives, it includes the energy used, where relevant, in manufacturing, transportation, retail operations, office-based service delivery and end-user devices.
As a validation, the study presents three use cases, selected as being clear contributors to data centre growth and with clear physical alternatives, i.e. video streaming vs Blu-ray discs, ebooks vs printed books and AI-powered translation vs human translation.
For each of these, the energy consumption is calculated under a range of assumptions, with the finding that in all three use cases, the digital option either matches or substantially undercuts the electricity use of the physical alternative.
In the video streaming vs Blu-ray case, they are relatively close in term of electricity consumption. In the central scenario, streaming uses slightly less electricity than a UK-manufactured Blu-ray. However, if the Blu-ray is manufactured overseas, physical delivery results in lower UK energy use as much of the production energy is consumed abroad.
In the ebook vs printed book case, ebooks consistently consume less electricity than printed books. Even when books are manufactured overseas, the UK energy use remains orders of magnitude higher than that of a typical ebook download.
The AI translation case reveals the most dramatic contrast, with electricity use for a single AI-powered translation task below 0.05kWh in all scenarios, while the same task carried out by a human translator in an office consumes up to 120kWh.
The report says the results provide some initial evidence that digitalisation can contribute to the achievement of net zero, by reducing electricity consumption in the UK compared with physical counterfactuals for a given level of activity.
However, a major limitation of the study is that it only covers three examples out of the many use cases that exist for data centres, and it is unclear to what extent results from those can be extrapolated to others.
Thus, it is suggested that the research be extended by commissioning further work to apply the methodology to a greater number of use cases.
Among these smart meters is proposed, along with others, including cloud gaming, AI customer support, software as a service for office workers and music streaming.
