Cloud empires: Mapping the geopolitics of data infrastructures

In recent years, environmental concerns have strengthened awareness of the digital world’s basis in physical infrastructure. However, the enormous political and economic importance of this has not yet been fully explored.

These implications were what piqued the interest of Aalto University and Oxford Professor Vili Lehdonvirta. His European Research Council project will map hyperscale data centres around the globe, in the process examining some compelling questions. Where is AI located? And who is influencing it?

‘The digital world is not just a separate dimension in some virtual cloud. It is physical data centres, servers, cables, workers, energy and emissions,’ says Lehdonvirta. ‘Their location and centralisation will ultimately determine how policy can influence the digital world.’ 

Cloud has impact on the ground

At the heart of Lehdonvirta's investigation lies a quiet revolution. The production of digital services has moved from modest corporate data centres to sprawling hyperscale cloud infrastructures. This has real world implications for climate policy.

A recent report by the International Energy Agency (IEA) states that, while the number of internet users has doubled and the amount of data has increased by up to 20 times within the last decade, the combined energy consumption and carbon footprint of data centres has increased only very modestly.

‘Global hyperscale cloud infrastructures are more energy efficient and can also make better use of waste heat, for example in district heating networks’, says Lehdonvirta, who is interested in whether this will lead to increased “migration” outside national boundaries.

Lehdonvirta’s research also looks at how centralised computation creates new gatekeepers who have the power to decide who has access to advanced computational resources and under what terms, he says. It should not go unremarked — or uninvestigated — that these gatekeepers are large U.S. and Chinese technology firms. 

Shifting digital landscape

In Europe, concerns about digital sovereignty and data localization weigh heavily. The project will examine the way governments try to influence the geography of computing through policy, for example by attempting to locate data centres within their own jurisdiction or resisting the shift to cloud services in favour of keeping computing decentralised instead.

In some countries, such as the UK, computing is highly concentrated in the data centres of hyperscale cloud platforms, with large sections of government running on Amazon Web Services. In others, such as Germany, it largely operates on local servers residing on a public sector organisation’s own premises, Lehdonvirta explains. Meanwhile, is China spreading its influence towards Europe by sponsoring new data centres, or will its focus remain on Asia and the Middle East?

These national policies aiming towards digital sovereignty versus global influence interact with technology companies’ business strategies, Lehdonvirta observes. ‘Together they shape our digital landscape.’

‘If sovereignty has to be bought as a service, is it really sovereignty?’

Amidst all the geopolitical manoeuvring, the whereabouts of data remain an enigma for most users. This is an inherent tension in cloud infrastructures, says Lehdonvirta. The ability to replicate data in multiple locations poses a risk, but also creates an opportunity to balance load, prevent downtime and provide high-availability services.

Hyperscale cloud providers have noticed European anxieties around digital sovereignty and developed offerings that seek to address these concerns with data localization, encryption, and organisational arrangements — what Lehdonvirta calls “sovereignty-as-a-service”. 

‘If sovereignty has to be bought as a service, is it really sovereignty,’ he asks. ‘What if the additional service ceases to be provided? Is the data then available to intelligence services, for example?’ 

Lehdonvirta was recently awarded a €2.5 million Advanced Grant from the European Research Council for a five-year research project to develop methods for mapping the size and location of global cloud empires.