Friday 13 November 2026
Shaping the Invisible
Data centres are among the defining building types of the digital age. Yet their architecture has barely begun to emerge.
Yesterday, The Guardian visited EcoDataCenter in Falun, Sweden, and asked whether this might be a data centre we can all love.1 Data centres are increasingly unpopular – their opacity makes it hard to judge what goes on inside them, and whose interests they serve. Falun has a persuasive sustainability story: mass timber, renewable electricity, and reused waste heat. But environmental performance is only one part of the equation. What makes a data centre good architecture?
»Why is it so much easier to explain why this data centre performs well than what makes it compelling as a building?«
We know how to measure a data centre’s technical performance, but we’ve barely developed a vocabulary for discussing its architectural success. There’s a reason for this. Data centres didn’t begin as an autonomous building type. They emerged within existing offices and industrial premises, before developing into purpose-built facilities, often built to standardised designs. Architectural historian Fanny Lopez and urban planner Cécile Diguet observe that few have acquired the architectural presence of railway stations, telephone exchanges, or power plants. Most remain anonymous boxes and hangars, sometimes camouflaged or decorated.2
New technologies often begin by hiding their infrastructure: underground, at the edge of settlements, or disguised as something else. Eventually, they become too consequential to ignore. Factories followed a related trajectory – architect Peter Behrens’s 1909 turbine factory for AEG in Berlin, for instance, gave industrial production a powerful architectural expression.3
The data centre presents a different challenge. Its physical presence is substantial, but what happens inside is largely invisible, as is the network it serves. A building may cover several football pitches, consume enormous quantities of electricity, and require elaborate cooling systems, yet its purpose remains difficult to read from the outside. It is a local concentration of a global system: a place where computation, electricity, cooling, security, and communications infrastructure converge.
Architects have long grappled with how buildings communicate what they are and what happens inside them. Some make their identity inseparable from their form. The architects Robert Venturi, Denise Scott Brown, and Steven Izenour called these »ducks«, after a roadside building shaped like a giant duck. Others are essentially functional structures whose purpose is communicated through signs and applied decoration. These they called »decorated sheds«.4 Perhaps the data centre calls for a third approach. Its architecture doesn’t have to imitate the technology it contains (server racks make for rather uninspiring formal precedents), or rely on symbols pasted onto its façade. Maybe it could express something less literal, such as its condition as a node, a necessary physical concentration in an otherwise virtual network.
Time introduces another dimension. The equipment may be replaced every few years, while the building and its energy infrastructure are expected to last for decades. Computing hardware keeps getting smaller and more powerful, even as demand for computation grows. Will data centres shrink as equipment becomes denser, or expand as our insatiable demand for compute outpaces its efficiency? Either way, they must be designed to accommodate successive generations of technology. What should be permanent, and what replaceable?
A global network doesn’t require interchangeable local buildings. Data centres can respond to the climate, topography, industrial history, energy systems, and materials of their sites. They can respond to the people who live next to them. The challenge is to give local form to a global function, and make that relationship constructive rather than incidental.
Falun is an interesting attempt to develop this emerging typology. Yet The Guardian’s account is revealing. It asks whether we can love the building, then gives us reasons rooted almost entirely in engineering and environmental performance: timber construction, renewable electricity, efficient cooling, waste heat recovery. No architect is named, and the design itself is barely discussed. Why is it so much easier to explain why this data centre performs well than what makes it compelling as a building?
The data centre is an architecture of the invisible. That doesn’t mean its own architecture should remain invisible. We’re building the infrastructure of one of the most consequential technologies of our time. The buildings that sustain it should be more than containers.
References
1 Steve Rose (2026) »Datacentres are deeply unpopular. Has Sweden made one we can all love?«. The Guardian, 12 November 2026. https://www.theguardian.com/environment/2026/nov/12/datacentres-unpopular-sweden-ecodatacenter-falun
2 Fanny Lopez & Cécile Diguet (2025) »An Infrastructure without Architecture?«. Chapter 3 in Digital Conquest: Spatiality, Energy and Data Centres. Bristol: Bristol University Press, pp. 23–28. https://doi.org/10.2307/jj.22992784.10
3 Stanford Anderson (2000) Peter Behrens and a New Architecture for the Twentieth Century. Cambridge, MA: The MIT Press.
4 Robert Venturi, Denise Scott Brown & Steven Izenour (1977) Learning from Las Vegas. Cambridge, MA: The MIT Press.