“Reliable and Cheap”?: Examining the adoption of the cloud by European news media companies

Abstract

Since the middle of the 2010s, an overwhelming majority of European news media companies have adopted cloud infrastructures. While a growing number of scholars and policymakers are raising concerns regarding the implications of cloud dependence, a systemic exploration of why media companies decided to undergo such infrastructural transformation is still missing. Building on 41 semi-structured interviews with news media professionals in France and the Netherlands, the present contribution shows that the move to the cloud was justified by the supposed reliability and cheapness of these infrastructures. Unpacking what is meant by those categories, the present paper shows how the infrastructure, labour practices, and workforce of news media companies has adapted itself to the cloud, normalising new forms of infrastructural dependence. The paper then goes on to show that conceptualising these changes through the categories of reliability and cheapness inscribes them within the traditional socio-technical imaginaries of media infrastructure, foregrounding their invisibility. However, doing so also obscures the unequal power relations between cloud providers and news media companies.

Introduction

In the 2023 Annual Report of DPG Media, the biggest media group of the Benelux, there is a small bullet point that reads: “we have closed the data centre in Kobbegem. The function of this data centre has been taken over by cloud hosting. This is more efficient and sustainable, because our cloud supplier uses renewable energy” (DPG Media, 2024: 79). The bullet point is part of the section on sustainability goals, between the shift to electric cars for company vehicles, and a renovation of corporate offices. Nowhere in the document is mentioned that the closing of the data centre culminated DPG Media’s migration to the cloud. This case is emblematic of the whole industry’s approach to this digital-technical environment. By and large, the sector has adopted it as an upgrade of their old infrastructure, without necessarily questioning the implications of this process (Wyss et al., 2025).

However, a growing number of voices are raising the alarm regarding the risks of moving into the cloud. At its core, this technology allows to access computing services, both hardware and software, over the internet. Yet, its development and commercialisation raises urgent questions. ¹ The concentration of the cloud market around an oligopoly of three American companies – Amazon, Microsoft, and Google, also known as hyperscalers – is seen as particularly problematic in the light of their repeated abuse of dominant market position to further their corporate goals (Luitse, 2024) or U.S. political interests (Holt, 2024). Researchers within and outside academia highlight that cloud dependency is both a threat to national sovereignty and creates asymmetrical power relations, hindering the economic development of countries outside the U.S. (e.g. Baur, 2026; Baykurt, 2025; Bria et al., 2025). Building on these perspectives, media researchers have pointed out that uncritical cloud adoption poses specific problems to the news media sector. These include economic dependence (Radsch and Montoya, 2024), growing inter-sectorial inequalities between corporate and independent media (Ferrari Braun and Cath, 2026), and structural challenges to the informational sovereignty of countries in the U.S. sphere of influence (Ortiz Freuler, 2025).

Yet, while the body of critical research on cloud dependence keeps growing, a key question remains unanswered: why have news media adopted the cloud in the first place? Like any other infrastructure (see Larkin, 2013), adopting the cloud requires companies to restructure their organisation around the affordances of the new system (Cath, 2025; Narayan, 2023). Understanding why they decided to undergo such process, and how it has changed media companies, is crucial to both situate the problems mentioned above, and provide solutions that are adapted to the needs of the sector. In other words, any critical enquiry on, or potential alternative to, the dominant infrastructural paradigm in the media needs to understand how we got here in the first place.

Building on 41 semi-structured interviews with French and Dutch media professionals, the present paper shows that European news media companies think of the cloud as more reliable and cheaper than other alternatives. However, these characterisations are only true insofar the organisations are willing to undergo thorough restructurings to fit their activities to the affordances of this system. In this context, the usage of categories like “reliable” and “cheap” positions the cloud within the socio-technical imaginaries of media infrastructures, stressing its capacity to be invisible, but also obscures the institutional transformations and power relations intrinsic to cloud adoption. France and the Netherlands are typical examples of the different European media ecosystems (Hallin and Mancini, 2004), and interviewees worked in 13 outlets representing the different types of organisations composing the news media industry, including Public Sector Media (PSM), media groups, and independent media companies. Drawing on interviews with employees from these different organisations, the present article shows that cloud migration in the news media is systemically discussed through terms that downplay its disruptive nature, obscuring its effects and therefore rendering critical reflection on this infrastructure harder.

The following two sections positions this paper in the scholarly discussion around digital infrastructures and journalism, and introduce the research methodology. Subsequently, the analysis is structured around the two main categories mobilised by interviewees when explaining the move to the cloud, namely reliability and cheapness. Finally, the concluding section shows how the mobilisation of these categories inscribes the cloud within traditional socio-technical imaginaries of media infrastructures, which in turn downplays the disruptions and new power relations embedded in this infrastructure.

Current research

The cloud has an ambiguous position in the scholarship on digital journalism. The field of platform studies has paid attention to the infrastructural dimension of the relationship between news media companies and digital platforms. Scholars focus particularly on how Very Large Online Platforms (VLOPs) have become key channels of news distribution, monetisation and funding, granting them considerable power over the industry (e.g. Nechushtai, 2018; Papaevangelou, 2024; van Dijck et al., 2018). The legal scholarship has followed a similar route, with researchers arguing that contemporary media law is unfit to address the infrastructural power of VLOPs (Seipp et al., 2024; van Drunen, 2025). Yet, the specific role of the cloud within these platform ecosystems is rarely conceptualised, either on the academic literature or on media regulation. Journalism studies, building on the field’s long tradition of newsroom ethnographies, has shown that cloud-based software services like Piano and Chartbeat are transforming journalistic work (Christin, 2020; Petre, 2021), but they do not necessarily problematise the specific role of cloud infrastructures in this process. All these perspectives have come together to explore the impact of AI in news media (Dodds et al., 2026; Hartley et al., 2023). Research has shown that the rapid roll-out and adoption of this technology was made possible by the prevalence of cloud infrastructures (Simon, 2022) but scholarship’s focus still lies on understanding the nature of AI and its effects on the production, distribution, and monetisation of news, rather than on the underlying infrastructure making it possible. Looking to the field as a whole, it is easy to find mentions to the cloud, but papers on this specific technology remain rare.

However, as researchers working in critical software studies and internet governance have extensively shown, the impact that cloud infrastructures have on the companies that adopt them is far from negligeable. The cloud is a system to access software and hardware services over the internet. Rather than buying commodities, companies can access them as services for a fee (Ruparelia, 2023). This, in turn, fosters the development of agile production environments where the digital infrastructure of companies is constantly being tested and upgraded (Gürses and van Hoboken, 2017). Software services – like the aforementioned Piano, Chartbeat and an extensive suite of AI solutions – are easier to access through the cloud, but the companies who adopt this infrastructure become dependent on their cloud providers to perform mission critical tasks (Benlian et al., 2018; Chander and Gürses, 2024). In the long run, the outsourcing of internal expertise is detrimental for organisations, as they come to rely on a handful of providers – mostly American hyperscalers – and an associated ecosystem strongly shaped by said providers (Luitse, 2024; Widder and Kim, 2025).

Organisations with a public-service mission, like news media companies, are particularly affected by these changes. Cloud environments are designed to maximise profit and do not adapt themselves well to non-market activities (Cath, 2025). Migrating to the cloud means adopting a wholly new production environment, outsourcing the maintenance and upgrade of critical infrastructure to a third-party, and accessing most digital functionalities as services rather than commodities. In other words, it represents a seismic change in the systems and internal structures of the organisations adopting them, which is why the cloud is present in one way or another across virtually all the contemporary scholarship on digital journalism.

Taking stock of these developments, a number of scholars have started to research the effects that cloud environments have on the news media. This endeavour can be seen as part of the wider “infrastructure turn” in journalism studies (Young and Hermida, 2025), which is seeking to situate the production and consumption of news within the much wider and globally connected digital information ecosystem. Notable work in this line includes the research of Sjøvaag et al. (2025) on the ownership structures of all the technical systems necessary for the functioning of Norwegian news media companies, including the cloud. Ortiz Freuler (2025) undertook a similar approach for the Latin American news ecosystem, putting the accent on the predominance of U.S. firms in the technological stacks of the region, which represents a significant risk for the informational sovereignty of the countries who have historically been on the receiving end of Washington’s imperialism. Stressing the outsized market power of hyperscalers and their history of anti-competitive practices, Radsch and Montoya (2024) showed how much journalistic organisations have come to rely on cloud services, paving the way for the establishment of the Journalism Cloud Alliance (Geber and Szyszlican, 2025). Finally, and perhaps most notably, Jennifer Holt’s seminal book Cloud Policy (Holt, 2024) showed that the cloud is an infrastructure at the core of contemporary media ecosystems whose very nature was shaped by lack of consequential regulation and behind-the-scenes deals between hyperscalers and the U.S. security apparatus. These contributions have played a major role in charting the risks of cloud dependency for news media companies, both within and outside the imperial core.

Still, even within this specialised body of literature, the reasons of why news media companies adopted the cloud in the remain vague. A popular explanation is the fields’ well-known tendency towards isomorphism, meaning that once leading actors adopt a technology the rest are very likely to follow suit (Caplan and Boyd, 2018; DiMaggio and Powell, 1983; Simon, 2024). This tendency has been proven with many technologies before the cloud, and there is no reason to believe that the adoption of this infrastructure would follow a different logic. However, considering cloud’s ambiguous presence in current scholarship, its problematic nature as a production environment, and the myriad of dangers associated with cloud dependency for news media, it is worth exploring the matter more in depth. Analysing the reasons put forward by the workers of news companies when explaining their cloud migration contextualise the elements presented in this section, and ground them in the practices of news organisations. It is a necessary step to understand what the cloud means for the (news) media, situate this transformation within the wider power dynamics of the sector, and think of infrastructural alternatives.

Methodology

The present paper builds on 41 semi-structured interviews (Aberbach and Rockman, 2002; Harvey, 2011) conducted with news media professionals from France and the Netherlands in 2024. Interviewees can be categorised in four groups: managers, developers, journalists, and policy experts. ² Managers and developers were selected based on their familiarity with their company’s infrastructural strategies. Journalists and policy experts were subsequently interviewed to complement the insights provided by the other two groups (e.g. if a manager and a developer were interviewed, further interviews were conducted with journalists from the outlet and with policy experts familiar with, or representative of the outlet). 13 different outlets participated in the research: six Public Service Media (PSM), three media groups, and four independent outlets. The transcripts of the interviews were coded using Reflexive Thematic Analysis (Braun and Clarke, 2019). Interviewees were guaranteed anonymity in the informed consent form that they signed prior to the interview, some quotes presented below were therefore slightly altered to ensure that they could not be traced back.

The two case studies, France and the Netherlands, are representative of the different media systems present in Europe (Hallin and Mancini, 2004). The French media environment is characterised by a broad system of press subsidies in place since 1947 which, despite its significant shortcomings, has allowed many outlets to preserve their independence (Lardeau and Floch, 2013). While France is undergoing a significant process of media concentration, the landscape remains more diverse than in the Netherlands. The Dutch media ecosystem is characterised by an extreme level of concentration in the private sphere – where two Belgian companies, DPG Media and Mediahuis own a majority of the countries’ outlets – somewhat attenuated by a robust PSM both at national and regional level. France and the Netherlands therefore represent two distinct ways of organising the news media market while being subject to the same EU regulations. Despite these differences in market composition, the proportion of PSM, media groups, and independent media of each country was kept roughly similar in the research design. The empirical findings in the following section focus on similarities across the entire spectrum of both countries. By engaging with different outlets, with different size, means, procurement processes, and goals, it is possible to assert that the infrastructural dynamics presented below are representative of the entire sector, and not just a specific corporate form or a given market.

The research was therefore designed to ensure both breadth in terms of represented outlets, and depth in terms of engagement with the specific characteristics of each country. There are some limitations to this method, still. The reliance on interview data and the lack of examination of internal documents or ethnography of their everyday practices make it difficult to ascertain certain dynamics. Likewise, France and the Netherlands are representative of Western European dynamics, characterised by a very strong dominance of hyperscalers in the cloud provision market and, up until recently, a generalised trust towards American technology and infrastructure providers. It is possible that Eastern European countries, who have their own history and market composition for both media and tech, have a different relationship to the cloud and its providers.

Reliable and cheap

When asked whether they would ever consider switching back from a cloud-based infrastructure to a local one, the CTO of a media group left no space for doubt:

Having a physical location, with actual server racks running, is very expensive and not scalable at all. I didn’t even think about the full picture until we had a flooding in our offices and realised that these things don’t take water very well. [. . .] There’s no going back, no. It’s definitively not a thing that we would want, definitely not. From a cost perspective but also. . . It’s actually costs and reliability. Doing hardware updates yourself. . . Like Jesus, really?

This answer encapsulates the reasoning of the news media industry when it comes to cloud adoption. These systems are more reliable and cheaper than the alternatives. However, reliability and cheapness are much more complex notions than this nonchalant answer might led to believe. Unpacking their meaning will show how these categories hide significant operational changes while also redefining the expectations around media infrastructures.

Reliable

“I don’t know if somebody spilled water over it or if it was a leak, but it was a major logistical issue” (CTO, media group), “I remember that around 2017 or 2018 we had a collapse, our provider had an outage and that’s when we realised we had to be more resilient” (manager, PSM). The CTO who provided the opening quote to this section was not alone in conjuring visions of infrastructural collapse to justify the move to the cloud. Water damage and power outages were put forward time and time again to illustrate why it was necessary to change infrastructure. Yet, as an independent media developer pointed out, there are other ways to manage hardware that are not cloud-based. These mundane illustrations of machinic breakdown are useful to visualise why media organisations seldomly operate their own servers, ³ but fail to explain why they chose the cloud. Cloud infrastructure is seen as more reliable because it allows media companies to outsource their infrastructural operation while meeting changing industrial standards, often times set by cloud operators themselves.

The reliability of the cloud is chiefly constructed around its capacity to dynamically scale computational provision to meet instances of “peak load.” Peak loads are moments of heightened user demand that does not correspond with regular usage. Cloud infrastructures have different ways to scale-up the available computational power of a given organisation, from adding more computers to the organisation’s cluster to adding more power to the existing cluster (Ruparelia, 2023: 22–23). As such, they can easily meet unusual demand, whether it is expected or unexpected. One example provided by the CTO of a PSM were elections: “we know that we will have lots of people between 19:30 and 20:00, so we start scaling-up from 17:00 onwards.” Peak loads and how to manage them are one of the central tasks of media CTOs and developers.

The CTO of a media group put it in a succinct way when they said:

From a technological perspective, the most important question is to manage the peak loads, the moment when there are lots of readers in a very short time. It’s the riskiest moment, but it’s also the most important moment for the image that we give to our audience, for the advertisement revenue, for getting subscriptions. It’s our main infrastructural issue, having a website that is always available to readers, regardless of the intensity of the news.

Peak loads are conceptualised as the litmus test of infrastructure, the moment where the media is both fulfilling its journalistic mission by informing the public, while signifying to audiences that it is capable of delivering quality information, which can hopefully be leveraged to obtain more subscriptions or advertising revenue. In a dynamic attention economy, breaking news or seamlessly broadcasting a highly anticipated event (e.g. elections, sports) are moments where media brands build their reputation, cementing themselves as trust-worthy institutions to which audiences can subscribe, and where advertisers can market their productions. It is critical that the infrastructure works during those crucial instances.

The dynamic nature of cloud environments makes them particularly well-suited to these high-risk, high-reward moments on pressure on the infrastructure. As another CTO, this time from an independent media company, argued:

The strength of the cloud is that they are capable of adjusting all the machines, all the available power, in under a minute. We saw it a couple of months ago, when we broke a big political story. We were the first ones and our number of views went through the roof, because Google News picked us up. The website bugged for around 45 seconds, I was getting all sorts of phone calls, but it was solved before I could even write up a ticket to flag the issue. If we were working with a classic hosting company, it would’ve been a catastrophe.

What the CTO is describing here can be essentially boiled down to a new form of labour division, whereby most of the work required to manage the infrastructure is performed by people external to the media outlet. The advantage of the cloud is not simply having more power at your disposal, but rather outsourcing the maintenance and adaptation of your infrastructure to external organisations, which is able to provide a much quicker and dynamic service than media companies would be able to get either from their own workforce or through other infrastructural arrangements.

Reflecting on these differences, a fourth CTO commented,

When we were with our previous hosting company, we had to make them a ticket requesting the creation of a virtual machine. [. . .] We were living in a different time, in the 1990s. Having to make a ticket to ask your host to create a virtual machine brings me back to the 1990s.

Compared to other solutions, the cloud offers a dynamic production environment where the needs of the users are much more expediently addressed. Their vertical integration of the different services involved in the maintenance of infrastructure, coupled with their capacity to further invest in facilities, available machines, maintenance personnel, and user interfaces, make the experience of using cloud infrastructure much more seamless than other infrastructural arrangements. However, it should also be noted that industry standards – what it means to be living in the 1990s as opposed to living in the 2020s – are also intrinsically shaped by the companies offering cloud services.

The nature of peak loads is not neutral. Since the development of mass media there have been fluctuations in the publics’ attention; the news media have developed strategies to deal with them for the better part of the last two centuries. Yet, the fact that millions of people can rush towards one web page in the space of a few minutes speaks to the changes in the distribution of news operated by platformisation. For example, a CTO mentioned that one of the causes for unexpected peak loads were news aggregators, such as Google News, picking up a news item from an outlet’s webpage. In turn, the public has come to expect near immediate loading of webpages, particularly when searching for breaking news. As a developer commented, achieving this level of latency (i.e. delay in charging a web page) without using the services of hyperscalers is hard. The digital news ecosystem has accustomed audiences to constant availability of information, forcing news media companies to adapt themselves to this new reality. Adapting means producing content that fits the logics of platforms, but also having an infrastructure that meets the audience’s expectation of seamlessness. Herein lies the fundamental tension of cloud infrastructures: they are productive environments that outsource infrastructure provision, allowing companies to keep up with industrial standards that are largely set-up by cloud providers themselves.

Reliability – the first element of the answer to the question why news media moved to the cloud – is a complex category. Cloud systems are perceived as more reliable because they are better equipped to deal with instances of peak demand, which they do by outsourcing the labour of maintaining and upgrading the infrastructure. At the same time, the nature of those instances of peak demand is shaped by platform ecosystems where standards are set by the same companies that dominate the cloud provision market. Reliability essentially means that media companies can pay other parties to maintain key aspects of their infrastructure, including keeping up to date with the industrial developments pushed by those parties. In turn, this outsourcing allows for a restructuring of the workflows and workforce of media companies, which is why the cloud is perceived as cheaper.

Cheap

There is an overlap between the reliability of a system and its economic performance. Having a reliable infrastructure means avoiding crashes, last-minute makeshift solutions, and improves the audience’s perception of the outlet. Yet, when people say that the cloud is cheaper, they often refer to the fact that you “build only what you need, instead of paying for a lot and ending up using 10% or 20% of it” (developer, PSM), a sentiment echoed throughout the interviews. Building only what you need is a deceptively simple way of describing a restructuring of infrastructural provision and associated working practices. It is premised on a decoupling between the backend and the frontend of media organisations, where the cloud provider takes care of the former and the technical teams of the media company focus only on the latter, specifically on activities perceived to be value-generating. In other words, the cloud is cheap(er) because it allows news companies to (1) centralise their infrastructure, (2) focus on specific technologies rather than on the whole stack, and (3) hire cheaper workers.

A key aspect of this transformation is the streamlining of the digital environment, which allows companies to pool all their resources, centralising provision and reducing overhead costs. This dynamic is particularly evident in the case of media groups. A manager illustrated this point when explaining why it is commercially interesting for independent media to be bought by their company:

[when new media outlets are acquired by the group] they benefit from the knowledge of the engineers, they benefit from the standardization and of course they benefit from the investments that we made. You get an extremely fast compute, for example. So if you previously run compute on your own machines. . . you can now use a cluster of multiple nodes and have millions or trillions of data analysed in real time.

The CTO of a different media group explained the same process from the perspective of the corporation:

Our strategy is to build a single [cloud-based] platform for all our properties. In the past we either built different platforms ourselves, or acquired titles with their own technical infrastructures [. . .] but we progressively moved to a system where there’s one unique platform for all our brands. There’s an economic reason to this strategy change: it is more advantageous to have one platform operated by one team than duplicating activities over ten different ones. This allows us to rapidly answer to exogenous changes. For instance, when Google changes its algorithm, we sometimes have to adapt our technologies. Doing it once is much more convenient than doing it ten times over.

By moving into a cloud environment, media companies can develop one common infrastructure that is then deployed to all outlets simultaneously. There are economies of scale at play here, benefitting bigger organisations that are able to invest up-front in the development of a common platform. That dynamic also applies to the PSM. Both the French and the Dutch public service used to have a decoupled architecture where different channels, frequencies, and sometimes even programmes, would have their own autonomous website with an autonomous supporting platform. Over the last years that approach has completely disappeared, replaced by unified cloud-based platforms serving entire organisations.

These cloud-based media platforms are based on a decoupling of the backend and the frontend. Developing and maintaining hosting facilities is very expensive, particularly when they are regularly subject to stress tests through peak loads, and they do not directly bring revenue. Succinctly put by a PSM manager: “we don’t have the money, and the off-the-shelf solutions do it better.” On the other hand, frontend applications, such as Content Management Systems (CMS) or advertisement placement software, do significantly contribute to both the image of the outlet(s) and their capacity to attract revenue, either through subscriptions or advertisement. Asked about their strategic priorities at an infrastructural level, the CTO of a media group focussed exclusively on frontend applications:

In the consumer market we need to focus on driving more value for the consumer. Off-the-shelf often doesn’t allow that, [. . .] we need to think about what features and stuff we can add to increase the value perception of our product. [. . .] Then, in the advertising market, we have the necessity to move out of the commodity market [—].We’ve built products in advertising and tech that are not for sale anywhere else, so we’re currently moving out of the open internet commodity business, and trying to create our own little ecosystem.

The economic reasoning here is plain to see: by outsourcing their backend needs to cloud providers, media outlets are able to focus on technologies that they see as generating value. This exclusive focus on the lucrative part of the company’s activities comes with a reconceptualisation of the role of technology in the media. Rather than ensuring production and diffusion, functions outsourced to the cloud, media technologies are understood as tools to improve the user experience, both for audiences and advertisers. Fittingly, the CTO quoted above defined this strategy as “re-inventing the core.”

The strategic shift of the infrastructural focus is not simply a matter of hardware and software, it also affects the workforce composition of media outlets. A paradigmatic example is the outsourcing of maintenance staff:

For 24/7 availability you need staff available 24/7, and we don’t have unlimited financial means. I would need at least five people, five specialists to provide that service. When people are working for us, not as temporary staff but really working for us, it’s hard to get rid of them, to fire them. That’s a lot of wages, and if our financial situation were to change due to the political situation, it could be a problem. That’s why the organisation told me “we would like you to have a supplier provide those services and have a good service agreement”. (manager, PSM).

Instead of having five people on the payroll, media outlets can simply hire cloud providers that not only take care of the machines, they also deal with the temporary or permanent contracts of the people operating them. Paying only for what you use also extends to human resources. This is far from a minor issue when considering that backend developers are among the most expensive in the market. A different manager in a different PSM company commented that one of their main infrastructural issues was the fact that their platform was built on Java:

Java developers are among the most expensive developers in the IT business. The tech stacks of bank and insurance companies are based on Java, which makes them very expensive. We’ve made a transition to a different provider which would allow us to focus much less on backend and much more on front end technologies. The IT industry has a much greater variety of frontend developers that are very well versed in JavaScript, CSS styling, model forms of HTML but also node JS and stuff like that. So salaries are more competitive there.

For decades the media industry has struggled to attract IT talent, given that outlets can seldomly compete with the salaries offered by other sectors. The overwhelming majority of interviewees noted that recruiting tech talent used to be a problem, but it has become progressively easier. By offering a standardised environment, where elements such as programming languages can be changed, the cloud lowers the bar for media to hire technical profiles.

Yet, the separation between frontend and backend is not complete. In-house developers often connect the two sides of the infrastructure. Their task is crucial from a financial perspective because, if left completely unattended, the automatic scaling up and down of compute provision afforded by the cloud can easily make prices rise exponentially: “We set up an API to supply the servers, but instead of getting a 200, the server created a 500 machine” (CTO, independent media), “somebody writes a faulty query, it’s send throughout the entire infrastructure and we end up paying quite a lot, it happens often” (developer, media group). . . Interviewees repeatedly provided examples of small but expensive mistakes due to the system’s propensity to automatically scale up when requested. To reduce these mishaps, media outlets hire programmers with cloud-specific expertise that are able to improve the system’s architecture and troubleshoot problems, when the inevitable issue presents itself. Often times, this means people with experience working with the ecosystem of one specific provider, which – given that hyperscalers dominate the cloud market – often means people who have experience working with Google Cloud, AWS or Azure. The technical expertise available within a news organisation, and its workforce composition, is durably shaped by the organisation’s choice of cloud provider.

The idea that, with the cloud, you “build only what you need” has much more structural implications than what might seem at first. Conceiving an infrastructure in terms of discrete needs, rather than as a technological stack, is the business expression of the outsourcing logics on which the cloud is premised. If a third party can take care of backend needs, media companies can exclusively focus on their front-end, paying attention to elements such as user experience rather than, for instance, the material technology necessary for broadcasting. Infrastructure moves from a fixed investment necessary to operate – but without immediate commercial or journalistic benefits – to a service purchased from a third party. This shift not only changes the technical operations within a media company, it also changes its personnel. Expensive backend developers or maintenance personnel are becoming increasingly rare among media companies, instead being replaced by frontend developers with training or experience with specific cloud ecosystems. The role of these new workers is either improve the core products of the media outlet, or manage the integration between the in-house frontend and the outsourced back-end. Far from evident, building only what you need requires rethinking what media infrastructure is, and what type of expertise is necessary to develop and operate it.

Such approach echoes the observations put forward in the previous sub-section regarding “peak loads.” In both cases, interviewees look at their infrastructure from the perspective of their business operations, either focussing on key moments of heightened demand or expressing a clear will to centre their activities on the aspects of the infrastructure that they see as profitable. If the cloud is reliable and cheap, it is mainly because this infrastructure is premised on an externalisation of key technical tasks, which in turn changes the way media companies think about their technical systems.

Discussion & conclusion

The answer to this paper’s research question – why have news media companies adopted the cloud – is deceptively simple: they did so because it was more reliable and cheaper than other alternatives. However, cheapness or reliability are not neutral categories. Their meaning is co-constructed by the different actors involved in the process. The cloud is seen as more reliable because it is a scalable system that can adapt itself to peak loads, temporary moments of heightened demand that put a considerable level of stress unto the digital infrastructure. The cloud is seen as cheaper because it decouples backend and frontend, allowing media companies to focus on their value-generating activities and lowering the threshold for talent acquisition. In other words, the reliability and cheapness of the cloud depend on a handful of affordances designed to outsource the provision of infrastructure as a service. By invoking them, interviewees were situating their transition to the cloud within traditional socio-technical imaginaries of (media) infrastructures.

The critical literature on media and cloud calls these categories into question. While the cloud tends to be more reliable than other computational arrangements, the centralisation of its provision also means that relatively small problems can scale up into infrastructural breakdown, as evidenced in the 2024 Crowdstrike outage (Alegre, 2024; Ongweso, 2024). Perhaps more worryingly for news media organisations, American cloud providers have a long history of collaboration with the U.S. security apparatus (Holt, 2024; Hu, 2015) and Washington could mobilise the infrastructure’s affordances to punish countries or media operations that they perceive as unfriendly (Ortiz Freuler, 2025). This literature suggest that the cloud is not as reliable as portrayed by interviewees, and the same applies for its supposed cheapness. Hyperscalers have a long history of abusing their market power and engaging in anticompetitive practices (Luitse, 2024; Widder and Kim, 2025). The market dominance of a handful of American players grants them an oversized influence over the pricing logics of media infrastructure around the globe (Geber and Szyszlican, 2025; Radsch and Montoya, 2024). Yet, these concerns were barely mentioned by interviewees when describing their migration to the cloud and the decision-making process leading up to it.

In describing the cloud as reliable and cheap, interviewees mobilise socio-technical imaginaries (Jasanoff, 2015) associated with digital infrastructures, stressing the system’s capacity to conform to expectations of invisibility. As shown in Susan Leigh Star’s pioneering work (e.g. Bowker and Star, 1999; Star, 1999; Star and Ruhleder, 1996), informational infrastructures are designed and operated to produce a self-effacing effect, allowing agents to interact with them without having to think about them. In the specific case of news media companies in the 2020s, the aspiration to invisibility in the socio-technical imaginary of infrastructures is further compounded by two factors. First, the contemporary experience of navigating the digital space is characterised by a widespread expectation of seamlessness, itself nurtured by the cloud’s obscuring of its own technologies and labour processes (Hogan and Gonzalez Moserrate, 2024). By using the cloud, news media companies hope to offer a smooth user experience where visitors do not have to consider to underlying infrastructure allowing them to access content. Second, the migration to the cloud took place in a context of crisis for news media companies, who saw their business models and socio-cultural status being durably disrupted by digital platforms (Nielsen and Ganter, 2022; van Dijck et al., 2018). The cheapness of the cloud signifies a modest technical system that does not cost much but allows news media companies to further develop their value-generating activities. As such, it is rhetorically separated from the grandiose infrastructural projects common in both France (Stern, 2012) and the Netherlands (Tameling and Broersma, 2013) before the platformisation of news production and consumption. Reliability and cheapness echo traditional imaginaries regarding how informational infrastructures should work, rendered more salient by the conjuncture in which cloud adoption took place. Looking at the data through the concept of socio-technical imaginaries helps understand why interviewees decided to focus on the properties of the infrastructure, rather than on the shortcomings and potential dangers that are common in the critical literature. In this community of practice, invisibility of infrastructure is both a desirable feature and the product of intense work.

However, the invisibility of infrastructure is a complicated category. On the one hand, the experience of modernity is mediated by so many different technical systems that being aware of all of them would be maddening, if not flat out impossible (Laak, 2023). At the same time, it can be strategically used to obscure and sediment contingent relationships of power (Collier, 2011; Larkin, 2013). By making technical systems opaque, the relationships between agents that they foster are progressively naturalised, “learned as part of membership” (Star, 1999: 381), and depoliticised. In the context of cloud adoption by the news media, this process could potentially lead to a situation where the extreme inequalities between cloud providers and news media companies fade to the background, as a feature of the system rather than the contingent and problematic result of the infrastructure’s development and roll out.

The invisibility of the cloud is an aesthetic feature of the modern infrastructural ideal, but it is also part of the wider political project of hyperscalers (Hu, 2015; Plantin et al., 2018). The outsourcing of the media’s mission critical infrastructure to these providers grants them an outsized power that is seldom reckoned with. In addition to the explicit security and economic threats mentioned earlier in this section, the likes of Amazon, Google and Microsoft use their infrastructural power to shape innovation in discrete ways. For example, the discretionary roll out of Generative AI on different software services helped create a narrative about widespread adoption of this technology, thereby creating a self-fulfilling prophecy (Hao, 2025). Likewise, one could argue that infrastructural invisibility was strategically leveraged by cloud providers to avoid regulation. As Seipp et al. (2024) have shown, digital infrastructure remains largely absent from media law, despite the disproportionate concentration of its provision and the potential threads to media freedom and opinion pluralism that such concentration entails.

It is very interesting to note that the only interviewees who did express worry over the power that hyperscalers were accumulating over the media worked for independent media. These included the two outliers of the whole corpus: the only company that was using OVH instead of a hyperscaler and the only outlet whose infrastructure was not cloud-based. Again, the technical and the political are intertwined in this decision. The cloud’s affordances of scale (Narayan, 2026) disproportionately benefit larger media groups, as argued above. At the same time, those same media groups have often argued that they need to scale-up their operations to compete with digital platforms, effectively asking for less comprehensive media concentration regulations (Sjøvaag, 2023). The socio-technical relations between media groups and hyperscalers raise questions about the technological processes underpinning the transformation of the digital public sphere. It is important to understand the socio-technical imaginaries of infrastructural invisibility when discussing why news media companies migrated to the cloud, but researchers must keep emphasising the politics of the cloud, bringing them to the fore, and problematising them from the perspective of the different agents affected by this infrastructural change.

This paper has shown migrating to the cloud has important consequences for the internal operations of news media companies, far beyond the boring technicality presented in the introduction’s opening quote. Some of these consequences are relatively minor. Having better systems to handle peak loads is not a massive change, even though it shows that news companies are further aligning themselves with the attention logics of a platformised public sphere (Poell et al., 2022; Smyrnaios, 2025). Others are much more consequential. The cloud’s affordances push companies to focus on their value-generating activities – whether it is to produce more journalism or to sell more subscriptions and advertisement. Such productivist outlook contrast with other visions of media infrastructures for the public good that were once popular. The idea that the media industry could help co-construct a regulatory framework for a digital system, as the French newspapers did for the Minitel (Mailland and Driscoll, 2017), or develop open source CMS through a public broadcaster, like the Dutch VPRO did with Mmbase (Sanders, 2024), now evokes a very distant past. Independent media developing their own autonomous infrastructure to preserve their capacity to do journalism free from corporate pressures, a distinctive sign of projects like IndyMedia (Platon and Deuze, 2003), has become positively impossible in the age of hyperscalers. In this context, the outsourcing of infrastructural maintenance and upgrade to third parties is not simply a choice of convenience, it marks news media’s acceptance of a limited role in contemporary politics of technology, focussed on a handful of core activities and leaving the development of technical systems to others.