Co-opted Transitions Locking in our energy systems
Silvia Weko nov 2026 · essay · issue 02
Donald E. Davis, interior of a Stanford torus. Public domain.

Big Tech has been making the headlines for its thirst for energy and water. With the rise of LLMs and the data centre boom, these companies have clearly demonstrated how something that sounds so intangible — the cloud — can have severe environmental consequences. But there is another lesser-known side to this story: beneath the physical infrastructure of wind turbines, solar panels, and batteries, lies a sprawling digital infrastructure of data and code, an infrastructure that relies on cloud hyperscalers like Google, Amazon and Microsoft (GAM). At a time when Big Tech firms are seeking to expand their energy supply, clean energy systems are, in fact, themselves increasingly reliant on Big Tech.

Part of what is driving this is that decarbonised energy systems are complex, and running them requires storing and transmitting far more information than older energy systems did. Fossil-based energy systems produce energy in a fairly instantaneous manner when needed by burning fuel:a car, for instance, is just filled with petrol, and a power plant can simply burn more coal when demand for electricity increases. By contrast, clean energy systems are based around electricity — they are therefore far more interconnected, and need to utilise or store energy whenever it is plentiful. Since the sun does not always shine and the wind does not always blow, generated energy levels are bound to fluctuate constantly. In addition, renewable energy systems are more geographically distributed. Instead of a few large power plants, there are hundreds of thousands of small assets: solar panels, wind turbines, tidal barrages, and so on. And instead of a few central nodes sending power out to the edges of the system, many small nodes at the edges can send power back into it. For instance, with vehicle-to-grid technology, an electric vehicle’s (EV) battery can be charged when it is sunny and solar energy is abundant; then, later in the evening, some of that power can be sent back into the system when electricity demand spikes.

Moreover, because decarbonising means electrifying the entire energy system, managing demand becomes all the more complex. If your EV software is trying to make an optimal decision about whether to charge its battery or send power back to the grid, it needs to have information about energy supply and demand — across households, transport, heating, and heavy industry — right now, later today, tomorrow, and maybe even next week. Access to this information is crucial because large drops in supply or spikes in demand can lead to grid instability and blackouts.

Coordinating renewable energy systems thus involves gathering and storing millions of data points, then analysing them using machine learning techniques to extract meaningful statistics. This is part of what policymakers sometimes refer to as the “twin transition”, where the digitalisation and sustainability transitions reinforce one another.1 This is well known to energy-sector actors, who have long understood the importance of data. For example, Vestas, one of the leading wind turbine manufacturers, gathers detailed data on how its turbines produce electricity to flag issues, respond to weather patterns, and develop better technology.2

Increasingly, however, energy companies have been outsourcing this data collection and analysis to Big Tech’s digital infrastructure. In 2011, Vestas had the world’s third-fastest commercially-owned supercomputer. In 2021, they began using Microsoft Azure to run the parts of their analyses involving reinforcement learning models; by 2024, they had entirely replaced their on-premises supercomputer with Azure. Vestas is not alone in this: most energy actors today rely on the cloud to run their operations.

Indeed, electric utility companies around the world — public and private alike — have become increasingly reliant on Big Tech’s digital infrastructure to manage power production, consumption, and planning. Consider the following examples: in the United States, two of the largest electrical utility companies, NextEra Energy and Duke Energy, rely on Amazon Web Services (AWS) and Google Cloud for grid monitoring, planning, and forecasting; in Spain, the primary electrical utility, Iberdrola, also relies on an Amazon-developed technology; and in Chile, the national grid transmission operator, Eléctrico Nacional, is using Google X’s “Tapestry” initiative to simulate the grid and plan future expansions.


Why should we care about this? Does it really matter if energy sector actors use GAM’s digital infrastructures? Isn’t this a good thing, if it allows them to “innovate” more quickly? After all, one could read this as a more efficient division of labour, where GAM provide digital infrastructure — their area of expertise — allowing Vestas or similar firms to focus on their own innovations without wasting resources recreating the whole stack.

The issue becomes more apparent when we consider the problem of dependence. Once a business starts using GAM’s digital infrastructure, it becomes rather difficult for them to leave. This is by design, and GAM employ a variety of tools to ensure vendor lock-in. A first barrier, for instance, often involves charging prohibitively high fees or ensuring that contracts entirely prevent exit for a certain period of time. But the primary obstacle is often technical. By designing tools and systems that lack interoperability and are specific to their stack, GAM make switching providers prohibitively costly. For example: if a company wants to train a model that simulates wind farm performance, it would need to purchase access to custom chips, like AWS Trainium, which can only be communicated with in a specific programming language — Neuron SDK, in this case.3 This, in turn, encourages companies to also use Amazon SageMaker, a software-and-platform hybrid service that makes communicating with the chip easier. Because SageMaker is a black box, its users cannot access or reconstruct the foundations on which their technologies are built, making it even harder to exit.4 Once these actors are captive, Big Tech firms can raise their prices, demand more data, more favourable contracts, or whatever else they want.


Using their discursive and political influence, Big Tech tries to ensure that firms and governments plug into their services and then get locked in. This dynamic has been evident in their interventions in sustainability discourse, where they have attempted to build a Gramscian “common sense” about how society should be organised and the problems it faces, and to present their technologies as key solutions. Part of this framing involves offloading responsibility for climate change onto individuals. Indeed, Big Tech actors often highlight that their technologies can help people make better choices about their consumption habits.5 The technologies these firms control and profit from are thus framed as a solution to ecological crises. Individuals are urged to save energy and water by using Google and Amazon’s smart devices or find recycling centres using Alexa; businesses are urged to switch their legacy IT infrastructures to the cloud, which, according to the hyperscalers, has a lower carbon footprint. The solutions they promote also move in tune with their latest investment interests. For example, Google’s sustainability reports used to promote reducing the number of cars on roads, providing public transport information on Google Maps, and encouraging users to commute together with Waze Carpooling. After Waymo — Google’s self-driving electric car project — took off, focus shifted to EVs as key to decarbonising transport, now downplaying the importance of public transportation.6

While trying to orchestrate public discourse, these firms also lobby governments to take specific measures that promote their worldviews and further their interests. At COP30, for instance, Amazon and Google had representatives promoting AI technology as crucial for addressing the decline of biodiversity and enabling the conservation of nature.7 Big Tech companies also exercised significant influence over key proposals by the Brazilian COP presidency, such as the Climate Mutirão initiative, a digital platform to improve collaboration between governments, funders, and stakeholders.8

Another strategy employed by Big Tech to further their penetration into markets involves offering free access or reduced rates. To get smaller businesses such as start-ups to use their services, many firms offer incubator programmes that reward winners with “cloud grants”, a large number of credits that can be spent on their services. This ensures that promising new technologies will exist within — and depend upon — their ecosystems. AWS’s clean energy accelerator, for instance, awards up to $100,000 in cloud credits and is currently in its fourth funding round, while GenAI startups using AWS’s chips receive $300,000 in credits.9

Governments and public agencies are roped in using similar tactics. Microsoft offers free credits to government users, and free pilot programmes for the US Department of Defence/War. Google, meanwhile, provides a free 30-day trial of Google Workspace to public sector organisations. A blog post entitled “Boost your agency’s efficiency: Take Google Workspace for a test drive”, which touts the 30-day trial, reads:

“Public sector agencies are under constant pressure to do more with less. With shrinking budgets and expanding responsibilities, the need for efficient, secure, and modern tools has never been greater.”10

The irony, of course, is that tech companies themselves help create this pressure for public sector agencies to do more with less. For example, to lobby the EU, AWS commissioned a 2022 study which concluded that “moving 10% of government IT systems to the cloud could save EU taxpayers over 900 million euros a year”.11 Similarly, Google commissioned a report arguing that widespread adoption of AI in public administration would boost the efficiency of government processes and increase productivity by ten per cent, equivalent to 100 billion euros annually.12

In the sustainability sector, Big tech firms have been pushing to loosen energy permitting regulations for years. Today, they are increasingly also pushing for governments to use their technologies to do so. Take the example of Easypermits, a cloud-based tool that aims to “help public administration and local communities […] embrace and speed up the whole permitting process”. The tool is a collaboration between AWS, WindEurope, and Accenture, and was recently piloted in Denmark and tested in Poland.13 Recent lobbying around the EU Grids Package and Energy and AI Roadmap also pushes for digitalised planning and permitting using AI.


Big Tech firms have also begun to explore ways of profiting from sustainability transitions by directly creating and deploying their own innovations. These often use public data and resources, which they subsequently enclose and monetise. A key example is Google Earth Engine, a tool to enable the analysis of Google Earth data through its cloud AI. During an initial experimental phase, Earth Engine was offered for free. The goal was to accumulate data, which Google did very effectively, going from having around twenty-five petabytes of data in 2019 to around fifty-two petabytes and over one thousand datasets by the end of 2021. This data was gathered while the tool was used for different projects to map risks and environmental harms, including by non-profit organisations and universities such as the Global Surface Water Explorer, the Global Forest Watch, and DiSARM for malaria risk mapping. Google then curated these datasets in its Earth Engine Data Catalogue, which was freely available and had over ninety-thousand users. In 2022, once it had amassed valuable data and dedicated users whose labour helped improve the platform, Google turned its Earth Engine into a paid cloud service.

Amazon has also engaged in similar practices. The Amazon Sustainability Data Initiative, for instance, collects public data provided by institutions like the UK Met Office and NASA and encourages researchers to contribute more information. This is all then combined with proprietary internal data and used to enhance Amazon’s energy management technologies and optimise its renewable energy production and consumption in over 30 countries — which is crucial for profit generation today, given AI’s rising energy demands.14 When they don’t need it, these companies can also sell power: The New York Times estimates that tech firms’ subsidiaries sold around $2.7 billion on the wholesale electricity market in the past decade, most of it since the AI boom in 2022.


Put together, all of this demonstrates how and why sustainable energy transitions are imbricated with Big Tech’s digital infrastructures. Positioning themselves at the most valuable part of the chain, Big Tech firms can manage information flows and monopolise knowledge, leveraging their advantage over other actors in the clean energy field, who tend to be smaller and have lower profit margins and less capacity for investment. Through this involvement in energy transitions, they can further consolidate their position in the global political economy and expand their already enormous economic power. Even if these firms suddenly decided they were not interested in directly profiting from the clean tech market, their relationship with energy actors would still give them enormous structural power, since energy is one of the backbones of societal functioning. Beyond energy, Big Tech’s deepening control over crucial infrastructure already extends to sectors such as banking, logistics, telecommunications, and the military. Indeed, as recent outages demonstrate, there are few areas of the world which are not, in some way, reliant on GAM.15

This is why it is more productive to think of the cloud as infrastructure, rather than as a service. If infrastructures are objects that create the grounds on which other objects operate, then the cloud is a kind of digital infrastructure that creates the grounds for the physical infrastructures that enable society to operate.16

What, then, is to be done about this reliance on Big Tech? Do energy systems really need AI and the cloud? Can’t we decrease dependence on AI by building energy systems that don’t need as much data, and by relying on energy sources — like nuclear — that fluctuate less than renewables? I believe that we can decrease the dependence on data, but probably not do without it. Decarbonised energy systems are complex enough that large-scale data collection and processing power will always be required to coordinate them efficiently. Imagine, for example, a larger city without private cars, where transportation relies on a large fleet of electric buses, taxis, or shared vehicles. Significant data-processing capacity will still be needed to coordinate the charging of all vehicles, avoid major spikes in demand that can destabilise the grid, and plan for future consumption.

The problem is less that sustainability transitions are reliant on digital infrastructures; it is that these infrastructures are owned and controlled by Big Tech. So long as that remains the case, these firms will try to steer transitions in ways that benefit them. Big Tech firms are profit-maximising actors. They will, therefore — whenever possible — raise prices for locked-in dependents, like wind turbine manufacturers, transmission system operators, and utility providers. This will inevitably be reflected in our energy bills. In addition, Big Tech firms will likely continue to expand their own presence in energy-related markets. Rather than public transport, we will get more self-driving electric vehicles, à la Waymo. Ultimately, the benefits and risks of energy transitions will be even more unequally distributed than they already are.

If we understand the digital stack as infrastructure, we can conclude that it is a natural monopoly — inherently difficult and inefficient to break up. As with electricity, it is more efficient to have a single actor run and coordinate the grid. But such monopolies cannot be private because they will inevitably seek to capture value from their dependents. Smaller cloud alternatives, that have recently gotten a boost from European efforts at digital sovereignty — like France’s OVHcloud and the Schwarz group — are therefore no panacea. Even if they could compete with U.S. Big Tech, it would involve using the same playbook of lock-in and value capture.

The only solution I envision to this problem is public digital infrastructure, as Cecilia Rikap and others have argued: a democratically controlled stack.17 The alternative would be to continue allowing tech firms to expand their reach into areas like sustainability transitions, and shape them according to their own interests. For researchers and activists to whom renewable energy promised a step away from fossil-fuel monopoly power, this is not a cheerful message. But understanding sensors, chips, cables, and code as monopolised infrastructure is, ultimately, a crucial step in pushing back against Big Tech.

Notes

  1. 1. Joint Research Centre, “The Twin Green & Digital Transition: How Sustainable Digital Technologies Could Enable a Carbon-Neutral EU by 2050”, European Commission, 29 June 2022.

  2. 2. Palle List Clausen, “How Does Combined Data Add Value to Our Customers’ Assets?”, Vestas (blog), 27 July 2020.

  3. 3. Open-source, but controlled by AWS.

  4. 4. Dieuwertje Luitse, “Platform Power in AI: The Evolution of Cloud Infrastructures in the Political Economy of Artificial Intelligence”, Internet Policy Review, 2024.

  5. 5. A particularly fun example of this is Amazon touting the sustainability of online grocery shopping.

  6. 6. Cecilia Rikap and Silvia Weko, “A Green Transition Orchestrated from Big Tech Clouds?”, Globalizations, 2026.

  7. 7. Michael Pooler, Beatriz Langella and Attracta Mooney, “Muted Global Business Turnout at COP30 Reflects Subdued Mood”, Financial Times, 10 November 2025; see also Google, AI for Nature: How AI Can Democratize and Scale Action on Nature, November 2025.

  8. 8. André Corrêa do Lago, “Primeira Carta do Presidente da COP30, Embaixador André Corrêa do Lago”, COP30 Brasil, 10 March 2025; COP30 Brasil, “Global Mutirão: The COP30 Aims to Go Beyond Negotiations to ‘Heal’ the Planet”, 22 May 2025.

  9. 9. Amazon Web Services, “Clean Energy Accelerator”, 2026.

  10. 10. Katie Tobin, “Boost Your Agency’s Efficiency: Take Google Workspace for a Test Drive”, Google Cloud Blog, 13 December 2025.

  11. 11. Amazon, “Digital Transformation Can Grow EU Economy by 2.8 Trillion Euros”, About Amazon, 16 June 2022.

  12. 12. Martin H. Thelle et al., “The AI Opportunity for eGovernment in the EU: Adopting AI in Public Administration to Enhance Productivity and Create Value for Citizens and Businesses”, Implement Consulting Group, 2025.

  13. 13. WindEurope, “EasyPermits”, 2026; WindEurope, “Wind Energy Permitting Is Improving but Governments Still Have Work to Do”, 8 February 2024.

  14. 14. Silvia Weko, “New Sites of Accumulation? Why Intangible Assets Matter for Energy Transitions”, Review of Political Economy, 2025.

  15. 15. Patrick Smith, Ben Goggin and Steve Kopack, “Amazon Says It Has Resolved Internet Services Outage That Left Websites and Apps Lagging around the World”, NBC News, 20 October 2025.

  16. 16. Brian Larkin, “The Politics and Poetics of Infrastructure”, Annual Review of Anthropology, 2013.

  17. 17. Cecilia Rikap et al., “Reclaiming Digital Sovereignty: A Roadmap to Build a Digital Stack for People and the Planet”, LUT University, 4 December 2024.