articles
issues
support
about
Sovereignty After the Boom From dependency to digital dependency
Vinit Ravishankar sep 2026 · essay · issue 02
Abandoned weaving factory in Americana, São Paulo. Photo: Leonardo Salvato. CC BY-SA 4.0.

At this point in the data-centre boom, we have considerable empirical data on how these infrastructures are funded: how money flows, what sorts of deals are made, who owns what, and so on. Considerable recent journalistic work has also revealed how both these infrastructural buildouts and resistance to them are increasingly a profoundly global phenomenon that spans geographies and nation-states.1 There are many similarities as far as the political economy of these expansions is concerned. Particularly outside the United States and China, leaders of the so-called AI arms race, data centre expansions can be viewed as an attempt to win out from the generative AI boom/bubble through the ownership of material infrastructure. At the same time, the specificities of how expansions are coordinated and legitimised make for a useful lens to examine the politics of how national governments view their roles in a global web of accumulation.

Thus, both practically and discursively, there are considerable differences between a data centre in East London and one in suburban São Paulo or Mumbai. This essay is about the latter of these. As things stand, two of the prime recipients of data centre investments in the global South are Brazil and India, with expenditure estimated to surpass $45b and $126b respectively, by the end of the decade. Given the historical and political similarities between the two countries — key members of BRICS with massive domestic markets, regional hegemons, and “sub-imperial” powers in their own right — the similarities in how they justify these infrastructural expansions stand out. Both countries, attempting to explicitly build a rhetorical connection to their own historical struggles against imperialism, have become visible champions of an increasingly widespread yet murky notion of digital sovereignty.2 Such discourse on the importance of digital independence is not unique; similar rhetoric is commonplace in the European Union, for instance. In the post-colonial world, however, this discursive hijacking of an already fraught concept fits into a longer history of abortive efforts to escape colonial domination — and is likely to end the same way.

While vast differences exist between the colonial relationships that shaped modern India and Brazil — the former a classic instance of British administrative governance, and the latter of a settler-colony in the Americas — there have been a number of points of convergence in the post-colonial era. In particular, local political elites’ desire to avoid the dependency that characterised the colonial period led, post-independence, to plans for economic coordination inspired by core insights from the dependency theory tradition. The umbrella term, dependency theory, refers to a fairly diverse set of analyses, including, for instance, the Prebisch-Singer thesis, which held that the value of primary commodity exports would rapidly fall relative to the costs of manufactured commodities; or Samir Amin’s notion of an imperial rent drawn by the developed monopoly corporations situated in the global North. Nevertheless, the prescriptions that emerged as a result tended to converge onto some form of import-substitute industrialisation.3 Both India and Brazil thus resorted to strategic trade protectionism, blocking imports of manufactured commodities in order to kickstart domestic industrial production.

Such policies proved to be something of a failure in both countries, particularly when compared with the dynamism of East Asian economies. The reasons for this were manifold. On the one hand, imports did not cease, but simply shifted from finished commodities to industrial machinery, trapping these countries even further into dependency. On the other, neither country could take advantage of the economies of scale that export-oriented countries with weaker currencies could; their industrial projects thus floundered, never attaining profitability, and resulting in a widening of the gap with the global North.4 These experiments definitively drew to a close in both countries by the early 90s. In Brazil, while brutal labour discipline and export-oriented industrialisation were a mainstay during the military dictatorship, it was Fernando Henrique Cardoso’s Plano Real that consolidated a neoliberal Brazil, fully integrated into global systems of accumulation.5 In India, the Congress government, facing a balance-of-payments crisis in the late 80s, was forced into opening the economy to foreign investment in 1991. This was a slower process than in Brazil, yet succeeded in eventually firmly embedding the country into global accumulation, providing the material basis for the Indian National Congress’ (INC) dominance from 2004 to 2014.

As a result of this integration into the world economy, foreign direct investment (FDI) inflows into both countries grew drastically as they began actively courting foreign investment, offering a number of financial benefits and access to cheap labour and resources. The distribution of the gains that economic liberalisation enabled was heavily contested. In Brazil, Lula’s PT was somewhat successful at managing this distribution, particularly during the commodities boom that coincided with Lula’s second term. Yet numerous clashes with grassroots struggles that hamper accumulation have been a recurring feature, such as with the Movimento dos Trabalhadores Rurais Sem Terra, or the Movimento Passe Livre.6 In India, the INC were able to form a relatively stable governing coalition for a decade, partly due to working-class discontent with the BJP’s first full government term. The National Rural Employment Guarantee Act helped offset the crisis in the countryside, but eventually ran out of steam amidst a series of corruption scandals.7 The predominantly services-driven nature of FDI in India, meanwhile, resulted in domestic consumption being driven by a small fraction of workers, failing to create mass employment and to absorb the large volumes of labour displaced from the countryside.8

What does any of this have to do with digital sovereignty? Today’s massive data-centre buildouts in both countries require some degree of legitimisation, or at least some pretence of convincing citizens that these environmentally devastating infrastructural expansions are in their interest.

This has created an interesting tension. On a discursive level, these digital sovereignty programs borrow heavily from dependency theory’s heritage: narratives of independence, freedom from “digital colonialism”, data localism, and so on, are a solid part of this package.9 Yet simultaneously, the actual political economy of these buildouts bears little parallel to import-substitute industrialisation, whatever its flaws: today, foreign investors are explicitly encouraged to invest in constructing domestic infrastructure and leasing access to local firms, should it please them. And to state the obvious, it is abundantly clear that neither Brazil nor India have the capacity to develop and manufacture the hardware or the chips that contemporary artificial intelligence requires. The ability to do so remains the sole preserve of Taiwan’s TSMC, with the machines that enable chip production made by the Dutch firm ASML. Should fairy-tale visions of a world still ruled by free trade persist, it is worth underlining that access to these chips depends entirely on the largesse of the United States, where chip restrictions on China have bipartisan consensus. And beyond hardware, neither India nor Brazil realistically have the ability to train large foundation models that might rival ChatGPT or Claude. Contemporary models have moved on beyond merely training on text scraped from the internet, and today make use of a myriad of different labour patterns to acquire more “valuable” data, which remains proprietary to OpenAI or Anthropic.10 Competitive entry into this space, short of adoption by a Big Tech firm, is virtually impossible.

Faced with these obvious stumbling blocks, notions of the development of domestic productive forces have fallen by the wayside. Conversations have instead shifted to discussing territorial claims to data, or the utility of domestic data centres to some nebulous sovereign “AI technology” wrapper built on top of existing foundation models. Lead firms have cheerfully played an active role in encouraging these narratives to develop, positioning themselves as eager to facilitate these countries’ renewed ambitions for sovereignty, fully cognisant of the fact that these demands for sovereignty simply serve to accelerate a demand for their services.11

Yet this should not be viewed as a monolithic imposition from abroad. After all, the immense volumes of capital flowing into data-centre construction guarantee convenient opportunities for accumulation amongst the local bourgeoisie. These include a coterie of landlords and construction firms. Conveniently, they also include energy companies, who foresee massive expansions in power-generation infrastructure to keep up with the demands of energy-guzzling GPUs. In Brazil, the recent REDATA law suspends federal taxes on the import of components feeding into data centres, provided that they rely on “low-emission” energy sources, which is effectively a roundabout tax write-off for the energy sector.12 In India, domestic capitalists are playing a more active role. The two richest men in the country, Mukesh Ambani and Gautam Adani, compete to facilitate both the construction of the actual data-centres and the import of the chips that go into them, striking fixed-term energy commitment deals with Big Tech firms to guarantee cash flows into the future. The digital bourgeoisie that steers the growing tech sectors in both countries, and plays a critical role in legitimising these narratives, also stands to gain. It is precisely this class fraction that benefits from the peculiar combination of a nebulous sovereignty and immense inward FDI, given that the end result is likely to be a form of economic protectionism in the realm of value-added software services, combined with foreign infrastructural investment in more expensive material hardware assets. That this setup exacerbates structural dependency, while claiming to counteract it, means little to a class that stands to profit from it.

In keeping with this strange duality and looking towards global affairs more generally, the way these buildouts are coordinated has also shifted somewhat. In the post-import-substitute industrialisation period, companies would simply open overseas operations in Brazil or India, offloading their obsolete machinery to eke the last bits of productivity out of them, by taking advantage of cheaper pools of labour.13 Today’s data centres require far less labour, making wage suppression or currency arbitrage far less meaningful beyond the construction phase. Moreover, regimes of ownership and profitability have shifted since the 2008 financial crisis, birthing a period where large-scale asset managers play a larger role in coordinating production and consumption.14 The key agents coordinating infrastructural expansions today are therefore more likely to be the likes of BlackRock or Carlyle than Volkswagen or Suzuki.15 These asset managers are universal owners that eschew active portfolio management, and are drawn towards assets whose return profiles have low volatility, guaranteeing bond-like returns over longer timeframes.16 In Brazil, Ascenty — Latin America’s largest data centre provider, which is backed by the asset management company Brookfield and the real-estate investment trust Digital Realty — recently announced $1.2b in investments in four new data centre projects near São Paulo. In India, Blackstone has signed MoUs promising $6b in investments in two such projects in the Mumbai Metropolitan Region.

Given the general reluctance on the part of asset managers to invest in large infrastructural expansions without some form of protection from intrinsic risks, Brazil and India have grown to be formidable de-risking states, attempting to create the conditions to ensure that global financial capital is comfortable making long-term investments into tangible infrastructure.17 This goes beyond rapid deregulation of access to land, labour, and energy: both countries now offer a number of tax exemptions for data centre investors. Key to this process is the legitimisation of digital sovereignty as a discourse, as an active goal that the public sector seeks to put into practice by appealing to capital. Yet de-risking is, of course, not an elimination of structural risk, but a shift in who bears the brunt of it. In both countries, it is ultimately taxpayers, indigenous communities, and landless peasants who carry the costs as their tax contributions are siphoned off, their lands are expropriated, and they are thrust into proletarianisation in a world that has little need for their labour.

The irony of it all is that the presence of data centres is anything but the development of domestic productive forces. Despite the expectation that some of the newly built domestic capacity be reserved for domestic use, it is unclear what this will lead to, beyond short-term profits for the domestic bourgeoisie, and perhaps a task-specific wrapper built upon citizen data. Nor are these likely to be particularly sovereign wrappers, given that they will be built upon American foundation models and run on American chips. This is a reality which the discursive drift in sovereignty has helped create the conditions for, legitimising the idea that development stems from uncritically inserting domestic actors into speculative global supply chains, often at the cost of the very citizens that sovereignty was meant to defend in the first place.

What is the point of all this? Following the integration of the entire planet into the “planetary factory”, the economic strategy of many states in the global South has involved resorting to increasingly hare-brained schemes in the hope that they might transition “upwards” through the world-system. The only way to further the domestic accumulation of capital, the logic goes, is to play their cards right every time a Schumpeterian cycle presents itself.18 One of the bleaker consequences of this has been the death of more radical imaginings, and an erosion of the idea of the global South itself, as parts of the planet simply turn into sacrifice zones that fuel the forces of industry elsewhere.19 Other parts, like India and Brazil, end up with a sizable bourgeoisie that recognises its own interests, a middle-class that is perennially at risk of falling through the cracks, and a mass of cheap, disposable labour that underwrites the living conditions of the other two classes. It is also precisely these classes that have become the prime driving force behind both bolsonarismo and Hindutva, two of the world’s most dangerous and successful far-right movements. These include deeply-entrenched agribusinesses in Brazil and a crony-capitalist faction of billionaires involved in coordinating every facet of the Indian economy, together with an urban middle-class whose consumption standards have risen in tandem with these blocks.20

The sovereign AI boom is likely another iteration of such hare-brained schemes. Stripped of all revolutionary meaning, sovereignty becomes little more than a sad imitation of the Californian ideology wrapped in nationalist robes.21 In practice, it serves to legitimise expansions in domestic surveillance, projects of mass displacement, and an endless expansion in mineral extractivism — whether locally or from other parts of the global South, particularly the DR Congo.

Given that there is no burgeoning communist revolution in India or Brazil, the need of the hour is instead to engage with this discourse on two levels. The first is to puncture the techno-solutionist ideology suffusing outwards from Silicon Valley, the common-sense notion of development as the uncritical adoption of technology.22 More broadly, however, we need to engage with the notion of sovereignty itself, and to recognise that — insofar as it remains trapped within the bounds of the nation-state — it is an idea that will remain actively contested by a number of actors, and that can be exploited by domestic capitalists for their own ends. One of the biggest failings of the policy prescriptions that emerged from early dependency theory was the idea that the national bourgeoisie could drive development in newly-independent countries.23 Today, we find ourselves in a world where a similar narrative provides cover for neoliberalism’s even more haphazard regime of governance: de-risking asset manager portfolios, in lieu of sober industrial policy. The consequences will unquestionably be tragic.

Notes

  1. 1. Isabella Kaminski, “Datacentres are growing target of global climate-related legal cases, report find”, The Guardian, 25 June 2026.

  2. 2. Julia Pohle and Thorsten Thiel, “Digital Sovereignty”, Internet Policy Review, 2022.

  3. 3. Ingrid Harvold Kvangraven, “Dependency theory: strengths, weaknesses, and its relevance today”. In Erik Reinert and Ingrid Harvold Kvangraven, eds. A Modern Guide to Uneven Economic Development, 2023.

  4. 4. Aaron Benanav, A global history of unemployment: surplus populations in the world economy, 1949-2010, 2014.

  5. 5. This was also a testament to how the dependency theory umbrella could also include liberal capitalism: Fernando Henrique Cardoso, the architect of the Plano Real, published several works on dependency theory over the course of his academic career. The post-FHC consensus in Brazilian politics would also end up eventually pushing the former firebrand trade unionist Lula closer to the centre.

  6. 6. See: Bernardo Jurema, “Green Brazil?”, Sidecar, 19 September 2024; André Singer, “Lulismo 3.0: A Mid-Term Diagnosis”, New Left Review, Nov/Dec 2024; Vincent Bevins, If We Burn, 2023.

  7. 7. Achin Vanaik, “India’s Two Hegemonies”, New Left Review, Jul/Aug 2018.

  8. 8. Jayati Ghosh, “The Global Financial Crisis and the Developing World”, Indian Journal of Politics and International Relations, 2009.

  9. 9. Given geopolitical realities, these are at least more noble in Brazil than in India. Consider, for instance, that Twitter/X was blocked in Brazil due to unmoderated election misinformation, while the Indian government routinely interferes with social media platforms to stifle critique of the government. See: Amarabati Bhattacharyya, “Government Orders Trigger Sweeping Instagram Censorship Days After CJP Protest”, The Wire, 13 August 2026.

  10. 10. Vinit Ravishankar and Markus Kienscherf, “Train and deploy: Expropriation and accumulation in the generative AI economy”, Capital & Class, 2026.

  11. 11. Rafael Grohmann and Alexandre Costa Barbosa, “Sovereignty-as-a-service: How big tech companies co-opt and redefine digital sovereignty”, Media, Culture & Society, 2025; Anupam Guha, “The Red Herring Has Fangs”, Disjunctions, March 2026. For a recent work explicitly tying the notion of digital public infrastructures to the Gates Foundation, see: Cecilia Rikap, “‘Digital Public Infrastructure’: Cultural hegemony and states’ technological dependence”, Big Data & Society, 2026.

  12. 12. The use of the term “low-emission” was itself the result of a last-minute amendment by the centre-right senator Laercio Oliveira, who wanted natural gas to be included as a potential energy source.

  13. 13. Juan Iñigo Carrera, El capital: razón histórica, sujeto revolucionario y conciencia, 2013. See also: Luis Cortés, “In Technology’s Shadow”, Disjunctions, July 2026.

  14. 14. Scott Aquanno and Stephen Maher, “The Fall and Rise of American Finance: From JP Morgan to Blackrock”, 2024.

  15. 15. Benjamin Braun, “Asset manager capitalism as a corporate governance regime”, The American Political Economy: Politics, Markets, and Power, 2021.

  16. 16. While Big Tech’s cash flows also do fuel this expansion, there is a division in where these investments go; Big Tech firms prefer to invest in chips, while leasing infrastructures and energy supply from financial firms. See: Peter Rudegeair and Peter Santilli, “Why Big Tech’s AI Spending Is $3 Trillion Higher Than It Seems”, The Wall Street Journal, 16 August 2026.

  17. 17. Daniela Gabor, “The Wall Street Consensus”, Development and Change, 2021. This is also increasingly a tendency in the global North, particularly Europe; see: Daniela Gabor, “The (European) De-Risking State”, Stato e mercato, 2023.

  18. 18. Phil Neel, Hellworld: The Human Species and the Planetary Factory, 2025.

  19. 19. Giovanni Arrighi, “The African Crisis”, New Left Review, May/June 2002.

  20. 20. Patrick Heller, “The age of reaction: Retrenchment populism in India and Brazil”, International Sociology, 2020; Tyler Antonio Lynch, “Lula and the Ranchers”, Jacobin, 8 July 2024; Vanessa Chishti, “Narendra Modi Has Ruled in the Interests of Big Capital”, Jacobin, 22 May 2025.

  21. 21. Richard Barbrook and Andy Cameron, “The Californian Ideology”, Mute, 1995.

  22. 22. Oliver Nachtwey and Timo Seidl, “The Solutionist Ethic and the Spirit of Digital Capitalism”, Theory, Culture & Society, 2023.

  23. 23. This is not to exclude the more radical currents under the dependency theory umbrella: the 70s saw important theoretical contributions by Samir Amin and Ruy Mauro Marini, for instance, who were both less enthusiastic about the role of the national bourgeoisie.