The Political Economy and Political Ecology of AI in India (Part I)
Over the last few weeks, artificial intelligence (AI) has evolved from a technological spectacle to a political flashpoint across continents. At the recent high-profile AI summit in India, intended to showcase the country’s digital ambitions, it became a story of organisational chaos and corporate spectacle rather than the technology itself. A week before that, Indian opposition leader Rahul Gandhi questioned whether India was becoming merely a market and data supplier in a global AI economy rather than a decision-maker within it. At the same moment, in the viral snippet of the speech from the CEO of Palantir, Peter Thiel, remarked, suggesting that advanced data systems could shape electoral outcomes, intensifying fears about the political uses of analytics. In the British Parliament, its members engaged in a heated debate over the role of private technology firms in managing data from Britain’s widely applauded state-funded National Health Service, raising concerns about sovereignty, accountability, and public trust. These episodes may appear disconnected, but together they reveal a common anxiety. AI is no longer just a tool for productivity, but it is becoming an instrument of power.
The unease may again arise about the machines becoming conscious. However, it is more about reorganising authority. When political leaders worry about data depart national control, when public institutions hesitate to rely on proprietary algorithms, when citizens suspect that elections or welfare systems could be influenced by opaque infrastructures, they are responding to a deeper transformation. AI is quietly shifting its boundaries of innovation to that of governance now. It has started mediating key sectors such as health, education, policing, credit, and communication, which directly affect citizens’ everyday lives. Thus, the central question must no longer be what AI can do, but who it enables to decide over its production and use.
India’s discourse emerges within this transition with unusual clarity. Recent speeches in Parliament (India and elsewhere) situate AI as a site of dependency rather than merely an opportunity for growth. At present, the immediate question is whether the country risks exporting raw data while importing technologies such as AI.
The spectacle of innovation can coexist with a quiet transfer of control. Large language models are trained on the linguistic and cultural output of millions around the world, mostly in the global south, but the final products are owned elsewhere, leading to knowledge itself becoming a commodity extracted and resold. The celebration of technological progress often masks the real societal, collective inputs through labour, data, and natural resources, while the profits are captured by a few, and the real use of such technologies is limited to a smaller section of elites.
The historical anxieties about earlier, more crude forms of technology appear less quaint and more as lessons. Societies have always had an instilled fear of machines, not because they were inherently destructive, but also because they reorganised power. The printing press undermined religious monopolies while enabling new forms of ownership and censorship. The steam engines expanded production but intensified industrial discipline. Similarly, AI belongs to this lineage. It is not simply a smarter machine but a system that enables the redistribution of wealth, knowledge, labour, and authority, and this time on a planetary scale.
What makes AI stand out is that it is not merely a tool but an infrastructure in its own right. It depends on data centres, fibre-optic cables, specialised chips, a huge amount of energy, heavy water usage, cooling systems, and often an invisible workforce that trains and maintains it, mostly known as ghost workers. The seamless conversational AI interface hides an industrial apparatus of overwhelming scale. Governments announcing AI policies and projects means that they are effectively announcing massive infrastructure projects similar to railways or power plants, except that these infrastructures are frequently privately owned and globally integrated from the outset. Such infrastructures essentially determine countries’ dependence on other countries with more developed AI systems, which in turn determines which regions supply resources, who gains decision-making power, and who absorbs the costs.
What helps here in understanding this complexity and contradiction is India’s emerging data-centre boom. Prima facie, it appears to be the government’s integration into the digital future as a project, but in reality, it requires a vast amount of energy and water. By 2026, India’s data centre capacity is expected to more than double, going from about 870 MW in 2023 to over 1.8 GW. This will require power on par with that of a medium-sized city in India. Depending on the technology and the weather, estimates suggest that a single large hyperscale data centre can use between 10 and 50 million litres of water a year to cool itself. This growth is happening in a country where many families still struggle to access reliable water, and energy prices are rising. NITI Aayog has warned that almost 600 million Indians are under significant water stress. People who live in cities like Bengaluru, Hyderabad, and parts of Mumbai often rely on tanker water during the summer and pay several times the municipal rate for a basic need. The paradox is more striking than it appears: a technology celebrated for its immaterial intelligence intensifies competition over the most material resources, such as land, water, and power. When computing capacity expands, something else must contract, and it is rarely the consumption of the affluent.
The labour dimension of AI reveals a similar pattern of dispossession and displacement. Public discourse often frames AI as displacing human work, however, the system depends on a vast workforce that remains largely invisible. Millions of workers, mostly from the Global South, label images, annotate, filter violent content, transcribe speech, translate languages, and correct errors so that machine-learning models can function. This labour is often outsourced through layers of subcontracting to regions where wages are lower and protections weaker. These workers would never see the final product they helped create. They only encounter fragmented tasks delivered via digital platforms. Nonetheless, there are a few companies in India that take on contracts for data annotation and labelling for Silicon Valley tech giants, mostly through multiple contract chains in between. The crucial question then emerges: are such workers in India protected by the law? Because the effect is not the end of labour but its fragmentation into micro-tasks that are difficult to organise or recognise as employment. It indicates that factories are not gone but have started mediating through our living rooms, small offices and odd working spaces throughout the global south.
Recent unrest in India’s IT sector illustrates how AI reorganises work without necessarily replacing it wholesale. Layoffs are often framed as a technological inevitability, but coincide with corporate restructuring and cost reduction. Entry-level jobs are vanishing for sure, but the pressure to increase productivity is being shifted to those who remain and to the invisible/ghost workers in the supply chain. Workers are continually pushed to upskill, and insecurity is imposed on them as the price of their relevance. AI becomes less an option for labour than an apparatus for disciplining it and pushing them into more precarious positions, while simultaneously setting new benchmarks for speed and availability. The promise of efficiency thus translates into a redistribution of risk from corporations to individuals.
The political implications, however, extend beyond labour markets into public governance. Data has now become a strategic resource, and control over data infrastructures extends influence over state functions. The usage of private analytics companies within public institutions reflects anxiety about whether essential services should depend upon proprietary systems controlled by external actors. Even the British Parliament was shaken by the ramifications of such interference in their prominent state-funded National Health Service. The wealthy nations are also alarmed by the gradual erosion of autonomy when critical infrastructure becomes dependent on a handful of private companies. For countries with fewer resources, the stakes are even higher. These waters we trade in are quite dangerous, especially when you have no national-level firm policy to protect its citizens and their data.
Environmental costs add layers to this political economy of AI. In regions already confronting climate stress, this can intensify local vulnerabilities. The extraction of minerals required for this infrastructure links the political economy of AI to global supply chains originating especially in ecologically fragile zones. Thus, a technology marketed as futuristic relies on resource frontiers that resemble those of earlier industrial expansions. As a result, the profits and benefits of such technology get concentrated in wealthier regions, while extraction zones bear long-term environmental damage.
AI, its production and associated technologies are already concentrated in the hands of a few Silicon Valley companies. Even though the usage of the final products of AI technologies is being enjoyed by a smaller chunk of the population, specifically the urban middle classes, most of its benefits are being accrued by private companies to increase productivity and accumulate profits. While AI concentrates economic gains in a small number of corporations, the increased pressure on labour and natural resources is always pushed on regions with limited bargaining power. None of these outcomes is predetermined by technology itself, but they follow the already existing unequal arrangements within which technology takes shape.
The ongoing sense of awe about AI is partly due to how people in our society view technology as an independent force rather than a choice made by society. This lets us celebrate new ideas without having to discuss how to distribute their benefits. But history shows that technological changes always create winners and losers, and that the outcomes depend on political negotiation. The printing press, the steam engine, and the internet all made new things possible, but they also made old problems worse in new ways. AI is likely to follow a path that is just as contradictory.
A democratic response to such a challenge cannot be based on fear or uncritical enthusiasm. It would require recognising AI as a field of contestation. We must be able to address political questions arising from this conjuncture, such as who owns data, who regulates these models, who benefits from them, and who bears the costs. Public investment in this area becomes critical and demands prioritising open systems and public oversight. Labour laws must adapt to recognise new forms of work arising from these changes, and environmental regulations must account for resource footprints. This wouldn’t be anti-technology; those would be the attempts to align technological advancement with social priorities.
