Author: Abhishek Singh
Every industrial era has been an argument about who owns the machines.
In the nineteenth century, it was the power loom, the locomotive and the telegraph cable. Nations that controlled them set the terms of trade, laid the routes of commerce and decided who could speak to whom. Others became markets, suppliers of raw material, or colonies. India knows this story well.
Today, the machines are GPUs, cloud platforms, foundation models and the data that feeds them. The phrase of the moment is “digital sovereignty,” and it is now at the centre of policy debates from Brussels to New Delhi.
Three of those old machines map almost exactly onto today’s digital stack.
Textiles: the extraction loop. India exported raw cotton and bought back finished cloth from Lancashire mills. The value-adding step, and the profits, stayed in Britain. Today, raw data flows out from a billion Indian users and returns as AI models, ads and cloud services sold back at a premium.
Railways: infrastructure built for someone else. Indian railways were laid out to move goods from the interior to the ports, not to connect Indian cities. British investors received guaranteed returns paid by Indian taxpayers, and locomotives and rails were imported. Infrastructure designed by someone else shapes your economy around their needs. Proprietary cloud stacks and app stores carry the same lock-in.
The telegraph: the off switch. Around 1900, Britain controlled most of the world’s undersea telegraph cables. When war broke out in 1914, it cut Germany’s cables within hours and could read much of the world’s traffic. Whoever owned the wires decided who could talk. Swap cables for cloud, and the logic carries straight into the present.
This is not hypothetical. In July 2025, the EU added Indian refiner Nayara to its Russia sanctions package. Microsoft then suspended Nayara’s Outlook, Teams and other tools, despite fully paid-up licences and no requirement under US or Indian law.
A third country’s rule, enforced by a foreign platform’s compliance team, switched off communications of a major Indian energy company. That is the telegraph’s off switch, rebuilt in software.
The geopolitics of technology has flipped. The loudest voices for digital sovereignty are European: France, Germany, and the EU. These are former imperial cores, now worried about American platforms. This is more of a contest over who controls the stack. India’s risk is getting caught in other powers’ contests, just as Nayara was caught in Europe’s.
India is no longer just a supplier of raw material, but it still buys the machines.
Over the past two years, the IndiaAI Mission has built a national pool of tens of thousands of GPUs, rented out cheaply to startups and researchers. Homegrown models from Sarvam and IIT Bombay’s BharatGen now compete with global models on Indian languages. A national data platform, AIKosh, is gathering the datasets those models learn from. Meanwhile, the first chip plants in Gujarat have begun production, though they package and assemble chips rather than make the most advanced ones.
Read through the lens of history, a pattern emerges. India now owns the mills, the models that refine data into intelligence. It also holds the raw cotton, its vast and multilingual data, and the distribution networks, public digital infrastructure like UPI and Aadhaar. But the steam engines and locomotives, the chips and GPUs, are still imported, almost entirely from Nvidia and built with foreign tools. As Nayara showed, even the cloud software in between can be switched off from abroad.
There is one more analogy. Dadabhai Naoroji once warned of a “drain of wealth” from India to Britain. Today’s drain is talent: Indian engineers lead much of America’s AI research, and cheap GPUs alone will not bring them home. Frontier problems and competitive pay will.
History also shows how countries escaped this trap. Those that broke free of industrial-era dependency did not refuse foreign technology. They imported it aggressively while building the capacity to make it themselves.
Meiji Japan hired foreign engineers, bought Western machines and sent students abroad, then built its own shipyards, railways and universities within a generation. Bismarck’s Germany and the nineteenth-century United States did the same with British technology. In India, the steel plant envisioned by Jamsetji Tata, which began production in 1912, did not make Britain irrelevant. It gave India a foothold in a strategic industry.
What does this mean for India? Which layers of the stack should we own, and where should we lean on partners?
We must own the layers where India has an edge: data, Indian-language models, public digital infrastructure and applications. India’s real “rail gauge” opportunity is exporting UPI- and Bhashini-style public AI infrastructure to the Global South, on terms no American or Chinese firm offers. We must hedge the layers India cannot own yet- frontier chips and frontier models. That means multiple suppliers, open-weight models, local data residency, and contracts and laws that limit foreign switch-off rights. We should also invest in the slow layers: fab skills, research universities and semiconductor talent. Japan’s catch-up was driven by institutions, not hardware purchases. In addition, we should keep domestic markets competitive, so “sovereign” never means “controlled by one Indian company.”
In the age of cotton and cables, sovereignty meant owning the machines. In the age of code, no country owns them all, not even the United States, whose chips depend on Dutch lithography and Taiwanese fabs.
A more realistic definition is this: sovereignty is the ability to say no, and to survive someone else saying no to you. It means not having your email, payments or public services switched off by a decision taken in another capital. It means keeping the most valuable step, refining data into intelligence, at home.
India is moving in that direction. It has gone from raw-material supplier to local processor. The danger now is not that India does too little, but that it chases the wrong goal: trying to replicate the entire American stack, or building a domestic monopoly and calling it independence.
The lesson from history is not that technology is destiny. It is that dependency is a choice that compounds. We must choose carefully which dependencies to keep, and work steadily to own the ones that can be used against us.
The author is an IAS officer; Views presented are personal.