By Amit Kapoor and Mohammad Saad
In the 21st century, sovereignty is not merely political. It is digital, and achieving AI sovereignty has become a national imperative. While many countries aspire to AI sovereignty, global AI power remains concentrated in the US and China. According to Stanford’s AI Index Report 2025, in 2024 the USA led with 40 notable AI models, followed by China with 15. With these regions controlling nearly all LLMs that have achieved a global footprint, India faces the formidable challenge of developing indigenous LLMs capable of competing with these global giants, a capability it has yet to build. While demands for AI sovereignty are high, recent discussions have often overlooked that the challenge is not purely technological. It also reflects underlying policy gaps that remain unaddressed. As a result, dependence on foreign AI capabilities may remain inevitable, at least in the short run.
AI sovereignty refers to a nation’s ability to understand, develop, and regulate AI systems in a manner that allows it to exercise control over them. Achieving sovereignty in AI is crucial, as it enhances security in defence and governance. It also positions a nation as a global innovator and attracts investment, thus allowing the development of customized solutions. In this regard, India has attempted to achieve AI sovereignty through initiatives such as the IndiaAI Mission, the Semiconductor Mission, Digital Public Infrastructure, skilling initiatives, data localisation mandates, and the recent notification of the data protection rules. While these initiatives may take time to materialize, recent developments suggest that we could be moving toward greater AI dependency, which could run counter to the objective of aatm-nirbharta.
To illustrate this point, a study released by the Competition Commission of India in October characterised India’s AI ecosystem as a structure comprising four layers: compute, foundation models, tools, and applications. The report found that 67% of the surveyed firms operate only at the application layer, indicating that India is achieving sovereignty, of some sort, only in domain specific products and services. This means that the Indian AI market is rapidly moving towards the development of AI products that rely on compute infrastructure and foundational models controlled by foreign entities. In fact, brokerage firm Bernstein, in the context of foreign LLMs penetrating the Indian market at scale, described the situation as a “wake-up call” for India.
In recent years, India has taken several steps to strengthen its AI ecosystem, from expanding national compute capacity to 34,000 GPUs in 2025 to introducing production-linked incentive schemes for semiconductor manufacturing and broadening internet access to over 1 billion users. Despite these initiatives, several challenges continue to persist. These include a shortage of AI talent, limited hardware and chip manufacturing and a weak R&D ecosystem. Bain & Co (2025) warned that India could face a shortfall of over a million skilled AI professionals by 2027. In hardware, the Indian semiconductor market stood at $45–50 billion in 2024–25, compared with China’s $180–200 billion. R&D spending is also low, at 0.6% of GDP in 2024, versus 2.68% in China and roughly 3.5% in the USA. High AI development costs and growing energy requirements for data centres add further pressure.
Crucially, these obstacles to AI sovereignty are not new. Despite recognition and policy support, they have persisted over the years. What may be novel in analysing India’s AI sovereignty challenges is that these obstacles are rooted not in technology, but in deeper socio-economic and structural realities that the nation may be unprepared to confront. The issue of skilled workers migrating out of the nation illustrates this point. From a structural perspective, one may notice that this issue is grounded in the reality that the available number of jobs is disproportionately lower than the size of the workforce, and that we often encounter young researchers in the country complaining about low PhD stipends and low research grants. While the government’s recent establishment of the Anusandhan National Research Foundation (ANRF) for fostering and promoting R&D is a welcome step, the limited level of private R&D investment suggests that this alone may not be enough.
A similar weakness is evident in manufacturing which remains both limited in scale and relatively low in technological intensity. This makes developing high‑tech sectors like semiconductors particularly difficult. The country effectively leapfrogged from agriculture to services, with manufacturing contributing only about 17% of GDP, compared with roughly 25% in China in 2024. Although technology‑intensive exports have grown in recent years, overall industrial production still relies on less advanced processes, and the small size of the sector reduces incentives to innovate and invest in R&D.
The shortcomings of Indian academia are another structural issue that can be seen through the triple helix model of innovation, which stresses collaboration between academia, industry, and government as essential for technological development and commercialization. Often disconnected from industry realities, Indian academia has been criticised for failing to meet practical needs, reducing incentives for industry to invest in research. Finally, the concern that India is gradually becoming a data colony of the West follows a similar pattern. It is indeed a result of how India has dealt with problems of data privacy. It is difficult to deny that India has always been slow in developing data protection laws, unlike its counterparts in the European Union.
While the analysis of structural problems in India is far from complete, it allows one to see that the lack of AI sovereignty is not simply a result of insufficient initiative, but rather a symptom of broader macroeconomic issues spilling over into the ecosystem. This means AI sovereignty is achievable only in the long run, as structural problems take time and require a sustained strategy. Without such long-term commitment, the dream of an AI-sovereign nation will be pushed further into the future. That said, short-run decisions remain important, and India has, to some extent, recognized that a sudden attempt at sovereignty would leave the country isolated. Isolation would not solve gaps in semiconductors, foundational models, or data centres. While these deficiencies enable AI colonialism, cautious action may be the only practical option at present. Sovereignty remains a long-term goal and achieving it is a steep hill to climb.
The article was published with Economic Times on February 5, 2026.
