Artificial intelligence is no longer just a competition between technology companies. It is becoming a contest for dominance between states over computing power, semiconductors, capital, talent, infrastructure and technological influence. The United States and China have gained prominence as the two dominant centers of this race, while India is endeavouring to build an AI ecosystem of its own. India is certainly not absent from the AI race. It has a large pool of highly capable engineers, a growing startup ecosystem, a huge digital consumer market and significant experience in deploying technology at a humongous scale. Yet the gap with the United States and China remains substantial. The question, therefore, is not whether India has an AI strategy. It surely does have a strategy. But the more formidable question is whether that strategy is adequate to compete at the frontier.
Three models of AI development
The United States has built its AI advantage around scale. Its ecosystem directly combines frontier laboratories, enormous private capital, hyperscale cloud infrastructure, advanced semiconductor technology, data centers and a deep research base. Companies such as OpenAI, Google DeepMind, Anthropic and Meta are developing the foundational models on which the next generation of AI applications will depend.
The leverage is therefore not simply that America has better AI companies. It has an ‘in-tandem’ fortifying system:
Capable and qualified tech workforce – Advanced chips with massive industrial backing → massive computing capacity → frontier models → commercial success → more capital → larger data centers and research → further AI advantage
This feedback loop is one of the most substantial reasons why the US has always remained ahead of others.
China has followed a different but equally daunting path. It has combined state direction, a large engineering workforce, domestic technology companies, extensive deployment and an increasing emphasis on cost-efficient models. Chinese companies have also been pushed to further strengthen domestic counterparts to advanced foreign chips because of US technology restrictions.
China’s rise is very crucial because it shows that the AI race is not merely determined solely by who spends the most money. Chinese models have increasingly kept pace with efficiency and cost, while China has translated AI research into large-scale industrial and economic applications.
India, though having entered the race slightly later, meanwhile, has chosen a somewhat different route.
Its accentuation has been on democratising AI, making computing resources, models and technological capabilities available beyond a small group of giant corporations. The IndiaAI Mission, approved in March 2024 with an outlay of ₹10,371.92 crore, aims to expand access to compute and data, develop indigenous AI capabilities, attract talent, support startups and build responsible AI. This is a considerable strength for India. But at the same time, it also uncovers the major difference in the stage at which India is currently competing.
The US and China are gradually asking:
How do we build and operate AI infrastructure at an enormous scale?
But India continues to raise the question:
How do we make sufficient infrastructure available to our researchers, startups and institutions?
The gap reflects differences in scale and technological maturity, rather than ambition.
The GPU problem: ‘talent cannot replace compute’
This distinction becomes easier to understand when looking at GPUs. A Graphics Processing Unit was initially developed for graphics and gaming, but its ability to perform mind boggling numbers of complex calculations side by side, makes it a particularly useful tool for AI. A simple way to comprehend it is this: a CPU is like a small number of highly capable workers handling different tasks, while a GPU is like thousands of workers performing calculations concurrently with one another. Frontier AI therefore requires far more than researchers sitting at computers. It requires data, algorithms, specialised researchers, GPUs, data centers, lots of electricity and enormous amounts of capital.
India has achieved significant progress. More than 38,000 GPUs had been onboarded under the IndiaAI Mission, while another 20,000 GPUs were announced in February 2026. But the numbers need to be taken into careful consideration. India’s 38,000 GPUs are part of a national shared-compute programme involving multiple researchers, startups and institutions. They should not be placed in direct comparison with the enormous dedicated clusters being compiled by individual American technology companies for frontier-model development.
So the real question is:
Can India become a frontier AI power without frontier-scale compute?
The problem goes beyond the mere concern of buying GPUs only. India continues to depend heavily on foreign suppliers for much of its advanced AI infrastructure. Carnegie notes that India’s IndiaAI compute infrastructure has heavily relied on GPUs supplied by American companies, while cloud infrastructure and advanced semiconductor technology also remain largely dependent on foreign firms.
This does not mean that India instantly needs complete technological autarky. But it does mean that strategic autonomy becomes arduous if the critical infrastructure required to train and run India’s most important AI systems remains predominantly external.
The semiconductor and infrastructure gap
This is where India’s historical trajectory becomes pivotal. India has, over time, developed enormous strength in software and services, particularly after the economic reforms of the 1990s. Its companies built global credibility in information technology, outsourcing and software engineering. But software strength did not by default produce semiconductor fabs, advanced processors, hyperscale cloud companies or frontier AI laboratories.
Semiconductor manufacturing is distinct in nature from software development. It requires enormous capital, specialised equipment, sophisticated supply chains, reliable electricity and water, advanced research and years of gradually obtained industrial expertise. India has increasingly acknowledged this gap. Under the Semicon India programme, the government has approved major semiconductor projects, while Semicon 2.0 has committed another large-scale investment to expand the country’s semiconductor ecosystem. These collective efforts show that India is no longer simply overlooking the importance of the physical foundations of technological power. But there is a fundamental timing problem. India is building the missing physical foundations of AI power while the US and China are already implementing those foundations at a massive scale.
The historical infrastructure gap was discernible even before today’s AI boom. A 2021 analysis by Georgetown’s Center for Security and Emerging Technology noted that India had only three of the world’s top 500 supercomputers, compared with 215 in China and 113 in the United States. Those figures are now dated and should not be treated as a picture of 2026, but they clearly depict how long India has been grappling with a computing-infrastructure deprivation.
India has the talent, but can it retain and harness it?
Talent is one of India’s greatest advantages, but unfortunately it is also a lingering problem. India yields a large number of engineers and technically trained graduates. Yet frontier AI requires more than a large engineering workforce. It strongly demands highly specialised researchers, advanced laboratories, sustained funding and access to enormous computing resources.
This is where the concept of ‘brain drain’ becomes relevant.
Highly educated researchers tend to move towards those environments where they have better access to research infrastructure, funding, compensation and opportunities to work on enterprising technological problems. The issue is therefore not that Indian researchers lack ability or commitment. It is that India has historically not consistently provided an ecosystem able to provide comparable opportunities available in the world’s influential technology centers.
This generates a feedback loop:
Better research infrastructure + stronger funding + competitive opportunities → attracts top researchers → stronger research → stronger companies → more capital → better research infrastructure
Breaking this cycle is imperative. India therefore has to deal with not only the problem of producing talent, but also the problem of establishing institutions in which its best talent can stay, research and build at the frontier. This disparity is notable because engineering talent and frontier research capability are not the same thing.
Capital, energy and the missing layers
India’s private sector is slowly beginning to respond. The country’s AI sector attracted significant new capital in 2026, with major investments going into companies such as Neysa and Sarvam AI. A substantial share of recent AI investment has also gone towards infrastructure such as GPU clouds and computing capacity. So it would be absolutely misleading to state that India has no AI capital. The problem is the scale and depth of that capital compared with the American ecosystem.
And there is another layer that is often ignored: energy.
AI requires data centres, and data centres require enormous amounts of reliable electricity. As models become larger and AI adoption expands, access to power, land, cooling, transmission infrastructure and suitable data-centre locations becomes part of AI rivalry itself.
In other words:
More AI → more compute → more data centres → more electricity → greater infrastructure requirements
The AI race is therefore evolving into an infrastructure and energy race as well. This is why exercising control only the application layer will not be enough. A country can produce sophisticated AI applications while retaining dependence on foreign chips, cloud providers and foundation models. That may make it a major AI market without inevitably making it an AI power.
Does India actually need to beat the US and China?
This may be the most important question, in the contemporary era of Artificial Intelligence. India does not inherently need to reproduce the American or Chinese model. Nor can it realistically expect to dominate every layer of the AI stack jointly. Its opportunity may instead lie in becoming indispensable in specific targeted areas: multilingual AI, low-cost AI, large-scale deployment, digital public infrastructure, emerging-market applications and AI systems designed for diverse populations.
India already benefits from certain advantages that neither the US nor China can simply reproduce: its enormous domestic market, linguistic diversity, digital public infrastructure and experience of deploying technology at population scale.
But there is a line India cannot risk going beyond.
India does not have to win the entire frontier-model race, but it cannot afford to be absent from it either. If the country completely abandons foundation models and frontier research, it jeopardises becoming permanently dependent on foreign companies for the most geopolitically significant layer of AI. From the opposite perspective, trying to replicate every part of the American ecosystem would require colossal resources and may not be the most efficient use of India’s strengths.
The more viable goal is strategic autonomy through managed interdependence.
India can work with American, European and other technology ecosystems while simultaneously building domestic capabilities in the layers that are strategically critical.
India is not starting from scratch, but is that enough?
India’s position should also not be underappreciated. Stanford’s Global AI Vibrancy Tool placed India Fourth globally behind the United States, China and United Kingdom mirroring its strengths in talent, research, investment, infrastructure, policy and adoption. But the gap in overall ecosystem strength remains significant. That highlights a crucial reality.
India is not starting from scratch. It has talent. It has a market. It has digital infrastructure. It has startups. It has government backing. It has begun investing in compute and semiconductors. But having an AI ecosystem is not the same as dominating the most strategically important layers of the AI stack. The US has built enormous strength around chips, capital, cloud, frontier models and infrastructure. China has built strength around state coordination, domestic technology, cost-efficient models, chips and mass deployment. India is advancing around access, affordability, applications, digital public infrastructure and strategic autonomy.
The question is whether India can now broaden that model.
India is not losing the AI race because it lacks talent or ambition. It is lagging behind because the AI race has moved beyond software into chips, compute, capital, energy, infrastructure and frontier research, and India is still building depth across these layers.
The goal, therefore, should not simply be to ask when India will become “equal” to the US or China. The frontier itself will continue to move. The critical question is whether India can safeguard that, as AI becomes an increasingly important source of economic and geopolitical power, it is not merely a consumer of AI, but a country with enough capability to shape the technology, influence its rules and remain free from excessive external dependence in a world increasingly dependent on it.
Article by Sruti Bhaumik