Where Does India Stand in the Global AI Race?

When the United States published its AI Action Plan in July 2025, it framed artificial intelligence (AI) as a contest for global dominance. Whoever builds the largest AI ecosystem, the document argued, will set global AI standards and collect broad economic and military benefits. The line raises a question that India has not yet answered clearly: in a world where AI is becoming a contest of power, what position is it actually building towards? Power in the global AI economy is not evenly distributed. It maps onto a layered hierarchy. At the base sit advanced semiconductors and fabrication plants. Above that are cloud platforms and data centres that supply compute. At the top are foundation models and the applications built on them. Each layer confers a different kind of influence, but those who lead in models shape AI standards and control access. The United States occupies several of these layers at once -- it holds the dominant cloud platforms, the frontier model labs (OpenAI, Anthropic), and deep integration with the semiconductor design firms (Nvidia, AMD) that supply the rest of the world. China has responded by investing heavily in domestic semiconductor manufacturing capacity and building its own widely used models. Taiwan and South Korea have concentrated on advanced chipmaking. Each country has made a different strategic bet. India's comparative advantage has historically been in delivering IT and IT-enabled services at scale -- a skilled workforce at competitive cost, combined with a public digital infrastructure that enables rapid adoption. Anthropic's country brief on India recognises it as one of the top global markets for Claude, second only to the United States. The Stanford AI Index's Global Vibrancy Tool (2025), which assesses research, investment, talent, policy, and economic activity, placed India behind the US and China with particular strength in talent and adoption. India's IT Minister, Ashwini Vaishnaw, cited this data at Davos. But adoption and vibrancy are not the same as leverage in the AI supply chain. Are we building or renting? India's R&D spending stands at approximately 0.6 per cent of GDP, compared to 3 to 4 per cent in most innovation-driven economies, according to analysts. High-end talent retention has compounded the gap. Sriram Krishnan, the Senior White House Policy Adviser on AI, was born in Chennai and educated at SRM University before emigrating to the United States, where he became a US citizen in 2016. Karandeep Anand, CEO of Character.AI, was born in India and attended IIIT Hyderabad before completing an MBA at Northwestern University. Both were named among Time magazine's "Architects of AI", its 2025 Person of the Year collective. India, ironically, seems to consistently produce elite technical talent that ends up powering someone else's AI agenda. As Kak and Kapoor observe, many low- and middle-income countries believe that failing to participate in AI will deepen their marginalisation. The Indian government appears to have absorbed that anxiety. In her 2023-24 Budget speech, Union Finance Minister Nirmala Sitharaman announced three Centres of Excellence (CoEs) in AI. In 2024-25, these CoEs reportedly received Rs.255 crore (roughly $28 million). Also Read | The Indian economy's weird state of suspension The sum is not trivial, but it does not build the kind of research infrastructure capable of shifting India's position in the frontier model hierarchy. Training compute costs for GPT-4 were estimated at $78 million and for Gemini Ultra at approximately $191 million. Even accounting for differences in purchasing power parity, the gap is significant. A more revealing lens is the split between direct and indirect AI spending in the Union Budget. Direct AI spending covers schemes explicitly earmarked for AI such as the IndiaAI Mission under the Ministry of Electronics and IT (MeitY). Indirect AI spending covers AI-adjacent or AI-enabling schemes: compute capacity, semiconductors, cybersecurity, and the underlying infrastructure that makes AI development possible at scale. Among these numbers, the IndiaAI Mission allocation in 2026-27 stands out. Despite being the flagship AI programme, its allocation was halved to Rs.1,000 crore from the previous year. In 2024-25, 96 per cent of the budgeted amount remained unspent, according to the Actuals released in the latest budget. This is the same fund being used to co-sponsor the AI summit in Delhi. Parliamentary responses describe the IndiaAI Mission as having a total outlay of Rs.10,371.92 crore over five years, with the largest single pillar being IndiaAI Compute Capacity at Rs.4,563.36 crore. Spending on compute outpaces the mission's allocations for foundation models, datasets, and skilling combined. The pattern is consistent: compute and AI-enabling infrastructure take priority over building models. Taken together, these budgetary choices suggest that India is orienting itself towards becoming a destination for running AI workloads, not for producing the systems that run on them. The policy move On February 1, Reuters reported that India would offer a tax holiday -- zero taxes until 2047 -- to foreign firms using Indian data centres to provide cloud services to global clients. A TechCrunch report described the move as a bid to attract the next wave of AI computing investment, noting that power shortages and water stress remain real constraints in a country where reliable electricity and clean water are still unavailable to large portions of the population. Investment announcements around and after the AI Impact Summit reinforce the same compute-and-hosting logic. Reliance Industries outlined $109.8 billion for AI and data infrastructure and is building data centres in Jamnagar. The Adani Group announced a $100 billion commitment for AI data centres by 2035 and is constructing campuses in Visakhapatnam and Noida. Yotta Data Services committed $2 billion to an AI computing hub using Nvidia chips. Google, Microsoft, and Amazon together committed a combined $68 billion in AI and cloud infrastructure investment in India by 2030, according to Reuters. OpenAI has also partnered with the Tata group to secure 100 megawatts of AI-ready data centre capacity, with an ambition to reach 1 gigawatt. What these announcements collectively signal is that India is presenting itself -- and being treated -- as a destination with hosting capacity, jurisdictional stability, and market access. The timing matters. India commands roughly 55 per cent of the global IT outsourcing market. Generative AI has introduced a structural risk to that position. A 2025 Gartner estimate projected that nearly 80 per cent of customer queries will be resolved by AI agents by 2029. Markets have begun to price this risk: in February 2026, Indian IT stocks fell sharply amid investor concern that AI-driven automation would erode the outsourcing model. Against that backdrop, becoming a data centre and compute hub appears to be a considered strategic response -- trading one services model for another. The Vast.ai booth at the AI Impact Summit in New Delhi, India, on February 20, 2026. India's comparative advantage has historically been in delivering IT and IT-enabled services at scale. | Photo Credit: Ruhani Kaur/Bloomberg On February 20, 2026, India formalised a further alignment. On the final day of the India AI Impact Summit, India signed the Pax Silica Declaration, becoming the eleventh signatory to the US-led initiative. Focused on securing supply chains for critical minerals, semiconductors, and AI infrastructure, Pax Silica spans the technology stack from rare earth extraction to frontier AI deployment. By joining, India has aligned itself with a geopolitically structured framework in which hosting capacity, supply chain reliability, and proximity to US technology partners are understood as strategic assets. What indispensability actually requires The US model in AI combines dominant private-sector "hyperscalers" -- large-scale cloud service providers -- with sustained state-backed research and defence funding. DARPA (Defense Advanced Research Projects Agency) has backed AI-related programmes for years, including work on explainable AI, pushing frontier capability alongside strategic applications. China's approach has been more centralised: its 2017 State Council plan set objectives through 2030 and emphasised building an integrated industrial chain, national standards, and domestic capability at every level of the stack. A smaller country is instructive here. Taiwan has made itself indispensable not by attempting to compete across the AI hierarchy but by controlling a chokepoint within it -- advanced semiconductor manufacturing. A dominant share of the world's most advanced chipmaking capacity is concentrated in Taiwan and South Korea, which is why Taiwan's stability carries global economic significance far beyond its size. Also Read | AI is not India's problem. Governance is India cannot replicate Taiwan's path -- the capital requirements and decades of accumulated process knowledge involved in advanced chipmaking are beyond what India can reasonably build in a short window. But it can take seriously the underlying logic: indispensability requires owning something that others cannot easily substitute. At present, India's compute-and-hosting strategy does not meet that test. Data centres can be built in many jurisdictions. Tax holidays are replicable. Market access is real, but not unique. All the indicators point in the same direction. The AI summit's framing emphasised deployment and investment. The mission architecture prioritises compute capacity. The 2047 tax holiday is designed to attract foreign firms to route global cloud services through India. The infrastructure commitments run into hundreds of billions of dollars. MeitY Secretary S. Krishnan, speaking on the first day of the summit, encouraged private investment in data centres and AI-driven compute infrastructure. Taken together, this is a coherent positioning: India as a hub at the compute-and-services layer of the global AI economy. The strategy plays to India's genuine strengths. The open question is whether it is designed as a base from which to move up the hierarchy, or as the destination itself. If the former, the policy agenda must extend well beyond data centres and subsidised compute. It requires sustained investment in research institutions, conditions that retain AI researchers within the country, and long-horizon R&D funding that builds foundational capacity rather than maintaining service competitiveness. None of those conditions are currently visible in the Budget. India rode the Y2K wave into IT outsourcing dominance but consistently captured the labour-arbitrage tier of the stack rather than building the IP layer above it. No Indian firm owns a global operating system, a hyperscaler cloud, or a dominant enterprise software suite. The sector grew large by servicing others' products, not by producing its own. The data centre and compute-hosting bet follows the same logic: enter at the execution layer, scale on cost and capacity, and defer the harder question of whether to move up. Summing up, this appears to be India's outsourcing moment for AI. Sayamsiddha is PhD Student at The New School for Social Research, New York. Featured Comment