The ‘AI IP deficit’: Will India become the world’s largest user of intelligence it does not own? 

India may become one of the world’s largest consumers, deployers and implementers of artificial intelligence (AI). But will it own enough of the intelligence it uses? That is the uncomfortable question hidden inside a remarkably candid observation by India’s Chief Economic Adviser, V. Anantha Nageswaran. 

Speaking on September 25, 2026, Nageswaran identified three immediate external-economic challenges confronting India: (1) an unsettled relationship with the United States, (2) elevated global energy risks, and (3) India’s comparatively weak position in the AI ecosystem. But he made an important qualification. India is not absent from AI. It participates in applications and areas such as edge AI. 

The problem, he said, is that “the IP is not quite there.” Those few words may contain one of the most important economic-policy questions India will confront over the next decade. Because there is an enormous difference between using intelligence and owning intelligence. 

The next outsourcing paradox 

India has seen this movie before. The country’s information-technology revolution was an extraordinary economic achievement. Indian engineers built systems, maintained infrastructure, transformed business processes and helped digitize corporations across the world. 

Yet much of the highest-value software IP, operating systems, cloud platforms, databases and enterprise products remained owned elsewhere. India became extraordinarily good at making other people’s technology work. AI creates the possibility of repeating this architecture, except at a much larger economic scale. 

Consider India’s strengths. The country possesses a vast technology workforce, a powerful IT-services industry, rapidly expanding Global Capability Centers (GCCs), extraordinary digital public infrastructure and an enormous domestic market capable of generating AI applications at population scale. 

India’s GCC ecosystem alone generated an estimated $64.6 billion in FY2024 and employed more than 1.9 million professionals. NASSCOM expects the sector to approach $100 billion by 2030. More importantly, GCCs are increasingly undertaking product engineering, AI and sophisticated R&D rather than merely back-office work. India does not suffer from an AI participation deficit. It suffers from an AI ownership deficit. And those are entirely different problems. 

Follow the Intellectual Property 

The global AI economy can be understood as a sovereignty stack: Talent → Data → Compute → Chips → Models → IP → Applications → Distribution.  

India is unusually strong toward both ends of this chain. It has talent, data-generating scale, enormous opportunities for applications, distribution through its population, enterprises and digital public infrastructure. But several of the economically decisive layers in between remain comparatively weak. The numbers are sobering. 

Stanford University’s 2025 AI Index reported that China accounted for approximately 69.7% of AI patents granted worldwide in 2023, while the United States accounted for about 14.2%. India’s share? Approximately 0.37%. That statistic should command considerably more attention than another debate about whether India immediately needs to produce a model larger than OpenAI’s, Google’s or Anthropic’s latest system. 

Patents are not a perfect measure of technological power. Nor does every strategically important AI innovation become patented. Open-source models further complicate any simplistic equation between patents and capability. 

But the underlying asymmetry is difficult to ignore. India possesses millions of technically capable people while capturing a comparatively small share of formal AI intellectual property. That suggests an extraordinary paradox: We may have the minds without owning enough of what those minds create. 

From software rent to intelligence rent 

This matters because intellectual property determines where economic rents accumulate. Imagine an Indian bank deploying foreign foundation models for customer service, credit analysis, fraud detection and internal productivity. An Indian hospital uses foreign AI architectures for diagnostics. Indian manufacturers integrate foreign industrial AI systems. Indian universities use foreign AI research platforms. Indian startups build thousands of applications using foreign foundation-model APIs. 

India could experience enormous productivity gains. Indian businesses could flourish. 

Consumers could benefit. GDP could rise. And yet an invisible meter may continue running underneath the entire economy. Every API call, cloud workload, proprietary model license, specialized accelerator, enterprise AI subscription and foreign-owned technological dependency can potentially transfer some portion of the value created upstream. This is what might be called intelligence rent. 

The strategic question is no longer merely: How much AI will India use? It is: How much of the economic value generated by India’s AI usage will India ultimately capture? A nation can be digitally sophisticated and technologically dependent at the same time. 

India’s real bottleneck may begin before AI 

The deeper problem is India’s research intensity. The Economic Survey 2025–26 puts India’s gross expenditure on R&D at approximately 0.64% of GDP, compared with 3.48% for the United States, 2.43% for China and 4.91% for South Korea. India’s business sector contributes only about 41% of total R&D expenditure, compared with roughly 75% in the US, 77% in China and 79% in South Korea. That is not merely an R&D statistic. It is an IP-production statistic waiting to happen. You cannot indefinitely underinvest in research and simultaneously expect to dominate the ownership of technologies produced by research. 

India’s AI challenge cannot be solved merely by training more engineers. We already produce talent. The harder transformation is: Talent → Research → Invention → IP → Product → Company → Global Scale. Too much value can leak from this pipeline before Indian ingenuity becomes Indian-owned intellectual property. 

The good news: The architecture is beginning to change 

India is not standing still. The ₹10,371.92-crore IndiaAI Mission has expanded far beyond its initial goal of 10,000 GPUs. By 2026, more than 38,000 GPUs had been onboarded into the common compute infrastructure, with access being provided to startups, researchers and academia. Twelve teams were also shortlisted to develop indigenous foundation models. 

Models developed by Sarvam AI, BharatGen, Gnani and Socket have already been launched under this ecosystem, while another 20,000 GPUs are in the process of being added. 

The semiconductor strategy is simultaneously moving upstream. Semicon 2.0, approved in July 2026 with an outlay of ₹1,27,500 crore, explicitly includes chip design, R&D, equipment, materials, fabrication and talent development. By September, the government reported that 105 startups had begun developing chips and five semiconductor manufacturing plants had commenced commercial production. 

These developments matter because AI sovereignty cannot exist independently of semiconductor sovereignty, compute availability and research capability. But sovereignty must not be confused with isolation. 

Sovereignty is not autarky 

MeitY Secretary S. Krishnan offered a useful formulation this week: India should pursue “sovereign autonomy” while remaining integrated with global technology ecosystems.  

Autarky = We must produce everything ourselves. 

Dependence = Someone else can switch us off. 

Sovereign autonomy = We collaborate globally, but retain enough capability that nobody can easily coerce us technologically. 

India does not need an Indian equivalent of every American GPU, Chinese model, European industrial platform and Japanese semiconductor material. That would be economically irrational. Instead, India must identify the strategic layers where excessive dependence could become vulnerability and ensure credible domestic capabilities or diversified alternatives. The objective should be selective technological sovereignty, not technological nationalism. 

From “Make AI work for India” to “Owning what India creates” 

India’s AI strategy now requires a second phase. Affordable compute is necessary. Indian-language models are necessary. AI applications for agriculture, healthcare, education, governance and MSMEs are necessary. 

But none of these alone guarantee ownership. India needs substantially stronger incentives for universities, startups, GCCs and corporations to create commercially valuable AI IP in India; deeper industry-funded research; mechanisms allowing researchers to translate discoveries into companies; patient capital for deep-tech ventures; world-class AI research institutions; stronger university-industry laboratories; strategic public procurement of Indian technologies; and incentives that reward the creation of globally scalable products rather than merely domestic deployment. 

Most importantly, India’s enormous GCC ecosystem must gradually evolve from “innovation in India” to “IP from India.” That one preposition could eventually be worth hundreds of billions of dollars. 

Final thoughts: Who will own the intelligence? 

The nineteenth century rewarded countries that controlled industrial machinery. The twentieth century rewarded those that controlled capital, energy, brands, patents and software. The twenty-first century may increasingly reward countries that control machine intelligence. 

India possesses something immensely valuable: hundreds of millions of potential users, enormous datasets, extraordinary engineering talent, world-class digital infrastructure and a rapidly expanding technology economy. But possessing the world’s largest workshop is different from owning the machines inside it. And possessing millions of AI users is different from owning the intelligence serving them. 

India should certainly become the world’s largest laboratory for applying AI. But it must simultaneously become a serious laboratory for inventing, patenting, commercializing and owning it. Otherwise, we risk constructing the world’s most magnificent AI-powered economy on intellectual property rented from elsewhere. 

The danger is not that India will miss the AI revolution. India is far too large, talented and digitally sophisticated for that. The more unsettling possibility is that India could participate magnificently in the AI revolution and still capture disproportionately little of its deepest economic rent.  

The question is not whether Indians will use AI. They unquestionably will. The question that will determine India’s technological sovereignty is far more consequential: When a billion Indians eventually use machine intelligence every day, who will own the intelligence they are using?