With a massive young populace behind it, India’s AI market is projected to reach $131 billion by 2032, growing at 42.2% annually — as the country looks to stake its claim in the three-way race for AI dominance alongside the U.S. and China.

According to the Press Information Bureau (PIB) of the Indian government, the country witnessed its AI market growth from $2.97 billion to $7.63 billion between 2020-2024.

“The Indian AI market is projected to reach US$131.31 billion by 2032, growing at a compound annual growth rate (CAGR) of 42.2%, as per a report by Competition Commission of India,” read a statement from PIB.

India’s indigenous push in AI has intensified after the recently held New Delhi AI Impact Summit 2026, that saw investments from conglomerates like Reliance Group ($110B), Adani Enterprises ($100B), Google ($15B) and General Catalyst ($5B) to promote AI infrastructure in the country.

The New Delhi AI Impact Summit 2026 also saw India launching several Foundation Models including Sarvam-105B by Sarvam AI, Inya VoiceOS by Gnani.ai, Vaidya 2.0 by Fractal. Tech Mahindra in partnership with NVIDIA also introduced Project Indus 8B .Several India centric benchmarks such as IndicXTREME, IndicGLUE, IndicBias and MILU have also been produced by India’s research community. 

However, the road to place itself in the triumvirate race of AI dominance is tricky for India as it braces for job loss impact due to automation and its subsequent economic fallouts.

It is here that the larger question of who builds the AI that powers it comes into focus. That question sits at the heart of a new government white paper pushing for indigenous foundation models.

India Launches Its White Paper on AI for Indigenous Foundation Models

The White Paper titled , “Advancing Indigenous Foundation Models” released by The Ministry of Electronics and Information Technology (MeitY) as part of India’s AI Policy, priorities White Paper Series, highlights and elaborates on the importance of Foundation Models and India’s plans for building indigenous Foundation Models (FMs).

The paper clearly states how Foundation models form the core enabling layer in modern AI systems , which makes them transformative but also concentrates influence. 

Dependence on foreign models is highly undesirable as it will limit India’s ability to ensure transparency, inclusion and  alignment with its national goals. 

Indigenous Foundation Models, trained on more diverse datasets designed to reflect  India’s linguistic and social diversity will be more trustworthy and less biased and also mitigate the risk of under-representation which foreign models are most likely to exhibit. 

The white paper then goes onto explain the steps being taken to develop foundation models. Since the training of foundation models requires sustained access to compute and large datasets , the government is building a shared compute and data ecosystem for the same. The India AI mission addresses the data and compute requirements for building Foundation Models through the two dedicated pillars of the India AI Compute Portal(for compute) and the AI-Kosh-a unified repository of datasets.

Explained: What are Foundation Models? 

Foundation models are to AI , what operating systems are to computers. They form the underlying layer on top of which other tools and applications are built and function.

Foundation Models are essentially large deep learning neural networks which are trained on vast datasets that can perform a wide range of tasks and serve as the base upon which AI applications are built. They contain billions of parameters and can also have multimodal capabilities (they can understand and generate text, audio, images and other forms of data). 

By looking at the AI technology stack in its entirety , the significance of foundation models can be better understood. At the lowest level lies the energy infrastructure, the semiconductor chips and the data centers which form the physical base layer. The compute layer sits on top of this and it provides the massive processing capability required to train the complex models. The next layer consists of data infrastructure which has the large datasets on which the models are trained. Foundation models form the next layer and are a very crucial component of the entire AI stack.

At the top of the stack are the applications(the application layer) that everyday users interact with directly. The capabilities and performance of these applications are determined  by the foundation models on which they are built. Foundation models are thus perhaps the most important layer in the AI tech stack and which will determine how strong and robust a nation’s AI infrastructure is.

India Lists Framework for Governance of AI Foundation Models

The paper in its next section highlights the framework – the India AI Governance Guidelines 2025-  which will govern AI Foundation Models in India. The most noteworthy aspects are pertaining to the data protection laws, the copyright and IP policy pathways to make training data available and the benchmarks to measure the technical requirements.

The training of Foundation Models will make use of large collections of publicly available data. Additionally, the Indian government proposed a hybrid innovation model under which AI developers are to receive a “blanket license” for the use of all lawfully accessed content. Royalties become payable upon commercialization and will be managed through a centralized mechanism.

Foundation Models once trained will have to be evaluated by benchmarks that provide an objective and evidence based way to evaluate and regulate the performance of FMs.

What this means for Indian Startups ?

Historically speaking, building large scale AI systems required enormous resources which made it difficult for smaller companies to compete against global giants. By sharing compute infrastructure and providing open datasets the barrier to entry has been reduced drastically. 

The development of indigenous FMs allows startups to focus on building domain specific applications on top of foundation models. Startups can create India specific solutions using these India specific FMs. These FMs can be used to build innovative solutions for domains across agriculture, healthcare, finance, education and industry. 

Another promising area that the white paper also touches upon is the development of Small Language models or SLMs. SLMs can be designed and optimized for particular domains requiring less computational power making them suitable for emerging markets.

The Road Ahead

The development of indigenous FMs was essential to not only capitalize on the potential upside and benefits but also because the consequences of failing to develop domestic foundation models could prove to be costly – in terms of both monetary value in the long run and also in terms of autonomy and sovereignty. 

The development of Foundation models is ultimately about much more than technological capability. It is equally , if not more important that India participates meaningfully and proactively in shaping the future of AI and its own future in a world with AI.