India is evolving from an AI adopter to AI leader. Here's how

The private sector—including global giants Google, Microsoft and Nvidia—is collaborating with Indian startups and institutions to advance AI applications. Companies like Qure.ai and Wipro’s AI Research Lab focus on agriculture, health care, and finance, whereas startups like Cropin and Sarvam innovate in precision farming and conversational AI. “Challenges like talent migration, limited digital infrastructure and unstructured data may hinder progress, though the ecosystem is moving in the right direction,” says Counterpoint’s senior analyst Prachir Singh.

No wonder that OpenAI CEO Sam Altman, during his India visit earlier this year, remarked that the country had become his company’s second-largest market by the number of users, which had tripled in the past year. “What’s happening with AI adoption in India right now is amazing to watch,” he said. “We love seeing the explosion of creativity–India is outpacing the world.”

Different Strokes

The US has a market-led approach to AI development. Tech giants there have driven breakthroughs in LLMs and generative AI, leveraging their dominance in information economies and social media to access vast amounts of user-generated content.

China has a state-led approach, guiding industry leaders such as Tencent, Baidu and SenseTime to develop core AI capabilities in computer vision, autonomous driving and facial recognition. The aim is to enable smaller enterprises to access technology at a lower cost.

India, having rapidly adopted AI technologies developed in the US and China, is now building its own foundational models, prioritising responsible AI, infrastructure, data ethics and skilling. The growing partnership between the government and industry is likely to play a critical role. Experts believe this could lead to a bright future.

Scaling AI-related research and development, improving access to high-quality datasets, ensuring data privacy and promoting equitable access to AI benefits for all are the crucial next steps, says Tuteja of KPMG India.

India recognised the need for a strategic approach after struggling to catch up with global AI advancements, triggered by the launch of OpenAI’s models and China’s DeepSeek. This led to the development of a seven-pillar approach through a consultative process with key stakeholders and the launch of the IndiaAI Mission in March 2024, which aims to address the gaps, including compute capacity, lack of Indian foundation models, need for dataset platforms, support for skilling and startups, and development of applications and tools for safe and trusted AI.

The Union government has allocated ₹10,300 crore over five years to strengthen AI capabilities through the IndiaAI Mission, which focuses on developing a high-end common computing facility with 18,693 GPUs to foster indigenous AI models, enhance compute infrastructure and promote AI adoption.

GPU, short for graphics processing unit, is an electronic circuit designed to speed computer graphics and image processing. A GPU reduces the time needed for a computer to run multiple programmes, making it a vital cog in the wheels of emerging technologies.

In April, Bengaluru-based Sarvam AI became the first startup selected under the mission to build a foundational model which would be completed in six months, said co-founder Vivek Raghavan at an event where the announcement was made. Sarvam will get access to 4,096 Nvidia H100 GPUs for six months from the IndiaAI Mission’s common compute cluster to train its model. 

Also read: India developing unique AI regulation model: Ashwini Vaishnaw

Foundational Models versus Applications

Building AI is about connecting data, open protocols and people in a way that benefits everyone. After the initial euphoria around LLMs died down, relevance and context became the key to getting more productivity out of AI systems, says Shankar Maruwada of EkStep Foundation, which was co-founded with Nandan and Rohini Nilekani to champion ecosystems that build digital public goods.

“The new world would be multi-model and multi-modal, and so a single model would never be prominent,” says Maruwada. “It is only the beginning and with our vibrant technology ecosystem, India should focus on investing in early research on AI technologies with Indian datasets and build AI systems that our citizens need.”

Though some argue that India should let the Silicon Valley giants handle the development of LLMs, Sunil Gupta, co-founder, MD & CEO of Yotta Data Services, a data centre firm, believes the rise of generative AI has made countries invest in sovereign AI, which refers to a nation’s capabilities to produce AI using its own infrastructure, data, workforce and business networks. Having missed out on earlier opportunities in data centres and cloud computing, India can build a robust AI compute infrastructure, paving the way for the development of new services, says Gupta.

India needs a foundation model for three major reasons, says Union minister for electronics and information technology Ashwini Vaishnaw in an interview with Forbes India (see Q&A on page 20). First, the country’s cultural heritage and linguistic strengths must be reflected in the models. Second, the biases that exist in many parts of the world need to be kept out. And third, these technologies are already gaining strategic importance and it is important for India to have its own models. 

Generational Opportunity

Sumangal Vinjamuri, an AI and SaaS investor at Blume Ventures, says that currently most AI application activity from Indian founders is in the B2B software space.

 “There are three categories of applications we are most excited about: Full-stack vertical AI that can boost productivity and outcomes in specific industries, services-as-software that can deliver traditional human-like delivery with software margins, and platforms and tooling that can help enterprises accelerate building and adoption of AI workflows internally,” he says.

On the investment front, Vinjamuri believes that venture capital firms averse to investing in AI startups will be missing a generational opportunity to back the next wave of software entrants into the Fortune 500. “One challenge we have seen broadly while investing in the space in India is finding teams that are able to stay at the edge of where the state-of-the-art is and execute both product build and go-to-market execution at a high pace, consistently,” he says.

Despite the hurdles, India’s AI future is promising, with sovereign foundational models under development and a focus on indigenous LLMs tailored to local languages, Vinjamuri adds.

Alok Goyal, partner at Stellaris Venture Partners, has been cautious about India’s ability to capture opportunities in the upstream part of the AI value chain, such as GPU manufacturing, LLM development, and AI infrastructure. “While these opportunities exist, we are not convinced India has the strength to capture them yet. However, we believe in supporting these entities through long-term capital,” he says.

Consider the internet value chain in the 1990s. India did not capture the upstream opportunities like networking chips and hardware but still captured significant downstream value through IT-enabled services, generating a market cap of half a trillion dollars. This shows India can create value even if not dominant in the upstream part of the value chain. “I believe our bread and butter will come from applications, where India excels,” says Goyal.

Today, the country has the opportunity to leverage its experience in software and services, combined with the new capabilities offered by AI, to reimagine processes and create innovative offerings for the global market.

More than two years after ChatGPT’s launch, AI remains a focus area for venture capital and the business world, with startups shifting from developing new models to creating practical applications in fields such as agriculture, health care and finance.