India has the AI talent, but the IP is leaving: MeitY’s Abhishek Singh flags the brain drain problem

Additional Secretary Abhishek Singh outlines plan to scale GPU capacity and back sovereign AI models through private investment and domestic talent funding.

• India aims to retain talent and create AI IP domestically.
• MeitY to incentivise private investment in GPU infrastructure.
• Govt to fund AI adoption in health, education sectors.

Did our AI summary help?

India must invest in its own talent and compute infrastructure to ensure intellectual property is created domestically rather than overseas, Abhishek Singh, Additional Secretary at the Ministry of Electronics and Information Technology (MeitY), said in a conversation with Dr Nalin Mehta, Managing Editor, Moneycontrol.

“Our biggest strength has been talent. Almost every Big Tech company relies a lot on Indian talent. Unfortunately, our talent leaves India and goes to the West where they are able to access funding and support from the Big Tech ecosystem and create IP there,” Singh said.

He added that the recent AI Summit had shifted the conversation towards retaining and funding talent within India. “What the AI Summit has changed is that people are looking if the talent is here, why don’t we invest in the talent within India, so that IP gets created here,” Singh said.

Singh said India currently has access to 62,000 GPUs, which he described as sufficient to train the first set of indigenous AI models.

He said that if 10–20 crore Indians begin using AI services across sectors, India would require between 1 lakh and 2 lakh GPUs to meet demand.

He clarified that 38,000 GPUs were not procured directly by the government. “Had we procured it, it would have cost Rs 20,000 crore,” Singh said. Instead, MeitY incentivised private sector investment in compute infrastructure. He said the actual compute expenditure exceeds Rs 20,000 crore.

Singh also said the government is preparing a policy to incentivise private investment in GPU infrastructure, with part of the capacity to be offered to the wider AI ecosystem.

In strategic sectors such as defence, he said the government would establish its own GPU infrastructure within government data centres. For public sector applications in agriculture, education and healthcare, inferencing costs would be funded by the government in multiple ways.

Singh said Indian AI models improve with usage and that broader adoption would enhance their performance over time.

“In many Indian language benchmarks, Sarvam’s model has performed way better than any other top models,” he said, adding that sovereign Indian models can compete globally provided there is domestic usage.

He said comparisons between India’s AI investments and those in Western countries often overlook the role of private capital abroad. Referring to OpenAI’s proposed $500 billion Stargate initiative, Singh said it was not a US government investment but a private sector initiative, adding that some of that investment would also come to India.

He also referred to a proposal within a Working Group recommendation to set up a common compute facility to service countries in the Global South.

India has been positioning itself to build sovereign AI capabilities through a combination of public funding and private sector participation. The India AI Mission has prioritised access to compute infrastructure, model development and sectoral AI deployment.

In recent months, policymakers have emphasised the need to reduce reliance on foreign AI platforms and retain domestic talent. India’s technology workforce forms a significant part of global AI and software teams, with senior engineers and founders contributing to major US-based AI firms.

At the same time, access to high-performance GPUs has emerged as a key constraint globally, with governments and companies competing to secure compute capacity for training and deploying large language models.

Singh’s remarks come amid efforts to scale AI adoption across public services and strategic sectors, including agriculture advisory systems, healthcare diagnostics, educational tools and defence applications.

The government’s approach, as outlined by Singh, relies on incentivising private investment rather than direct state procurement of large-scale GPU clusters, while retaining strategic control over sensitive deployments.