On IT, India had early advantage. It hasn’t caught up with AI era

Why in the News

Anthropic suspended access to two of its frontier artificial intelligence (AI) models to comply with United States (US) Department of Commerce export controls. It restored access within three weeks. The episode shows frontier AI is now treated much like nuclear technology, and India, an early leader in IT services, has no frontier model of its own.

What is frontier AI, and who controls access to it?

1. What it is: Frontier AI means the most advanced general-purpose AI models, built by a handful of firms. Using them is like renting a service whose switch sits abroad.
2. Control of access: Countries that build these models can turn access on or off. The US and China have their own; India does not.
3. The takeaway: Without home-grown models, India’s use of the most advanced AI depends on another country’s export policy.

Why did India’s IT lead not carry into AI?

1. Low R&D: India spends about 0.84% of gross domestic product (GDP) on research and development (R&D), far below China and the US, so it lacks a base for frontier AI.
2. Services model under threat: Indian firms prospered by customising and maintaining software built abroad. AI can now write, test and maintain code itself, which strikes at India’s labour advantage.
3. Entry-level jobs: An Indian Council for Research on International Economic Relations (ICRIER) survey of 651 IT firms found 55% reporting a fall in entry-level employment, with senior-level jobs largely stable.
4. Resilience so far: Global capability centres (offshore units of multinationals) are booming, and AI-linked hiring still exceeds layoffs.

Is the IndiaAI Mission enough?

1. IndiaAI Mission: The Ministry of Electronics and Information Technology (MeitY) launched it in 2024 to build India’s AI ecosystem.
2. Shared compute: Its shared computing pool has grown to over 45,000 GPUs (graphics processing units, the chips that train AI), offered at subsidised rates to start-ups and researchers.
3. Scale gap: OpenAI’s projected 2026 costs for training and running its models, $45.8 billion, are not directly comparable with India’s budget but show the resource gap.
4. Ecosystem, not just money: China’s experience shows the task is to build a research and innovation ecosystem for home-grown models, not simply to spend more.

Does hosting data centres and chip units build AI capability?

1. Policy shift: Policy now focuses on attracting data centres, which matter for training domestic models and for data sovereignty (keeping data under Indian law).
2. Thin gains: Hardware is imported and most jobs end with construction, so the gain is small if the models running there are foreign-owned.
3. Incentives on offer: Foreign cloud providers serving global operations from Indian data centres get a tax holiday (tax exemption) until 2047; States offer cheap land, water and electricity.
4. Power demand: AI data centres are projected to add about 26.3 GW (gigawatts) of demand by FY32. They need round-the-clock power, which intermittent renewables cannot supply without costly storage.
5. Chip units: Of the 12 approved semiconductor facilities, nine only assemble, test and package chips made elsewhere. Hosting alone risks locking India into the lowest-value-added link, the stage of the chain that earns least.

Challenges

1. Night-time power gap: Until storage and firm power exist, coal partly fills the gap, so stretched grids and other consumers bear the cost.
2. No leading-edge chips: None of the approved facilities makes leading-edge AI chips, which would need a generational leap in manufacturing.
3. Investor caution: Demanding too much in return for incentives can deter investors.

Way Forward

1. Compute-access guarantees: Require data centre investors receiving state support to guarantee compute access for Indian users.
2. Sovereign cloud capacity: Secure cloud capacity under Indian control as a condition of incentives.
3. Local linkages: Tie support to local linkages in hardware and services, so infrastructure builds domestic capability.

Conclusion

In its “landlord role”, India hosts infrastructure for models built elsewhere, so support without conditions subsidises foreign firms instead of building Indian capability. The test is whether incentive deals attach such conditions without driving investors away.

Key numbers

1. R&D spending, FY24 comparison: China 2.6% and the US 3.4% of GDP.
2. IndiaAI Mission budget: about $1.2 billion over five years.
3. Anthropic’s projected 2026 costs: $19 billion.

Government Initiatives on AI and Semiconductors

1. BharatGen: A government-funded multimodal large language model for public services and Indian use cases.
2. Digital India Bhashini: Speech and translation tools across the 22 Scheduled Languages.
3. India Semiconductor Mission 2.0: Builds on a ₹76,000 crore framework, with fiscal support of up to 50% for fabs (chip factories) and chip design.

Matching Previous Year Question

“[2026, GS3, 15 marks] What is agentic Artificial Intelligence (AI)? Explain its working. Describe its applications with suitable examples. Discuss the advantages, risks and challenges associated with agentic AI systems.”