Enterprise generative-AI adoption accelerating across BFSI, IT services and manufacturing, tracked sector-by-sector in NASSCOM's AI Adoption Index (published with EY).
AI Market in India
India's AI market is in an early, fast-moving, and still hard-to-size phase. Vendor market-size estimates for 2025 range from roughly $10 billion to $13 billion, with 2030-2035 forecasts diverging even more sharply depending on methodology (compound annual growth rate estimates from research firms range from about 18% to 39%). What is more firmly documented is the surrounding activity: the Indian government approved a ₹10,372 crore (~$1.25 billion) IndiaAI Mission in March 2024 to subsidize GPU compute access for startups, academia and industry; Microsoft, Google and Amazon collectively pledged roughly $67.5 billion in India-specific AI infrastructure between October and December 2025, including Google's $15 billion gigawatt-scale data-center campus in Visakhapatnam; and Indian AI startups raised $1.3 billion in 2025, roughly 3x the $430 million raised in 2024. At the same time, the regulatory foundation for AI-specific liability is not yet settled: India notified the Digital Personal Data Protection Act Rules in November 2025 but most substantive obligations only take effect from May 2027, and the government has explicitly opted not to pass a standalone AI law, instead extending the DPDP Act, IT Act and existing IP law to cover AI systems.
What this market includes.
The intersection of the Artificial Intelligence industry with the Indian market: AI software, cloud/compute infrastructure, and services (including generative-AI applications) that are built in India, sold into India, or consumed by enterprises, government bodies and consumers operating in India. This covers three layers that move somewhat independently: (1) the compute/infrastructure layer (data centers, GPU capacity, government-subsidized compute), (2) the enterprise/IT-services layer (AI-augmented delivery by Indian IT majors and global-capability-centre (GCC) adoption), and (3) the domestic startup/foundation-model layer (India-focused large language models and GenAI applications). It excludes AI activity by Indian-founded companies operating purely outside India, and excludes hardware manufacturing (semiconductors) except where directly tied to AI compute deployment.
The market in numbers.
Every figure below carries its source and the date it was last verified, per Analyzing Markets' trust standard.
What's driving demand.
The structural and cyclical forces behind current and forecast demand in this market.
Government-subsidized compute: the IndiaAI Mission subsidizes roughly 40% of GPU cost, targeting under $1/hour GPU access to lower the capital barrier for startups, academia and researchers to train and serve models domestically.
Hyperscaler capital expenditure landing inside India: Microsoft's $17.5B and Google's $15B India-specific commitments (the latter a gigawatt-scale data-center campus in Visakhapatnam built with Airtel and AdaniConneX) are moving AI compute capacity onto Indian soil rather than serving India from overseas regions.
A large, comparatively low-cost, English-fluent technical talent base: India has the world's second-largest installed AI talent pool per NASSCOM, with close to 70% of early-career tech talent rated 'AI-proficient' and about 23% 'AI-native.'
Indian IT-services majors (TCS, Infosys, Wipro, HCLTech) re-platforming existing global outsourcing and GCC (global capability centre) relationships around AI-augmented delivery, converting existing enterprise contracts into new AI-specific revenue lines rather than requiring net-new customer acquisition.
Domestic sovereign-AI and multilingual-model ambition: startups such as Sarvam AI and Krutrim, backed by government interest in India-specific foundation models, target the roughly 3.7x year-on-year surge in GenAI-startup formation NASSCOM documented for 2025.
How the market is structured.
The market has a two-tier structure with very different concentration at each layer. At the compute/infrastructure layer, supply is concentrated among a handful of global hyperscalers (Microsoft, Google, Amazon/AWS, Nvidia as the dominant GPU supplier) plus the government's own IndiaAI Mission compute facility; this layer is capital-intensive and consolidating. At the application/services layer, concentration is low: NASSCOM's 2025 landscape mapping counts well over a hundred active Indian GenAI startups, alongside India's large IT-services majors (TCS, Infosys, Wipro, HCLTech, Tech Mahindra) monetizing AI primarily through enterprise services and GCC delivery rather than product sales. Competitive intensity for enterprise AI-services contracts is high because Indian IT majors compete both against each other and against the same global hyperscalers now investing directly in India, while the startup layer competes for a talent pool NASSCOM's own data suggests is not growing fast enough to meet 2027 demand.
Regulatory environment.
India has chosen not to pass a standalone AI law. MeitY notified the Digital Personal Data Protection (DPDP) Act and its Rules on 13 November 2025, but most substantive compliance obligations only take effect 18 months later, from 13 May 2027. Alongside the Rules, MeitY released IndiaAI Governance Guidelines in November 2025 confirming the government's approach: extend the DPDP Act, the IT Act and existing intellectual-property law to cover AI systems (including algorithmic due-diligence duties for significant data fiduciaries) rather than legislate AI specifically. For a market this reliant on foreign hyperscaler capital and cross-border data flows, this leaves a roughly two-year window in which the operative AI-specific compliance regime is still provisional.
Where the openings are.
Entry points that current market structure and demand trends make more attractive right now.
- Subsidized GPU compute access under the IndiaAI Mission (targeting under $1/GPU-hour) lowers the capital barrier for startups and mid-size enterprises to train or serve models domestically rather than renting overseas capacity.
- Enterprise GenAI-adoption gap outside BFSI and IT services: NASSCOM's sectoral Adoption Index shows manufacturing, retail, healthcare and public-sector AI adoption still lag the leading sectors, leaving unaddressed demand.
- India-specific and multilingual foundation-model tooling remains comparatively underserved relative to English-first global models, an opening domestic players like Sarvam AI and Krutrim are targeting.
- Downstream services around the hyperscaler capacity build-out (Microsoft, Google, Amazon): MLOps, data-labeling, model-evaluation and AI-compliance advisory demand should scale with the underlying infrastructure investment.
What could go wrong.
Conditions that would weaken or invalidate the opportunity above.
- Talent demand-supply gap: Deloitte-NASSCOM project India's AI-talent demand rising from roughly 610,000-650,000 to more than 1.25 million by 2027, a pace of growth (25-35%) that could outstrip supply and push up delivery costs and wage inflation.
- Regulatory uncertainty: the DPDP Act's substantive obligations do not bite until May 2027, and the government has explicitly deferred a standalone AI law, so the AI-specific liability and governance regime enterprises will actually operate under in two years is not yet settled.
- Compute-supply dependency: sovereign compute build-out still depends on imported GPUs (chiefly Nvidia) and on hyperscaler capex decisions made outside India, exposing the market to global chip-allocation and export-control risk.
- Market-size estimate divergence: leading research firms' 2025 India AI market-size figures vary by more than 25% ($10.15B-$13.05B) and their 2030-2035 growth forecasts vary far more (CAGR estimates from roughly 18% to 39%), a sign this market is still too early-stage to size with precision; any single point estimate should be treated as directional, not exact.
Related market intersections.
Other industry-geography pairings connected to this market.
Artificial Intelligence -- global industry pillar
VIEW ANALYSIS →India -- country pillar
VIEW ANALYSIS →Where this came from.
Every source cited on this page, in the order it was added.