India's AI development has been constrained by two persistent financial gaps: research and development spending hovers at just 0.64% of GDP, and startup investment-until a sharp turn in 2026-trailed global peers by billions of dollars. The IndiaAI Mission and the first sovereign large language model releases signal a deliberate policy reversal, though scaling to frontier capability remains a heavy lift.
Research and development funding gap
India's gross expenditure on R&D as a share of GDP sits well below the global average. China allocates roughly 2.6% of GDP, and the United States about 3.6%, according to the Lowy Institute. Private industry funds only 36% of India's total R&D, compared with over 70% in the US, China, and South Korea, an analysis by the Carnegie Endowment for International Peace found. The gap is visible in AI patent grants, where India trails the top two nations in both volume and quality.
Investment in AI startups reveals a capital chasm
Indian AI startups raised $643 million across 100 deals in 2025, a 4.1% increase, while US AI funding surged 141% to $121 billion over 765 rounds, according to Tracxn data reported by TechCrunch. Investors directed the limited Indian capital toward application-layer companies, steering clear of capital-heavy foundational model builders. The first quarter of 2026 brought a sharp reversal: Indian AI startups raised $3.94 billion, led by Neysa's $600 million round and momentum from the AI Impact Summit. New Delhi reinforced the shift with a $1.15 billion Fund of Funds and a Rs 1 trillion ($12 billion) Research, Development and Innovation scheme spanning AI, quantum computing, and deep tech. Much of this remains at the announcement stage, so its durability is unproven, but the funding direction has clearly changed.
Research output: China leads, India climbs slowly
China dominates high-impact research across 66 of 74 critical technologies tracked by the Australian Strategic Policy Institute's Critical Technology Tracker, published in December 2025. India's ranking improved to 50th from 43rd a year earlier, but the country's weakness is concentrated in compute infrastructure and frontier AI model categories, rather than the breadth of its research base.
India's sovereign AI model bet
The IndiaAI Mission selected Sarvam AI in April 2025 to build the country's first sovereign large language model ecosystem. Sarvam's original mandate specified a 120-billion-parameter model, but in early 2026 it released two open-source models-Sarvam-30B and Sarvam-105B-under the Apache 2.0 licence on Hugging Face and AIKosh. Both are mixture-of-experts reasoning models trained from scratch using government-supported compute infrastructure. A separate awardee, Soket AI, retains the 120-billion-parameter model mandate.
The training process exposed the hardware constraints. Sarvam's models were trained on a few thousand GPUs over several months, far below the cluster-months of more than 10,000 GPUs typical for frontier-scale AI training. All sovereign model training continues to rely on imported NVIDIA chips, underscoring India's dependence on foreign semiconductor hardware. The IndiaAI Mission has now demonstrated that the country can build and release competitive open-source large language models, but closing the capability gap in frontier AI systems remains the harder challenge. For teams working on Generative AI and LLM applications, these open-source models offer a new base to fine-tune and benchmark against, though compute requirements still limit large-scale adoption.
Why this matters for IT and development professionals
The shift from an application-only AI market to one that includes sovereign foundational models changes the technical workload for IT teams. Deploying mixture-of-experts models like Sarvam-105B on limited GPU clusters demands optimization skills in model quantization, distributed inference, and memory management. As government-backed compute infrastructure scales, professionals will need to design systems that balance on-premise and cloud resources while keeping model performance within cost constraints. The policy push also creates demand for expertise in fine-tuning open-source LLMs for domain-specific tasks, a practical area where India's AI talent can build an edge. For those tracking this evolution, the AI for IT & Development resources provide guidance on deployment patterns and tooling choices that align with these infrastructure realities.
Your membership also unlocks:
