India’s digital success story is dazzling. From UPI to Aadhaar to ONDC, the nation has built public technology that now serves over a billion people. We have mastered the art of rapid innovation — moving fast, iterating faster, and creating products that scale effortlessly. Yet beneath this triumph lies an imbalance that threatens long-term resilience. India has become exceptional at building user-facing applications but remains weak in creating the deep scientific and infrastructural foundations that power them.
For over a decade, India’s research and development expenditure has stagnated at around 0.6-0.7 per cent of GDP, far below countries like China or South Korea, which invest over 2 per cent. While we celebrate technological adoption and digital inclusion, we have neglected the invisible layers that sustain it — compute infrastructure, chip design, and foundational algorithms. UPI’s 20-billion-plus monthly transactions demonstrate world-class execution, but not technological independence. Every major Indian start-up — from fintech to e-commerce — relies heavily on foreign cloud providers and machine learning APIs. If those providers were to alter pricing or policies, much of India’s digital ecosystem would face severe disruption.
India’s dependency makes us fast at the surface but fragile at the core. The problem is not lack of ambition but of structure. Our venture capital ecosystem rewards speed over depth. Investors chase traction within months, not discoveries that require years of patient work.
In the public sector, procurement processes still prioritise low-cost, quick delivery over innovative research. Success is measured by compliance and completion, not discovery. Talented researchers face a stark choice between underfunded academic labs and start-ups that prioritise immediate outputs over long-term inquiry. Over time, this erodes the culture of scientific curiosity. Companies, too, remain fixated on vanity metrics — downloads, user counts, and transactions — while neglecting the creation of open datasets, foundational models, and reusable benchmarks that build lasting capability. The result is an ecosystem optimised for speed rather than substance.
Recent government initiatives such as the IndiaAI Mission, BHASHINI language platform, and India Semiconductor Mission mark encouraging progress. They recognise that compute, data, and hardware matter as much as applications. But declarations alone cannot bridge a decade-long gap. What India needs is a decade of patient capital, sustained scientific depth, and a national compact to own the digital foundations of the future. A path forward begins with rebuilding the core-compute, data, and talent. Compute must be treated as public infrastructure, not a private luxury. India must now democratise access to high-performance computing through a National Compute Grid. This grid could make GPUs and research infrastructure accessible to universities, start-ups, and MSMEs under a transparent, fair, and innovation-friendly model. Simultaneously, India should establish an Open Data and Benchmarks Facility that curates anonymised datasets and standardises them for research use. Public leaderboards and benchmarks would drive accountability and foster innovation. Investing in talent is equally essential — offering doctoral fellowships paired with guaranteed compute access and creating return pathways for scientists abroad.
The second step is reforming rules — from price to purpose. Government contracts must reward value creation, not just low cost. Procurement should be linked to measurable outcomes such as energy efficiency, model accuracy, or benchmark improvement rather than project completion alone. A protected line of funding should be created for long-cycle, curiosity-driven research insulated from budget cuts. Public workloads — citizen chatbots, translation tools, and knowledge retrieval systems — should preferentially use models trained in India. Sovereign demand will nurture ecosystems; ecosystems will build resilience. Finally, partnerships among government, industry, and academia must be rewired.
Success must be measured differently. Within two years, India should host at least three open, India-grade large language models trained on diverse datasets. Dozens of public benchmarks should exist across language, speech, and code. Academic compute access should be available within a week, and at least two Indian chip design houses should produce inference-grade silicon within three years. R&D spending must steadily rise towards one per cent of GDP. Digital sovereignty is not isolation; it is independence. India doesn’t need to replicate China’s walls or Silicon Valley’s wealth. It needs its own model-rooted in openness, shared infrastructure, and patient collaboration. The journey from “apps before atoms” to “atoms before apps” will be neither quick nor glamorous, but it is essential if India is to stand tall on foundations it built itself, in languages it owns, for citizens it empowers.
The writer is Chief AI and Digital Consultant, Wadhwani Foundation