Artificial intelligence in India no longer moves in bursts. It moves in layers. Policy, capital, talent, and infrastructure are now building on top of each other in a way that is harder to separate into neat categories. The week between March 23 and March 29, 2026, offered a clear view of that shift. What looked like a series of announcements was, in reality, a coordinated push across government, industry, and startups to reshape how AI is built and used in the country.
Government Picks 10 Startups for Global Push
On March 25, the Ministry of Electronics and Information Technology announced a shortlist of ten AI startups selected for an international acceleration program under the IndiaAI umbrella. The program, linked with Station F and HEC Paris, is meant to place Indian companies directly into foreign markets rather than treating them as suppliers of backend services.
The startups selected reflect a pattern that has become more visible in recent months. Several are focused on healthcare diagnostics, where AI is used to process medical data in low-resource settings. Others are working on agricultural systems that rely on predictive models for crop management. A third group is building language models tailored for Indian languages, an area that has gained attention after the launch of sovereign models such as Sarvam AI and BharatGen.
Officials described the program as a way to move Indian AI companies closer to global capital and customers. That shift matters because many startups still rely on domestic demand, which can be uneven. Access to overseas markets changes the scale at which these firms can operate. It also exposes them to stricter standards on privacy and data use, which are increasingly part of international trade in technology.
Compute Support Expands Under IndiaAI Mission
A day later in Bengaluru, the government outlined how it plans to support domestic AI work through subsidised compute access. The IndiaAI Mission is now rolling out thousands of GPUs across the country, along with a network of data labs meant to lower the cost of training models.
The numbers are not small. More than 38,000 GPUs are expected to be available, with hundreds of data labs planned. For startups and research groups, compute cost has often been the biggest barrier. Renting capacity from foreign cloud providers adds both expense and dependency. By offering local alternatives, the government is trying to shift that balance.
This also ties into the idea of sovereign AI, where countries build their own systems rather than relying entirely on foreign models. The models introduced earlier this year have already been tested in areas such as agriculture and healthcare, where language and local context matter. The ability to train and run such systems domestically reduces exposure to external pricing and policy changes.
Industry responses suggest that this support could change who gets to participate in AI development. Smaller firms and academic groups, which often struggle to afford large-scale computing, may find it easier to build and test models. That could widen the range of applications being developed, especially in regions outside major cities.
Big Business Moves In on AI Infrastructure
While the government focuses on access, large companies are focusing on scale. In Mumbai, both Reliance Industries and Adani Group signalled that they are moving faster on AI-related investments following the February summit.
Reliance has indicated that parts of its large investment plan will be directed toward data centres and computing capacity. Adani is looking at energy-linked infrastructure that can support large computing clusters. These moves reflect a practical reality: AI systems require not just software but also power, cooling, and physical space.
The entry of such firms changes the structure of the market. Instead of relying only on foreign providers for large-scale compute, India may see domestic options emerge, backed by companies that already operate in energy and telecommunications. This could affect pricing, access, and even the location of data centres, which are often placed near power sources.
Executives have framed these investments in terms of job creation and long-term economic value. Whether that translates into widespread employment depends on how these facilities are used and who gains access to them. Data centres themselves do not employ large numbers of people, but they support industries that do.
Skilling Push Expands with New Centres
In Hyderabad, a quieter but equally important shift is underway. New centres focused on AI training have begun operations, tied to programs from NITI Aayog and MeitY. These centres are working on advanced areas such as agent-based systems that can carry out multi-step tasks with limited human input.
India already has a large pool of software engineers, but AI work requires a different set of skills. Training programs are now focusing on model development, data handling, and responsible use of AI systems. There is also an effort to expand participation beyond major cities, with training reaching smaller urban areas.
One area receiving attention is inclusion. Programs aimed at women and underrepresented groups are being expanded, partly to address gaps in the workforce. Another focus is on practical use cases, where trainees work on real-world problems rather than abstract exercises.
The scale of this effort reflects the belief that talent will determine how far India can go in AI. Without enough trained workers, even large investments in infrastructure may not produce the desired results. The link between skilling and deployment is becoming clearer as more companies look for people who can move from theory to application.
Funding Momentum Builds for AI Startups
By March 28, attention shifted back to capital. Indian startups raised more than $200 million during the week, with a large share going to AI-focused firms. The funding came across stages, from early seed rounds to larger investments in companies already generating revenue.
This surge follows policy moves such as the Startup India Fund of Funds 2.0, which targets deep-tech companies. Investors appear to be responding not just to individual companies but to the broader direction of policy and infrastructure. When compute becomes cheaper and markets expand, the risk profile of startups changes.
The types of companies receiving funding also point to where demand is growing. Language models, voice-based systems, and enterprise tools are drawing attention. These are areas where Indian companies can build products suited to local needs while still selling abroad.
At the same time, challenges remain. Hardware development, in particular, continues to lag behind software. Building chips and specialised equipment requires a different scale of investment and longer timelines. Data governance is another area where rules are still being shaped, affecting how companies collect and use information.
Government Turns to Agent-Based Systems for Public Services
The week began with discussions in New Delhi about a concept that is moving from theory into practice: agent-based AI systems. These systems are designed to handle tasks that involve multiple steps, decisions, and interactions, rather than simple responses.
Under the IndiaAI Mission, pilot projects are being considered for areas such as agriculture support, healthcare triage, and administrative workflows. The idea is to reduce the load on human staff while improving response times for citizens.
This raises practical questions. Data privacy remains a concern, especially when systems handle personal information. There is also the issue of accountability. When decisions are made by automated systems, it becomes harder to assign responsibility for errors.
Still, interest in these systems reflects a shift in how AI is being used. Early applications focused on generating text or analysing data. The next stage involves systems that act, not just respond. That step increases both the potential and the risk.
