Frugal AI: Why India Must Build Practical AI Solutions Instead of Following Silicon Valley

Artificial intelligence is advancing rapidly, but building powerful AI systems is only one part of the challenge. Making these technologies affordable, practical and sustainable for everyday use is becoming equally important. Serish Gandikota, co-founder of the Frugal AI Hub at the University of Cambridge’s Judge Business School, believes countries such as India should develop AI solutions around their own needs rather than simply follow trends emerging from Silicon Valley.

Gandikota’s approach, known as frugal AI, focuses on achieving useful outcomes with limited resources. Instead of concentrating only on developing larger models or investing in expensive computing infrastructure, it encourages organisations to examine the problem they want to solve and determine whether AI is the most appropriate solution.

What Is Frugal AI?

Frugal AI is not simply about making existing technology cheaper. It involves rethinking how a problem can be solved while using resources efficiently.

Gandikota draws a parallel with affordable medical innovations, including a portable baby warmer developed as an alternative to conventional incubators. The example demonstrates how redesigning a solution can make an essential service accessible in environments where electricity, funding or infrastructure are limited.

The same principle can be applied to AI. Organisations should first identify the outcome they want to achieve and then select the technology, model and computing infrastructure that best fits their requirements.

This approach also requires effective governance, well-defined workflows, reliable data and change management. Without these foundations, even an advanced AI system may fail to deliver lasting results.

Moving Beyond AI Experiments

Many companies and public institutions have tested AI through pilot projects, but turning those experiments into systems that work consistently at scale remains difficult.

The Frugal AI Hub, launched at Cambridge in April 2025, works on helping organisations move from initial trials to sustained deployment. Its work includes examining the total cost of AI projects and assessing whether the benefits justify the investment.

The initiative has also explored deployment partnerships beyond Cambridge. An agreement with the Andhra Pradesh government, signed in November 2025 under the Amaravati Quantum Valley initiative, is intended to support collaboration on frugal AI and the development of a deployment lab in Amaravati. Discussions with Telangana have also been reported.

The central question is not simply whether an organisation can adopt AI, but whether it has the resources, processes and institutional capacity to use it effectively over time.

Why India Needs Its Own AI Strategy

India faces different infrastructure and resource constraints from those encountered by major technology companies in the United States. Competing directly in the race to build ever-larger AI systems could require substantial investment in computing capacity, electricity, cooling and data centres.

Gandikota argues that India should not assume it must replicate the same approach. Instead, it can build on its experience with digital public infrastructure, including Aadhaar and UPI, while developing AI systems that address local challenges.

Language is another important consideration. Many of the world's languages remain poorly represented in AI systems, creating barriers for people who do not use widely supported languages. India's linguistic diversity makes locally relevant language technology particularly significant.

AI tools must also be dependable when used in areas such as agriculture, where inaccurate information can affect livelihoods.

Practical Applications Matter More Than Hype

One example of frugal AI in India is MahaVISTAAR, a voice-based advisory service for farmers. Such systems must handle regional dialects, pauses in speech and background noise while providing accurate, up-to-date information.

Achieving this requires more than a capable language model. The supporting information must be maintained, relevant institutions must coordinate their work, and the system must deliver reliable advice in real-world conditions.

The example highlights a wider lesson: successful AI deployment depends on the complete system surrounding the technology.

The Future of Sustainable AI

Frugal AI does not reject advanced models or large data centres. Rather, it distinguishes between developing cutting-edge technology and deciding what an organisation actually needs.

As businesses and governments increase their AI investments, the ability to deploy, manage and sustain these systems could become a major source of advantage. For India, the opportunity lies in combining technological capability with affordable infrastructure, local knowledge and practical problem-solving.

Ultimately, the success of AI will depend not only on how intelligent systems become, but also on whether institutions can put them to effective and lasting use.