In a world where OpenAI, Google, and Microsoft are racing to build ever-larger AI models at skyrocketing costs, India is charting a different path. The country's frugal AI startups — led by companies like Sarvam AI, Krutrim, and AI4Bharat — are proving that artificial intelligence doesn't need to be expensive to be effective. Their mission: bring AI to 1.4 billion people, not just the 1% who can afford $200-per-month ChatGPT Plus subscriptions.
The contrast could not be starker. Global AI models cost approximately $2.50 to $3.00 per hour for computation. India's government-subsidized indigenous AI model will cost less than ₹100 per hour — roughly $1.20 — after a 40% subsidy. That's not just a price difference; it's a philosophical divide about who AI is built for.
The Language Tax
India is home to over 1,600 dialects and 22 official languages. When Silicon Valley builds AI models trained primarily on English data, they create what researchers call the "language tax" — the extra cost and reduced performance that comes from trying to force a monolingual system into a multilingual reality.
"When ChatGPT came out in November 2023, it simply blew my mind," says Vivek Raghavan, co-founder of Sarvam AI and former researcher at AI4Bharat, an initiative launched at IIT Madras in 2020. "Here was a truly deflationary technology in every sense of the term. I immediately saw a way for India to achieve huge breakthroughs in health and education. An AI tutor would cost far less than today's schools and provide much more personalized instruction."
But Raghavan recognized that global models wouldn't work out of the box. They needed to be affordable, scaled for India's infrastructure constraints, voice-enabled, and built in Indian languages. That realization became the founding principle of Sarvam AI — and a growing movement of Indian startups committed to frugal AI.
Frugal by Design
Frugal AI isn't just about cutting costs — it's about building leaner, more efficient systems from the ground up. Unlike the brute-force approach of scaling model sizes indefinitely, frugal AI prioritizes efficiency, smaller footprints, and deployment on low-end devices.
The Saving Voices Project recently built a speech AI system for the Soliga tribe in southern India. With just five hours of voice data, researchers created a text-to-speech model that runs on low-powered devices and can operate offline for extended periods. The Soliga language has no written script, fewer than 2,000 speakers, and no internet access — making it invisible to commercial speech technology. Frugal AI changed that.
"By design, these systems use less compute, less memory, and less energy, which directly translates into a smaller carbon footprint," says Arjuna Sathiaseelan, founder of the Saving Voices Project and CTO of the Frugal AI Hub at Cambridge University. "This is perhaps the most important dimension of frugal AI — it's about building efficient systems, not just cheap ones."
The Sovereign AI Movement
India's push for sovereign AI isn't just about affordability — it's about strategic independence. With U.S. and Chinese companies operating more than 90% of the world's AI data centers, countries like India face the prospect of digital colonialism: dependent on foreign infrastructure, subject to foreign policy shifts, and unable to control their own technological destiny.
The Indian government has responded with the IndiaAI Mission, offering subsidized compute capacity and supporting domestic model development. Union Minister Ashwini Vaishnaw announced that India's indigenous AI model would cost significantly less than global alternatives, with high-end computing available at ₹150 per hour and common compute access dropping below ₹100 per hour after subsidies.
Reliance Jio, India's largest telecommunications company, has committed $110 billion to building AI infrastructure, promising to bring the same "extreme affordability" it brought to mobile communications to the AI era. Akash Ambani, head of Reliance Intelligence, announced plans to make AI "dramatically more affordable" for every Indian by 2030.
Startups Leading the Charge
Beyond government initiatives, Indian startups are delivering frugal AI solutions across healthcare, education, agriculture, and legal services.
Sarvam AI is deploying voice-enabled, multilingual conversational agents that allow rural patients to access medical advice, schedule appointments, and consult doctors through WhatsApp and low-bandwidth interfaces. Their Sarvam 2B and Sarvam-M models are optimized for Indian languages and run efficiently on smartphones.
Krutrim, founded by Bhavin Kirit Shah, is building sovereign AI models for India while also creating consumer applications. The company has positioned itself as India's answer to ChatGPT, with models trained on Indian data and optimized for Indian use cases.
BharatGen offers the Param-2 model at just ₹5 per million output tokens — a fraction of the cost of global alternatives. Sarvam charges ₹10 for its 30-billion-parameter model and ₹16 for its 105-billion-parameter version, making enterprise-grade AI accessible to startups and small businesses that would be priced out of OpenAI or Anthropic.
Other notable players include Haptik (conversational AI for enterprise), Yellow.ai (AI automation), Qure.ai (medical imaging), and Fractal Analytics (enterprise AI), each bringing frugal innovation to their respective domains.
The Global Implications
India's frugal AI movement isn't just relevant to India — it offers a blueprint for the Global South. From Indonesia to Nigeria to Brazil, countries that have been priced out of the AI revolution are looking to India's approach as proof that sovereignty and affordability are achievable.
The launch of DeepSeek in China last year energized advocates of frugal AI worldwide. China's open-source models have become foundations for developers globally, while countries including India, Mexico, and Malaysia aim to reduce their reliance on expensive chip imports and build domestic AI capabilities.
"The current trajectory of AI development is unsustainable economically, environmentally, and socially," says Sathiaseelan. "Model sizes have exploded, leading to significant energy and water consumption, and yet billions of people remain excluded from AI's benefits. Frugal AI addresses these failures."
🔥 Hot Takes
1. The "bigger is better" approach to AI is a luxury problem. Silicon Valley's arms race for ever-larger models is solving a problem most of the world doesn't have: unlimited compute budgets. India's frugal approach proves that 80% of the value can be captured with 20% of the resources — if you design for constraints from the start.
2. Language diversity is a feature, not a bug. The global AI industry treats multilingual support as an afterthought — bolted on after the English model is built. Indian startups are building multilingual from day one, creating systems that are actually usable by non-English speakers. This isn't just inclusive design; it's better engineering.
3. Frugal AI is the only path to mass adoption in the developing world. A ₹399 ($4.75) ChatGPT subscription is still too expensive for most Indians. When Reliance Jio promises to make AI "dramatically more affordable," they're not being charitable — they're recognizing that the only way to serve 1.4 billion people is to make AI as cheap as a cup of chai.
The Bottom Line
India's frugal AI revolution is more than a market strategy — it's a philosophical statement about who technology is for. While Silicon Valley builds monuments to compute power, Indian startups are building tools for real people: farmers using voice AI for crop advice, rural patients accessing telemedicine, students learning in their mother tongue, indigenous communities preserving endangered languages.
The question isn't whether frugal AI can compete with expensive alternatives. The question is whether the expensive alternatives can survive when billions of users simply can't pay the price. India's answer — sovereign, multilingual, affordable AI — may well define the next era of artificial intelligence.
This story is based on reporting from Rest of World, IndiaAI.gov.in, The Economic Times, and Mint. Startup cost data verified from public filings and interviews.