BharatGen, India’s government-backed sovereign AI consortium, has warned that the country risks becoming a permanent consumer rather than a producer in the global AI economy unless it builds its own foundational models. BharatGen chief executive Rishi Bal made the warning in an interview with Business Today on the sidelines of the India AI Impact Summit 2026, where the consortium launched its 17-billion-parameter multilingual model, Param2.
Bal’s central argument is that cheap access to foreign AI tools is not a neutral convenience. He described global AI firms as effectively “price dumping” services into India, arguing that low-cost foreign models are crowding out the conditions needed for a domestic AI ecosystem to take root. “If we don’t build these capabilities ourselves,” Bal said, India risks remaining dependent on systems it does not control.
Param2, unveiled on February 16 at the summit, is a Mixture of Experts model built from the ground up by BharatGen and trained extensively on Indian data, supporting all 22 scheduled Indian languages. BharatGen is led by IIT Bombay through the National Mission on Interdisciplinary Cyber-Physical Systems under the Department of Science and Technology, with a consortium that includes IIT Kanpur, IIT Hyderabad, IIT Madras, IIT Mandi, IIIT Hyderabad, and IIM Indore.
A Hard Trade-off, by the CEO’s Own Admission
Bal did not present the case as simple. He acknowledged that inexpensive foreign AI tools deliver real, near-term economic benefits, including productivity gains that function as a direct contributor to GDP growth. “It’s a hard conversation, because in the short term, the productivity gains that you’re going to get are a GDP growth vector,” Bal said. “You have to manage that short term … and long term where we have to make sure this price dumping needs to stop.”
That tension, between the immediate gains of using whatever AI tools are cheapest and most capable today, and the long-term risk of having no domestic alternative if access terms change, sits at the center of India’s broader AI policy debate in 2026. A Bernstein report cited by Moneycontrol made a similar argument independent of BharatGen, concluding that a domestic foundational model has become a strategic necessity if India wants to retain control over the next generation of digital infrastructure rather than depending on systems built and governed by the US or China.
What India Has Actually Built So Far
BharatGen’s Param2 is one of several outputs from the IndiaAI Mission, the government’s central vehicle for sovereign AI development, approved by the Cabinet in March 2024 with an outlay of roughly 10,372 crore rupees, or about 1.25 billion dollars. The mission selected 12 organizations and consortia in April 2025 to build indigenous foundational models, distributing subsidized GPU access rather than backing a single national champion.
The most visible result besides BharatGen is Sarvam AI, a Bengaluru-based startup that released two open-source models in February 2026: a 30-billion-parameter model and a 105-billion-parameter Mixture of Experts model with a 128,000-token context window. Unlike earlier models built under the mission, both Sarvam models were trained from first principles, using IndiaAI Mission-provisioned compute rather than fine-tuning existing foreign base models. Sarvam has since raised 234 million dollars at a 1.5 billion dollar valuation, with HCLTech taking a 10.46 percent stake, and the government is reportedly set to convert compute subsidies into a small equity stake of its own through compulsorily convertible debentures.
At the same summit, IT minister Ashwini Vaishnaw announced 20,000 additional GPUs for the national compute pool, bringing the IndiaAI Mission’s common compute facility to more than 38,000 GPUs. Gnani.ai also launched a voice-cloning system supporting 12 Indian languages from as little as 10 seconds of audio input, adding to a growing, if still early-stage, base of indigenous AI infrastructure.
The Gap Between Sovereignty and Self-Sufficiency
Independent assessments of the mission complicate the sovereignty narrative BharatGen and the government have promoted. Sarvam AI itself counts Nvidia among its investors, meaning the country is funding both compute infrastructure and model development while the underlying training hardware remains entirely foreign. A report from The Ken found that more than three-quarters of the GPUs deployed through IndiaAI-linked infrastructure providers were sitting underutilized in recent quarters, even as the startups the capacity was built for continued opting for Microsoft Azure, Amazon Web Services, or Google Cloud, often because venture funding came bundled with credits on those platforms.
Earlier models built under the mission also fell short of the “built from scratch” framing now associated with Param2 and Sarvam’s newest releases. Analysis from Times of Israel blogger Yashwant Singh noted that Sarvam’s earlier model and a 14-billion-parameter model from Fractal Analytics were fine-tuned on top of foreign base models such as Mistral and DeepSeek rather than trained independently, and that Indian developers still report limited access to the most advanced GPU clusters used by frontier labs elsewhere.
BharatGen’s own position reflects this gap. As of early 2026, the consortium has no polished, consumer-facing product comparable to ChatGPT or Gemini. Param2’s 17 billion parameters sit orders of magnitude below frontier models from OpenAI or Google, and BharatGen has instead positioned its models as public digital goods rather than direct competitors, designed for deployment by government departments, banks, hospitals, and courts, including in secure, air-gapped environments without internet connectivity.
Why the Government Is Betting on Public Infrastructure, Not a National Champion
BharatGen and the IndiaAI Mission’s broader strategy rests on a comparison to India’s earlier digital public goods, such as the UPI payments system and Aadhaar identity infrastructure. Rather than building one model to compete head-on with global frontier labs, the approach distributes funding and compute access across multiple consortia and startups, betting that a diversified domestic base, optimized for cost efficiency and Indian-language performance, can deliver strategic value even without matching the trillion-parameter scale of the leading US and Chinese labs.
Bal has framed this distinction explicitly, arguing in a separate comment around Param2’s launch that most real-world AI use cases do not require trillion-parameter models, an argument that doubles as a justification for the mission’s smaller, more targeted approach relative to the capital available to OpenAI, Anthropic, or Google.
Whether that strategy closes the gap BharatGen warns about depends on factors largely outside any single consortium’s control: whether India can secure durable, long-term access to advanced chips that survives shifts in US export policy, whether the underutilized GPU capacity already built finds paying demand, and whether sovereign models can move from research benchmarks to the kind of consumer adoption that would actually reduce reliance on foreign platforms. For now, the warning from BharatGen’s own chief executive is also an admission: the foundation exists, but the case for self-reliance is still being made against the more immediate appeal of cheaper, more capable foreign tools.
Frequently Asked Questions
What is BharatGen, and who runs it? BharatGen is India’s government-backed sovereign AI initiative, led by IIT Bombay through the National Mission on Interdisciplinary Cyber-Physical Systems under the Department of Science and Technology. Its consortium includes several IITs, IIIT Hyderabad, and IIM Indore, and it develops foundational AI models trained on Indian languages and data, released as public digital goods rather than commercial products.
What is Param2, and why does it matter? Param2 is a 17-billion-parameter Mixture of Experts multilingual model launched by BharatGen at the India AI Impact Summit 2026. It supports all 22 scheduled Indian languages and is designed for deployment by government departments, banks, hospitals, and courts, including in secure environments without internet access.
What does “price dumping” mean in BharatGen’s argument? BharatGen CEO Rishi Bal used the term to describe how global AI companies offer services in India at prices low enough to discourage investment in domestic alternatives, even while acknowledging this delivers short-term productivity gains. His concern is that this dynamic could leave India permanently dependent on foreign-controlled AI systems.
How much has India invested in sovereign AI, and what has it produced? The IndiaAI Mission was approved with an outlay of roughly 1.25 billion dollars in March 2024 and has since expanded its GPU compute pool to more than 38,000 units. It has funded foundational models from BharatGen, Sarvam AI, Gnani.ai, and several other organizations, though most of the underlying compute hardware remains foreign, primarily from Nvidia.
Are India’s sovereign AI models actually competitive with global models like GPT-5 or Gemini? Not at the same scale. Param2’s 17 billion parameters and Sarvam’s largest model at 105 billion parameters are both significantly smaller than leading frontier models. On Indian-language benchmarks, models like Sarvam 105B have shown strong, sometimes leading performance, but neither BharatGen nor Sarvam currently offers a consumer product positioned to directly compete with ChatGPT or Gemini at a general level.
Sources
- Business Today: ‘Price dumping of AI into India’: BharatGen CEO Rishi Bal warns cheap foreign models threaten local ecosystem
- PRESS Insider: Why India needs to build its own foundational AI model
- Indian Defence News: India’s BharatGen Ushers In Sovereign AI Era With PARAM-2 Launch
- The Hans India: India bets big on public digital infrastructure model for AI with BharatGen
- Medium (Shorya Bisht): BharatGen: India’s AI Moonshot in a World Already Owned by Giants
- The Ken: India has two sovereign-AI programmes. One is official, the other isn’t
- TechTimes: Sarvam AI Hits $1.5 Billion Valuation as HCLTech Bets $150 Million on India Sovereign AI
- Inc42: Sarvam And The Sovereign AI Dream
- The Times of Israel (Blogs): India’s Innovation Gap and the Rise of Sovereign AI
- explainx.ai: India Sovereign AI Status 2026: IndiaAI Mission, Sarvam Models, Gaps & Geopolitics
Stay in the Loop
For more stories, breakdowns, and unfiltered takes on what is really happening in Indian and global business and tech, follow TheFounder Nation.
Instagram: thefoundernation
We cover what the mainstream business press won’t.
© TheFounder Nation | All rights reserved Word count: ~1330 | Read time: ~7 minutes Primary keyword: BharatGen foundational AI India | Secondary: Param2 BharatGen, IndiaAI Mission, Sarvam AI, sovereign AI India, AI price dumping, India foundational model, Rishi Bal BharatGen Meta description: BharatGen CEO Rishi Bal warns India risks becoming a permanent AI consumer without sovereign foundational models like Param2. WordPress tags: BharatGen, IndiaAI Mission, Sarvam AI, sovereign AI, Param2, foundational AI models, India AI policy, AI infrastructure WordPress category: AI News
