India’s Sovereign AI Reality Check

India was quick to respond to Sam Altman when he underestimated the country’s ability to build something even remotely comparable to ChatGPT during a visit in early 2023. 

But within just seven months, India had an answer of its own. Krutrim, the country’s first AI unicorn, launched a family of large language models (LLMs).

Following this, LLMs by Sarvam AI, BharatGen, CoRover.ai, Gnani.ai, Soket AI became the focal point for sovereign AI. Indian founders had started work on their own arsenal in response to the growing dominance of OpenAI, Google Gemini, Anthropic’s Claude, and xAI’s Grok.

Indeed, the India AI Impact Summit in Delhi this year was a culmination of this wave of sovereign AI that emerged in the aftermath of Altman’s contentious remarks.   

The grit was undeniable, and the shared ambition was clear — to build something powerful and deeply rooted in the Indian context, something global players would struggle to replicate with the same authenticity and nuance as Indian founders.

But the race to make a truly Indian LLM seems to have lost some of its steam. Just take the case of Krutrim, which launched with such fanfare but has now petered out after its efforts to build an LLM did not pan out. 

So what exactly went wrong, and can Indian AI companies disrupt the global AI industry in a major way?

Missing From The Global AI Conversation

Despite pouring hundreds of millions of dollars into sovereign AI, India is still waiting for its DeepSeek moment. When it comes to LLMs, Indian models rarely figure in the global conversation, and even those models which were launched during the India AI Impact Summit have failed to plug this gap. 

Well, for one, in the bid to build India’s own ChatGPT, the focus of many of these companies increasingly shifted towards Indian languages, and one can argue that they overlooked the challenge of building an Indian AI model that can work in the global context. 

Thus far, only DeepSeek, which is of Chinese origin, has managed to break into the conversation and be spoken of in the same breath as OpenAI, Claude, Gemini, xAI or Meta. 

Much of the ecosystem became consumed with proving that Indian models could speak Indian languages. This was indeed a strong differentiator in the beginning, but it was not enough to move the global market. 

According to Ed Huang, the cofounder and CTO of TiDB, an AI-native open-source database company, India can nurture the next big consumer AI company, but it seems its LLMs are only part of the country’s narrative rather than a moat in the global arena.

He added that the core focus on India gives relevance to the likes of BharatGen or Sarvam AI. But it also restricts their global competitiveness. 

Having said that, he believes that India’s multilingual experience is relevant globally. Therefore, it makes more sense for us to cater to the rest of the world, which will bring in new revenue streams. “I think Indian [AI] startups should position multilingual AI not simply as “supporting Indian languages,” but as infrastructure for large-scale human-AI interaction across regions and cultures.”

Where’s The Moat?

Next comes the question of why Indian AI labs have not yet produced a general model that is competitive enough to enter the global conversation.

According to Chaitanya Choudhary, the founder of Workers IO, a US-based AI startup, Chinese labs are currently leading much of the open-source model ecosystem, which makes it even harder for Indian models to stand out internationally.

He added that new Indian models are not bad but only competitive with last year’s models in terms of benchmarks. 

Touting Opus 4.6, GPT 5.5, Qwen 3.5, and Minimax 2.6 as some of the best models to pick for AI founders presently, Shayak Mazumder, cofounder of Adya.ai, said that there is no particular motivation for founders like him to use Indian LLMs unless they are truly locking horns with their global counterparts.

“Now, there has to be a compelling reason for a model built in India to be adopted. Government deployments is just one use case, as preferences naturally tilt towards models developed within India,” he added.

Even commercially, there are challenges. Some of the leading Indian frontier AI labs are still operating at ARR levels in the range of $10 Mn to $12 Mn, far below the scale achieved by global players. In several cases, these numbers are only marginally ahead of Indian AI application startups that are not building foundational models themselves.

According to Artificial Analysis benchmark data, global frontier AI models from OpenAI, Anthropic, Google and Alibaba are currently clustering around intelligence scores of 57-60 with context windows nearing 1M tokens, significantly ahead of the current benchmarked scores of some of India’s top names.

On a similar note, Sunil Maurya, founder of one of India’s largest Voice AI communities, Unio, said that users naturally gravitate towards the smartest and most capable models available from OpenAI or Anthropic. On the enterprise side, while privacy and on-premise deployments remain important, companies can already deploy highly capable open-weight models like Google’s Gemma series without depending on custom Indian LLMs.

Beyond The Nationalism

The challenge is no longer proving that India can build foundational AI models. The bigger concern now is how to make Indian models globally relevant. It might seem strange for some that the pressure has risen so drastically, especially because sovereign AI as a concept is still new for much of India. 

But the AI race moves at a radically high pace. The focus areas need to shift rapidly, and what was the priority one year ago is almost taken for granted now. Indian LLM makers need to step up their game and get ready to compete with AI giants from the US and China.  

Even as the likes of OpenAI and others have moved onto multimodal AI and agentic models, Indian companies are yet to break through despite having the full backing of policymakers and raking in funding in recent months.

Undoubtedly, players like Sarvam AI are doing everything they can to make a global impact, but even domestic adoption is limited to developers and researchers, not large enterprises. On the consumer AI side, the high-profile launch of Sarvam’s chatbot app and partnerships with smartphone OEMs have given the company something of a lead in this space, but is that enough in the face of the might of OpenAI, Google Gemini and Anthropic, which seem to release platform updates every few days?

Indian chatbots like Indus face flak on social media for not being prompt in addressing bugs and latency. Whether the likes of Sarvam can prove the naysayers wrong will likely depend on three things — intelligence parity, distribution and differentiated use cases. 

China’s DeepSeek did not enter global conversations merely because it was Chinese. It entered because developers found it competitive, open, cheap and deployable at scale. 

Indian labs need a similar inflexion point, which is hardly in sight. India’s state of AI affairs gets visibly clear when one looks at the developer and user ecosystem. To say the least, Indian LLMs are still figuring out a way to engage developers, which seems like the lowest-hanging fruit. 

Despite these grave issues, India’s multilingual advantage and enterprise IT services foundation are strategic strengths that can give Sarvam and others a leg up over the international competition. But these are potentials that have yet to be fulfilled. 

Unless Indian LLMs consistently match frontier models on reasoning, speed, developer adoption, and ecosystem integrations, they risk becoming extras in the global AI race.

Inc42’s AI Summit: India’s Largest Invite-Only AI Gathering

Inc42’s AI Summit 2026 is designed as a gathering point for the people building India’s AI ecosystem beyond the hype cycle. The idea is to move the conversation away from flashy demos and focus instead on what is actually working in production across India’s unique realities, from multilingual users and low-cost economics to patchy infrastructure and billion-user scale.

The summit will bring together more than 600 attendees, including AI founders, enterprise leaders, GCC executives, operators, policymakers, investors, and product builders. 

Across two stages and 25+ sessions, conversations will span AI-native product development, agentic AI systems, ecommerce transformation, infrastructure, enterprise adoption, and the future of AI-led business models. The event will also feature hands-on workshops and practical playbooks aimed at helping companies move from experimentation to execution.

The AI Summit will host keynotes, fireside chats, and panel discussions featuring leaders such as V Vaidyanathan, MD and CEO, IDFC First Bank; Rahul Chari, cofounder and CTO, PhonePe; Khilan Haria, CPO, Razorpay; Mohit Saxena, cofounder and CTO, InMobi; Rishikesh SR, cofounder, Rapido; Anish Achuthan, cofounder and CEO, Open Financial Technologies; Ankush Sabharwal, founder and CEO, CoRover.ai; Vaibhav Khandelwal, cofounder and CTO, Shadowfax; Karthik Rajaram, general manager and country head India, ElevenLabs; Rohan Nayak, cofounder and CEO, Pocket FM, among others. 

More importantly, Inc42 hopes the summit becomes a space where India’s AI community can openly exchange lessons, failures, and frameworks that are relevant to building in India, rather than simply replicating Silicon Valley narratives. With founders, CTOs, operators, and investors all in one room, the goal is to create deeper collaboration across the ecosystem while spotlighting the next wave of Indian AI startups and builders.

Top Stories From India & Around The World

• Neysa’s $1.2 Bn AI Bet Clears Hurdle: The CCI has approved the Blackstone-led acquisition of a controlling stake in AI cloud startup Neysa Networks, formally clearing one of India’s largest AI infrastructure funding rounds. Neysa plans to deploy over 20,000 GPUs across India as part of its sovereign AI cloud expansion.
• AI Surveillance Inside Homes? Pronto faced backlash after reports revealed it was piloting opt-in video recording inside customer homes to generate training data for robotics and physical AI systems. The controversy has now triggered wider debates around consent, surveillance and whether India’s labour economy is quietly becoming AI training infrastructure.
• Mythik Raises Fresh Capital: Jason Kothari-led Mythik has raised an additional $5 Mn, taking its valuation past $50 Mn. The startup uses AI to recreate Indian mythology and folklore into short-form video content for global audiences.
• Anthropic Eyes $900 Bn Valuation: The AI startup is reportedly closing a new funding round that could value it above $900 Bn, potentially making it the world’s most valuable private company. Investors including Sequoia, Dragoneer, Altimeter and Greenoaks are expected to co-lead the financing.
• Vibe Code Android Apps: Google has launched the ability to build native Android apps directly inside AI Studio for free. Since launch last week, users have already created more than 2.5 Lakh Android apps, with Google claiming that over 99% of creators had never built an Android app before.

Startup In The Spotlight: Angoor AI

As businesses increasingly manage customer interactions across fragmented channels such as WhatsApp, calls, Instagram and emails, a new category of AI-native platforms is emerging around unified, autonomous customer engagement. Bengaluru-based Angoor AI is positioning itself within this shift by transforming traditional CRMs from passive data repositories into active, AI-led customer interaction systems in the global CRM market, projected to become a $321 Bn opportunity by 2034.

Founded in 2023 by IIT Bombay graduates Arpit Agrawal, Anuj Agrawal and Rishabh Kumar, the startup was born out of the observation that most businesses handled communication channels in silos, without building a unified layer of customer intelligence underneath. Instead of simply storing customer information, Angoor AI aims to continuously generate and act on customer intelligence through every interaction.

Today, the startup offers an agentic AI-native platform that helps B2C and D2C brands automate customer engagement across voice calls, WhatsApp, Instagram, websites, email and messaging platforms. Businesses can build AI-orchestrated workflows with human-in-the-loop interventions, enabling automation across customer support, marketing, sales and engagement functions within a single system.

Under the hood, Angoor AI is building what it calls an “active CRM” layer, where AI agents not only manage workflows but also continuously capture behavioural insights, conversation context and engagement signals across channels. The platform is designed to help businesses move beyond static customer databases toward adaptive systems capable of responding, learning and optimising interactions in real time.

The startup currently serves clients across India and the UK, while also expanding its voice AI capabilities as part of its broader product roadmap. Going forward, Angoor AI plans to deepen its AI-led communication stack and enter the US market by late-2026, positioning itself at the intersection of CRM software, conversational AI and enterprise automation.

[Edited by Shishir Parasher & Nikhil Subramaniam]
[Creatives by Varshita Srivastava & Abhyam Gusai]

The post India’s Sovereign AI Reality Check appeared first on Inc42 Media.