Sarvam, India's first AI unicorn, is attempting to build a foundational AI model with over a trillion parameters from scratch, potentially bringing India into competition with global AI powerhouses like OpenAI and Anthropic. The startup announced this ambitious goal at its Epoch 2026 conference on July 30, marking a significant moment for India's AI ambitions. But whether Sarvam can actually deliver a model that people choose to use, rather than one they're forced to use, remains the real test.

What Has Sarvam Actually Built So Far?

Sarvam isn't starting from zero. The startup has already developed several specialized AI models designed specifically for Indian use cases and languages. These include Saaras V4, a speech recognition model that works in both Indian languages and English, with particular strength in detecting overlapping conversations and multi-speaker interactions. The company also built Vision 2.0, designed to digitize Indian documents, which has already processed over 35 million documents and is being used by the Odisha government.

The startup's flagship offering is the Sarvam 105B AI model, which now works with AI agents. According to benchmarks shared by Sarvam, this model outperforms OpenAI's GPT-5.4 Mini and Google's Gemini 3.5 Flash on certain tasks, particularly voice call capabilities and instruction following. Additionally, Sarvam recently launched Bulbul V4, a text-to-speech model that can add emotions like laughter or emphasis to make audio sound more natural, especially for code-mixed Hinglish, which global models still struggle with.

Why Does India Need Its Own AI Models?

The case for sovereign AI models has become increasingly urgent. After the US temporarily banned Anthropic's Mythos 5 and Fable 5 models for all foreigners, companies using foreign-made AI models began worrying about supply chain and national security risks. Until now, Indian companies had limited options: they could use US-based models from OpenAI or Anthropic, or turn to Chinese alternatives like DeepSeek V4.

Sarvam's progress demonstrates that India can develop competitive AI infrastructure. As one cybersecurity expert noted, Sarvam's achievements prove that the necessary infrastructure exists in India, models can be built, and compute resources are available. This matters because it challenges the narrative that AI development is exclusively the domain of Western and Chinese companies.

What Are the Major Obstacles Sarvam Faces?

Building a model is fundamentally different from building a business. Sarvam faces two critical hurdles that could determine its long-term viability:

• User Adoption Challenge: Sarvam must convince companies to actually choose its models on merit, not out of patriotic duty or government mandate. The Indus by Sarvam app has over 100,000 downloads on Google Play, but this pales in comparison to OpenAI or Google's user bases. Most Indian tech professionals work in English-first environments, limiting Sarvam's addressable market.
• Revenue Model Problem: Sarvam has not yet proven a sustainable business model exists. No customers are currently paying a premium for Indian large language models (LLMs), and even if forced to switch, companies would likely negotiate prices down dramatically rather than pay full rates.
• Frontier Model Gap: While Sarvam excels in specialized areas like Indian language processing and document digitization, it does not yet have a frontier model that matches the cutting-edge capabilities of OpenAI's GPT-4 or Anthropic's Claude. The upcoming trillion-plus parameter model could help close this gap, but it remains unproven.

"Sovereign AI shouldn't quietly become second-best AI. The actual test is whether Indian companies choose these models on merit, when nobody's watching. That's the bar," said Gurleen Khurana, co-founder of AI consulting firm SimplifyGenAI.

Gurleen Khurana, Co-founder at SimplifyGenAI

One expert explained the market timing problem bluntly: Sarvam's achievements make the technical path clearer, but they don't create a market that doesn't exist. That requires either a government mandate forcing adoption or genuinely superior products that outperform global alternatives. Sarvam currently has neither.

Where Is Sarvam's Real Competitive Edge?

Sarvam's strength lies not in competing head-to-head with OpenAI on general-purpose models, but in dominating specific niches where global models fall short. For Indian language voice work, code-mixed Hinglish text-to-speech, and document digitization tailored to Indian formats, Sarvam has built defensible advantages. Companies working with Indian brands in creative spaces are actively evaluating Sarvam's models precisely because they handle these specialized tasks better than alternatives.

"If you're building Indian language applications, Sarvam becomes relevant, which I believe is their strength. Most techies in India work in English-first environments," explained Govind Rammurthy, CEO of cybersecurity firm eScan.

Govind Rammurthy, CEO at eScan

How Is Sarvam Positioning Itself for Growth?

Sarvam recently achieved a major milestone by becoming India's first AI unicorn, reaching a $1.5 billion valuation after raising $234 million in funding led by HCLTech. The startup also opened its first research lab in San Francisco, giving it direct access to top AI talent in Silicon Valley, including many Indian engineers who moved to the US for higher salaries and access to computing resources.

The company also recruited Devendra Singh Chaplot, a star AI expert who previously worked at xAI, Mistral, and Thinking Machine Labs, as a part-time advisor. This move signals Sarvam's ambition to accelerate its frontier model development by tapping into the global AI research community while maintaining its India-focused mission.

Steps to Understanding Sarvam's Path Forward

• Monitor Model Benchmarks: Watch for independent benchmarks comparing Sarvam's trillion-plus parameter model against OpenAI GPT-4 and Claude 3.5 on standard AI evaluation tests. This will reveal whether Sarvam can close the frontier model gap.
• Track Enterprise Adoption: Follow which Indian enterprises voluntarily adopt Sarvam models without government incentives. Organic adoption by companies like Flipkart, Infosys, or TCS would signal genuine competitive advantage rather than patriotic purchasing.
• Assess Revenue Growth: Look for quarterly revenue reports and customer testimonials showing whether Sarvam has built a sustainable business model with paying customers willing to commit long-term contracts.
• Evaluate Specialized Use Cases: Test Sarvam's models in Indian language processing, document digitization, and code-mixed speech recognition to understand where it genuinely outperforms global alternatives.

Sarvam's journey reflects a broader shift in AI development. Rather than trying to beat OpenAI at its own game, the startup is building models optimized for India's unique linguistic, cultural, and business needs. Whether this strategy proves sufficient to sustain a billion-dollar company depends on whether Indian enterprises and developers ultimately choose Sarvam because it's the best tool for the job, not because it's Indian.