Sarvam AI to Receive First IndiaAI Mission Grant Worth INR 220 Crore

Sarvam AI, a AI startup based in Bengaluru and founded by Vivek Raghavan and Pratyush Kumar, is progressing toward being the first beneficiary of the IndiaAI Mission. This mission is the flagship initiative from the Indian government to aid and boost the development of AI that is made and owned in India. As per several industry insiders, Sarvam AI is expected to receive support to the tune of about INR 220 crore. This support is expected to come mostly in the form of having nearly 4,000 top-tier Nvidia H100 GPUs available free of cost for a period of six months. These GPUs are among the best available for training AI models.

Indigenous Models for Indian Voices

Sarvam has already advanced in this area, having founded a language model with 2 billion parameters last year. This model, when put to the test in the context of Indian languages, was found to yield results that outperformed those of several globally recognized language models. Sarvam's model used tokenization techniques that were lauded for their efficiency and precision, enabling the generation of prompt regional language responses with far greater ease than had been the case up to that point.

Sarvam in the process of scaling up both his infrastructure and the model itself. The AI company is working on a new version of its language model that will come equipped with 70 billion parameters and will be capable of functioning in both Indian languages and English.

Government’s Grand AI Vision

The Mission IndiaAI, with support exceeding INR 10,000 crore, has been conceived to set in motion the domestic innovation ecosystem in artificial intelligence. Its primary focus, stated quite clearly, is on three components: on the one hand, it wants to create the very necessary IndiaAI compute capacity; on the other hand, it aims to establish an IndiaAI Innovation Centre; and, finally, the datasets required for AI in India, which are currently lacking, should be made available through the IndiaAI Datasets Platform. The agencies that set up this mission seem to have realized that AI requires large-scale computing facilities and have set about trying to create one.

Competition and Collaboration in the AI Arena

Alongside Sarvam, other startups like Soket AI Labs, Gnani.ai, and Gan.ai are also in the fray. Soket AI Labs has proposed a massive 120-billion parameter open-source Indic LLM under the EKA Project, while Gnani.ai is focused on conversational AI, aiming to develop speech-to-speech models for customer service. Gan.ai, known for AI-powered video personalization, has also submitted its proposal. The urgency to support such projects has been amplified by rapid advancements in countries like China, whose DeepSeek model spurred India’s resolve to build its own foundational AI technologies.

Currently, India is advancing large-scale language models, and it aims to have several of them ready by year-end. Hence, the timeline fits well with the government's ambition to have a more significant presence in the global artificial intelligence race.