/images/ppid_59c68470-image-178551769567135620.webp)
What is the story about?
Indian artificial intelligence startup Sarvam AI has launched Indus, a full-stack sovereign AI platform built around a custom 100-billion-parameter model developed entirely in India, as the company seeks to position itself as a domestic alternative to global AI providers for enterprise customers.
The launch marks Sarvam AI's expansion from developing foundation models to offering an end-to-end AI platform spanning infrastructure, models and applications. The company is targeting businesses that want to deploy AI while keeping sensitive data within India, amid growing focus on data sovereignty and enterprise adoption of generative AI.
Speaking to CNBC-TV18, Sarvam AI Co-Founder Pratyush Kumar said the company is looking to address a gap in the market by providing an integrated AI stack from a single Indian provider.
"For the first time we have an Indian player which can provide a full stack that's running in India," Kumar said, adding that sectors such as banking increasingly want customer data processed within the country rather than on overseas servers.
The launch comes as enterprises are increasingly moving beyond experimenting with generative AI and looking for production-ready platforms that combine computing infrastructure, AI models and deployment support. Today, many businesses rely on multiple technology vendors for different parts of their AI stack.
Enterprise focus
Sarvam AI said Indus is aimed at a broad customer base, including startups, small and medium enterprises and large organisations.
According to Kumar, the platform enables businesses to build AI agents for research, data analysis and presentation creation, deploy AI-powered voice applications for customer service and create multilingual content.
"People knew us as building models and putting out research components. What we are doing with Indus now is to provide full-stack applications, which people can relate to," he said.
He added that while smaller companies can adopt the platform directly, larger enterprises typically require deeper integration with existing systems and clearer return on investment. Sarvam AI has built a specialised deployment team to support enterprise implementations.
Capital remains critical
Kumar said the company will continue to raise capital as demand for AI computing infrastructure grows.
"Of course, we will continue to raise. Capital is one of those key ingredients because you need these AI GPUs, and they are not cheap," he said.
He said the company remains focused on ensuring investments in AI infrastructure translate into measurable returns for customers, enabling further expansion of its computing capabilities.
Government backing and sovereign AI
Kumar declined to comment on reports that the Government of India may invest in Sarvam AI, saying he would do so only after official confirmation.
He said, however, that the IndiaAI Mission had supported the development of the company's first foundation model and described the government as an important stakeholder in India's sovereign AI ambitions.
"We are very much an India-first player," Kumar said, adding that Sarvam AI is working with several state governments to deploy AI-based solutions.
He argued that India needs domestic AI capabilities as changing geopolitical dynamics and export controls could affect access to foreign AI technologies over time.
Kaze rollout delayed
Kumar also said the rollout of Sarvam AI's Kaze smart glasses has been delayed by global supply chain issues.
He clarified that Kaze is being developed primarily as an enterprise product rather than a consumer device, with manufacturing underway in Chennai. The company is also working with the National Association for the Blind and government agencies on the product.
Long-term strategy
Looking ahead, Kumar said Sarvam AI intends to build capabilities across AI infrastructure, foundation models and enterprise applications as businesses increasingly adopt AI at scale.
"The bottom line is very clear. We need to be owners of the majority of the AI tokens that are used in the country," he said.
He added that enterprises are increasingly exploring customised AI models alongside open-source alternatives, creating demand for companies that can offer technology across the entire AI stack.
The launch marks Sarvam AI's expansion from developing foundation models to offering an end-to-end AI platform spanning infrastructure, models and applications. The company is targeting businesses that want to deploy AI while keeping sensitive data within India, amid growing focus on data sovereignty and enterprise adoption of generative AI.
Speaking to CNBC-TV18, Sarvam AI Co-Founder Pratyush Kumar said the company is looking to address a gap in the market by providing an integrated AI stack from a single Indian provider.
"For the first time we have an Indian player which can provide a full stack that's running in India," Kumar said, adding that sectors such as banking increasingly want customer data processed within the country rather than on overseas servers.
The launch comes as enterprises are increasingly moving beyond experimenting with generative AI and looking for production-ready platforms that combine computing infrastructure, AI models and deployment support. Today, many businesses rely on multiple technology vendors for different parts of their AI stack.
Enterprise focus
Sarvam AI said Indus is aimed at a broad customer base, including startups, small and medium enterprises and large organisations.
According to Kumar, the platform enables businesses to build AI agents for research, data analysis and presentation creation, deploy AI-powered voice applications for customer service and create multilingual content.
"People knew us as building models and putting out research components. What we are doing with Indus now is to provide full-stack applications, which people can relate to," he said.
He added that while smaller companies can adopt the platform directly, larger enterprises typically require deeper integration with existing systems and clearer return on investment. Sarvam AI has built a specialised deployment team to support enterprise implementations.
Capital remains critical
Kumar said the company will continue to raise capital as demand for AI computing infrastructure grows.
"Of course, we will continue to raise. Capital is one of those key ingredients because you need these AI GPUs, and they are not cheap," he said.
He said the company remains focused on ensuring investments in AI infrastructure translate into measurable returns for customers, enabling further expansion of its computing capabilities.
Government backing and sovereign AI
Kumar declined to comment on reports that the Government of India may invest in Sarvam AI, saying he would do so only after official confirmation.
He said, however, that the IndiaAI Mission had supported the development of the company's first foundation model and described the government as an important stakeholder in India's sovereign AI ambitions.
"We are very much an India-first player," Kumar said, adding that Sarvam AI is working with several state governments to deploy AI-based solutions.
He argued that India needs domestic AI capabilities as changing geopolitical dynamics and export controls could affect access to foreign AI technologies over time.
Kaze rollout delayed
Kumar also said the rollout of Sarvam AI's Kaze smart glasses has been delayed by global supply chain issues.
He clarified that Kaze is being developed primarily as an enterprise product rather than a consumer device, with manufacturing underway in Chennai. The company is also working with the National Association for the Blind and government agencies on the product.
Long-term strategy
Looking ahead, Kumar said Sarvam AI intends to build capabilities across AI infrastructure, foundation models and enterprise applications as businesses increasingly adopt AI at scale.
"The bottom line is very clear. We need to be owners of the majority of the AI tokens that are used in the country," he said.
He added that enterprises are increasingly exploring customised AI models alongside open-source alternatives, creating demand for companies that can offer technology across the entire AI stack.
