Loading...
Building India's complete AI computing stack
Krutrim AI: The Startup Trying to Build India’s Own AI Stack
When the global AI race began to accelerate, India had a problem. It had millions of engineers, a huge internet population and a growing digital economy — but it did not yet have a major homegrown AI company building the technology stack from the ground up.
Bhavish Aggarwal wanted to change that.
In December 2023, the Ola founder introduced Krutrim, a Sanskrit word meaning “artificial”. The idea was bigger than simply creating another chatbot. Krutrim wanted to build an AI ecosystem designed around India — from language models to computing infrastructure.
“India has to build its own AI.”
That idea became the foundation of Krutrim’s journey.
Krutrim entered the market at a time when companies such as OpenAI, Google and Microsoft were rapidly advancing generative AI.
Instead of trying to compete only through a consumer chatbot, Krutrim focused on something that could give it a distinct identity: India’s linguistic and cultural diversity.
Its foundational model was trained on more than 2 trillion tokens, according to the company. Krutrim said the model was designed to work across Indian languages and could switch between languages during conversations.
The company eventually presented Krutrim as a family of models, including its base model and the more advanced Krutrim Pro, with plans for capabilities involving vision, speech and task execution.
The ambition was straightforward: make AI feel more natural for people who do not primarily communicate in English.
But Krutrim's bigger ambition was never limited to an LLM.
The company described its goal as building a complete AI computing stack. That meant working across several layers:
AI models for language and other applications
Cloud infrastructure for developers and businesses
Computing hardware and data-centre infrastructure
AI services and APIs for enterprises
This approach was important because training and running advanced AI models requires enormous amounts of computing power.
Krutrim therefore began developing its own cloud platform and working with different hardware providers to improve the performance and cost of AI workloads. Its engineering team has publicly discussed optimisation work involving Nvidia, Intel and AMD hardware.
The company was not simply trying to build an Indian chatbot. It was attempting to build the infrastructure underneath India's AI economy.
Krutrim's rise was remarkably fast.
In January 2024, the company announced a $50 million equity funding round led by Matrix Partners India, giving it a valuation of $1 billion.
That made Krutrim India's first AI unicorn, barely a month after its foundational model had been introduced.
For Aggarwal, it was also familiar territory. Ola and Ola Electric had already taken him through the process of building large technology businesses in India.
But AI was a different game.
The cost of computing, access to GPUs, model development and competition from global technology giants meant that building a serious AI company required far more than a good product.
In February 2024, Krutrim released its AI assistant in public beta.
The assistant was positioned as an Indian alternative to global conversational AI products and supported multiple Indian languages.
But the bigger lesson from Krutrim's early journey was that building an AI model is only one part of building an AI business.
The company gradually moved its attention toward infrastructure and enterprise customers.
That shift became much clearer in 2025 and 2026.
By 2026, Krutrim had repositioned itself as a domestic AI cloud services company.
The company reported approximately ₹300 crore in FY26 revenue, around three times its previous year's revenue, along with its first annual net profit and a reported profit-after-tax margin of more than 10%.
Krutrim also said that it had more than 25 large enterprise customers across sectors such as telecom, financial services and healthcare, while much of its GPU capacity was being used for external workloads.
At the same time, the company paused its chip-design initiatives and concentrated more resources on its cloud infrastructure.
This marked an important change in the company's story.
Krutrim had started by asking:
“Can India build its own AI model?”
It was now asking a broader question:
“Can Indian businesses build and run AI on infrastructure built for India?”
Today, Krutrim's story is less about creating one ChatGPT competitor and more about building layers of India's AI infrastructure.
Its cloud platform aims to provide computing resources for AI developers and enterprises, while its models remain part of the broader ecosystem.
The company is also connected to technologies such as Ola Maps, which uses AI, large-scale data and mapping infrastructure to build location services suited to Indian conditions.
That fits into the larger philosophy behind Krutrim: build technology locally, understand India's specific requirements and gradually reduce dependence on external infrastructure.
Krutrim's journey shows how quickly India's AI landscape has changed.
It began with a bold vision of creating an Indian foundation model, became a billion-dollar company almost immediately, experimented with a consumer AI assistant and then moved deeper into the less glamorous but commercially important world of cloud infrastructure.
The next chapter is still being written.
But the idea behind Krutrim remains clear: India should not only use the AI revolution — it should also build some of the infrastructure that powers it.
And that is what makes Krutrim's story bigger than a chatbot.
It is a story about an Indian startup trying to move one layer deeper into technology — from using AI to building the systems on which AI itself can run.
