Sarvam AI eyes trillion-parameter model with $234 million Series-B, U.S. research lab
Bengaluru: Bengaluru-based Sarvam AI has announced plans to build a one-trillion parameter foundational model, a move that puts the Indian startup in direct competition with the world’s largest AI labs.
The company said it has opened a San Francisco office and a Bay Area research lab to support the project. It has also raised $234 million in the first close of its Series-B funding round at a valuation of $1.5 billion.
Foundational models are general-purpose AI systems capable of performing new tasks without explicit programming. Their capabilities are shaped by the datasets they are trained on. These massive neural networks identify patterns across vast amounts of data to create a “foundation” that can be applied to tasks like coding, summarisation, and image generation.
Sarvam AI’s trillion-parameter ambition carries clear geopolitical implications, as nations race to control large-scale AI infrastructure. But the company acknowledged the scale of the challenge, with training, compute, and data requirements for such models running into billions of dollars.
The funding and U.S. expansion signal Sarvam’s intent to build and deploy large models both for India and global markets, at a time when open and proprietary AI models are reshaping technology and policy worldwide.
