Inside India's First AI Unicorn: Krutrim's Master Plan
“India cannot rely on western models. We need our own foundational intelligence and compute infrastructure.”
Executive Biography
Bhavish Aggarwal is an Indian tech entrepreneur who founded Ola Cabs, Ola Electric, and Krutrim AI. An alumnus of IIT Bombay, he has spearheaded indigenous industrial manufacturing and sovereign AI infrastructure.
Began his career as a researcher at Microsoft Research India before founding Ola Cabs in 2010, taking on global rideshare giants before pioneering gigafactory battery cell development and generative AI.
Key Strategic Rules
- ›Sovereignty requires vertical integration: controlling hardware, compute stack, and indigenous foundational weights.
- ›India's linguistic diversity requires language models trained directly on regional corpora rather than fine-tuning English bases.
- ›Energy density and low-cost electricity will determine the winning AI data centers of the 2030s.
Founding Milestones
2010
Founded Ola Cabs
Revolutionized urban mobility in India from a small apartment in Mumbai.
2021
Ola Futurefactory
Constructed the world's largest two-wheeler electric factory in Krishnagiri, Tamil Nadu.
2024
Krutrim Unicorn
Krutrim became India's fastest unicorn, releasing Indic LLM APIs and sovereign cloud infrastructure.
The Interview Transcript
Startup Journal: Why is building domestic foundation models a national imperative rather than just a commercial enterprise?
Bhavish Aggarwal:If 1.4 billion people interact with digital knowledge entirely through lenses encoded with foreign worldviews and cultural nuances, we concede intellectual sovereignty. India produces huge datasets; our models must speak our cultural vernacular.
Startup Journal: How do you manage the immense compute cost differences between India and Silicon Valley?
Bhavish Aggarwal:By optimizing model architectures specifically for inference efficiency on commodity enterprise chips and locating our high-density data centers near renewable solar installations in South India.