The AI industry is entering its infrastructure phase. As models scale from millions to trillions of parameters, the constraint is no longer just talent or ideas, it's compute. Access to GPUs, tooling, and technical support is becoming as decisive for a startup's trajectory as access to capital.

India is increasingly central to this shift, with one of the world's largest developer bases, a fast-growing AI startup ecosystem, and a national push toward building sovereign AI infrastructure. But turning that potential into products requires the right partnerships between capital, compute, and technical expertise.

We recently had the privilege of hosting Unnikrishnan A R for a Zero-Hour Conversation. He leads NVIDIA's developer and startup ecosystem efforts across South Asia, and has been closely involved in scaling the Inception program, which today supports over 5,000 startups globally by putting real resources behind founders.

𝗪𝗵𝗮𝘁 𝘀𝘁𝗮𝘆𝗲𝗱 𝘄𝗶𝘁𝗵 𝘂𝘀:

𝗧𝗵𝗲 𝗲𝗰𝗼𝗻𝗼𝗺𝗶𝗰𝘀 𝗮𝗿𝗲 𝗺𝗼𝘃𝗶𝗻𝗴 𝗳𝗮𝘀𝘁𝗲𝗿 𝘁𝗵𝗮𝗻 𝗺𝗼𝘀𝘁 𝗿𝗼𝗮𝗱𝗺𝗮𝗽𝘀 - new architecture generations are landing annually now, each cutting inference cost roughly 10x, which changes the math on when and how much compute to commit to.

𝗗𝗮𝘁𝗮 𝗶𝘀 𝘁𝗵𝗲 𝗿𝗲𝗮𝗹 𝗺𝗼𝗮𝘁, 𝗻𝗼𝘁 𝘁𝗵𝗲 𝗺𝗼𝗱𝗲𝗹 - in domains like healthcare and voice, India's scale and diversity of data is a genuine structural advantage, but it only compounds if it's put to work early.

𝗢𝗽𝗲𝗻 𝗺𝗼𝗱𝗲𝗹𝘀 𝗮𝗿𝗲 𝗰𝗹𝗼𝘀𝗶𝗻𝗴 𝘁𝗵𝗲 𝗴𝗮𝗽 𝘄𝗶𝘁𝗵 𝗳𝗿𝗼𝗻𝘁𝗶𝗲𝗿 𝗺𝗼𝗱𝗲𝗹𝘀 - with full data, checkpoints, and training recipes now being shared, startups can build deep, differentiated products without depending on a single closed provider.

𝗖𝗮𝗽𝗮𝗰𝗶𝘁𝘆, 𝗻𝗼𝘁 𝗮𝗺𝗯𝗶𝘁𝗶𝗼𝗻, 𝗶𝘀 𝗜𝗻𝗱𝗶𝗮'𝘀 𝗿𝗲𝗮𝗹 𝗰𝗼𝗻𝘀𝘁𝗿𝗮𝗶𝗻𝘁 - compute provisioning now takes months, so the founders who commit early to infrastructure are the ones who compound the advantage.

#LifeAtKalaari is all about continuous learning, and this session reflected that spirit.

‍