India's largest cluster of graphics processing units (GPU) was about 2,000- 2,500 as compared with 20,000 plus in the US, said Shridhar Garge, Head, Strategy, Planning & Programs, NVIDIA, illustrating the relevant AI infrastructure gap that impedes growth potential for the biggest breakthrough technology in the world.
The graphics processor and chipset major is the world’s sixth-most valuable company by market capitalization, at nearly $1.2 trillion, and the NVIDIA stock has surged nearly four times in a year as discussions around AI grab more column inches and dominate the air waves. The US chip giant's products play a crucial role in AI systems.
Garge also said that for clusters that have about 1000 GPUs, India would have around four or five.
"There are certain things that are limiting the thought process of Indian startup founders - infrastructure being a very important one," Garge said at the India Internet Day 2023 in New Delhi last week.
A GPU cluster is a group of computers that have a GPU on every node. Multiple GPUs provide accelerated computing power for specific computational tasks, such as image and video processing and training neural networks and other machine learning algorithms.
Similarly, Rajan Anandan, MD, PeakXV Partners and Surge, said that for application-level companies like those involved in healthcare, education and customer support among others, not having this infrastructure in place is a real constraint. Thus, he said it boils down to how much capital the startups can raise but also the access that they can get.
"One thing the government can do is to build a very large-scale (GPU cluster)," Anandan said. "Why shouldn't India build a 100,000 GPU cluster and then we give it to Indian developers, Indian research institutions, Indian startups to build on it and make it very affordable?"
Costs & volumes
He said this would be playing on our strengths of doing population scale projects at ultra-low cost and high volumes.