India's Future AI Game Plan: From Local Know-How to Global Innovation If India can break the cost models in AI computing, these innovations can be exported globally positioning the country as a leader in the AI space, said Anil Nanduri

The Indian government is laying the groundwork for India's ambitious goal in Artificial Intelligence—ensuring infrastructure development aligns with national priorities of innovation, self-reliance, and inclusive growth to become an intelligence capital. The young talent pool of the nation is the cherry on the cake. Earlier in March, India AI Mission 2.0 was approved by the Indian government to make AI in India and make AI work for India.

Innovation in Infrastructure

While drawing light on 'government strategy on compute in India' Dr. Thomas Zacharia, senior vice president,(AMD) emphasized overcoming power constraints to achieve exascalecomputing, and urged India to influence global AI infrastructure through strategic investments. He also highlighted the importance of scaling AI applications beyond data centers to include edge computing, envisioning transformative effects on sectors from healthcare to industrial optimization.

"My 30-year experience says India's strength is data and the large pool of talent, and this is tailoring the infrastructure. There will be an insatiable appetite for compute cycles in India. India should adapt algorithms to use lower precision math formats for significant improvements in energy efficiency," said Zacharia.

Investment in R&D and Learn Local KnowHow

While discussing what India aims to achieve Vishal Dhupar MD, Asia South, NVIDIA emphasized the crucial need to understand Indian vernacular languages and cultures to codify this knowledge to foster unique AI applications and innovations to position India as the "intelligence capital".

"Our common sense is very different, our sensibilities are very different. India needs to be the intelligence capital, just like Paris is the fashion capital. If we codify our know-how, the algorithm is known, compute you know how to source," said Dhupar.

"Our mission is from India for India and for the world as well," added Sunil Gupta Co-Founder, MD & CEO of Yotta Data Services while talking about investment Scale.

Breaking Barriers for Software Developers

While speaking on where the government should strategically invest in R&D capital over the next few years, Anil NanduriVice President, Head of AI Acceleration Office, Intel said that the talent in India is exceptional, but to fully leverage this potential, it is crucial to democratize access to computing for AI development. Making these resources broadly available will enable developers to innovate more freely and efficiently.

"Determining the desired outcomes is essential for strategic investments. If India can break the cost models in AI computing, these innovations can be exported globally, positioning the country as a leader in the AI space," said Nanduri.

Cost of GPUs

While shedding light on how India should approach building its computer infrastructure Sambit Sahu, Senior Vice President, Krutrim-Ola said the cost of GPUs is a critical factor. Designing and building GPUs at a lower cost can dramatically increase the number of GPUs affordable for AI development. India has a unique opportunity here, given its substantial talent pool—20 per cent of the global semiconductor task force is based in India. India can develop GPUs without the high margins charged by Western companies.

"To understand the scope of AI training, consider that training a trillion-parameter model like OpenAI's GPT-4 requires around 25,000 GPUs for three months. This illustrates the immense computational power needed for such advanced AI models," said Sahu.

"India must address the urgent need for AI infrastructure. With a minimum requirement of 10,000 GPUs, the focus should be on scaling up to meet the needs of the entire country, not just individual startups," added Gaurav Aggarwal, Vice President, AI, Reliance Jio.

Monitoring and Utilization

While sharing the government perspective on compute capacity, Col. A.K Nath director general CDAC underscores the importance of robust monitoring frameworks to optimize compute utilization effectively. "We have devised ways to monitor and advise users," he emphasized, highlighting efforts to ensure efficient usage of available resources like India's Param Siddhi AI, ranked among the world's top supercomputers. This monitoring not only enhances efficiency but also guidesfuture procurement decisions.