India needs AI factories, to target accelerated computing: Nvidia MD

Chip-making giant Nvidia, which recently surpassed Apple to become the highest-valued company in the world, has called for the setting up of AI factories in the country to take the next leap in AI revolution.

 

Batting for an AI factory in India, Vishal Dhupar, managing director of Nvidia in South Asia, said the need is to establish manufacturing intelligence in the country, consume it domestically, and export the surplus.

 

“The number one opportunity for India and for Tamil Nadu is to set up manufacturing plants for intelligence—factories for AI. I am not calling them data centres,” Dhupar said while addressing Tamil Nadu’s flagship IT meet, Umagine 2025, on Friday.

 

AI factories are ecosystems or hubs that boost innovation, collaboration, and development in the field of AI by bringing together computing power, data, and talent to create cutting-edge generative AI models.

 

He added that just as India leapfrogged to mobile technology, the country should also shift faster to accelerated computing rather than general-purpose computing.

 

“India has no legacy. In telecom, how we leapfrogged to mobile—just like that, we should leap to accelerated computing, not general-purpose computing. It is sustainable computing. Accelerated computing means augmenting the graphics processing units (GPUs) and CPUs,” Dhupar added.

 

 Accelerated computing powers applications ranging from gaming consoles and self-driving cars to the world's most powerful supercomputers. Interestingly, the California-based company has a first-mover advantage and three decades of expertise in GPUs, giving it a market monopoly over rivals.

 

The company holds a 90 per cent market share in the AI chip market, compared to competitors like Advanced Micro Devices (AMD) and Intel, which have much smaller shares.

 

During the third quarter that ended on October 27, Nvidia posted record revenue of $35.1 billion, up 17 per cent from the second quarter and up 94 per cent from a year ago. It also posted a record quarterly data centre revenue of $30.8 billion, up 17 per cent from Q2 and up 112 per cent from a year ago.

 

Dhupar noted that the Indian government has already taken steps to make a significant leap in the AI sector, starting with a strategy paper in 2018 and followed by the declaration of the AI Mission.

 

“The government has created conditions to bring all the infrastructure together. The first step forward is extremely good. TCS and others are working on how we integrate workflows. This is just the beginning, and a lot more needs to be done,” he said.

 

On the sustainability focus in electricity consumption for data centres, he said the future will see data centres with a capacity of even 250 kW, and India has a big opportunity in hosting such facilities.

 

This is because it requires less populated areas with ample power opportunities and water availability. At present, 40 kW to 125 kW is considered to be high-density data centres, while globally, there are extreme-density centres that consume up to 200 kW.