India's need for high-performance computing rises as AI use expands

Government and private companies are placing orders for high-performance machines

The earth science ministry in June awarded a $100 million contract to Eviden, a division of Paris-based Atos, to build two supercomputers for weather modelling and climate research by the Indian Institute for Tropical Meteorology (IITM) and the National Centre for Medium-Range Weather Forecasting (NCMRWF).

The supercomputers will have a combined power capacity of up to 21.3 petaflop – a unit for measuring a computer’s speed – and will rely on a technology that cools systems using warm water. The supercomputer for NCMRWF, which is in Noida, will have a computing capacity of 8.3 petaflop for weather and climate modelling and to support advanced numerical weather research across India. The supercomputer at IITM, Pune, will have a capacity of 13 petaflops of power for atmosphere and climate research. 
 

More is better
 

The news about the contract is of significance for it shows the growing use of high-performance computers (HPC) by government and private companies.
 

Arvind Bajaj, senior vice-president & country head, of big data & security, Atos, said the company had provided HPCs for India’s National Supercomputing Mission (NSM). “We have about 14 sites running as part of the NSM. These are mainly for academic and research-based work. This is the first one that will be used for the meteorological needs in India from Atos,” he said.

The usage of HPCs in the government sector has strengthened in the last few years. “We see a huge potential in more specialised supercomputers. The market in the last three to four years is also shifting from small machines to large machines and more specialised use cases in various sectors. The Government of India is investing and acquiring some large capacity and capabilities to solve concrete problems, such as weather and climate or others. We are seeing the demand going up,” said Bajaj.
 

Investment opportunity
 

Atos won in 2018 the biggest tender by the Centre for Development of Advanced Computing (C-DAC) for providing HPC under the NSM. Starting 2019, the company has delivered HPCs to state organisations like the Indian Institutes of Technology, Indian Institute of Science, Institute for Scientific and Engineering Research, and National Agri-Food Biotechnology Institute

Emmanuel Le Roux, group SVP advance computing, HPC, quantum and AI at Atos, said India was among its two largest markets outside of Europe. “We are also investing in India. We have a competency centre, where we are delivering and developing some of our IPs (intellectual property) in Bangalore. We also invested in a factory so that we could build HPCs under the Make-in-India campaign. We see a huge opportunity in India,” he said.
 

HPC adoption is expanding because the solutions that these systems provide have a societal impact. “After Covid, countries have realised that access to high computing becomes a sovereignty aspect. Also with AI (artificial intelligence) and the data generated, companies are realising that they are not able to extract value from the data (without using HPC),” said Le Roux.

 In the private sector, the use of HPC is powered by cloud computing. Hewlett Packard Enterprise (HPE), one of the largest players in cloud computing, has said the adoption of HPC is increasing, according to media reports.
 

Technology provider Dell Technologies has said organisations building infrastructure in India for generative AI systems like ChatGPT, the popular AI chatbot that responds to a range of written queries, were creating a demand for HPC solutions.

“High-performance computing traditionally has been with enterprises, education and research. Enterprises have used it in areas like simulation, computer-aided design, and manufacturing but the evolution of AI has created this new buzz,” said Srinivaschary T, lead solution architect at Dell Technologies India.
 

Such computing is needed as the graphics processing unit (GPU), which is a chip or electronic circuit capable of rendering graphics for display on an electronic device, has taken the front seat as an AI accelerator, according to Srinivaschary. 

The performance-intensive computing as a service (PICaaS) market – it includes HPC for AI – is expected to grow from $22.3 billion in 2021 to $103.1 billion in 2027, according to the International Data Corporation (IDC). The segment is projected to have a compound annual growth rate (CAGR) of 27.9 per cent between 2022  and 2027.
 

“The government is also pushing very hard to get more installations to be listed in the top 500 computers in the world. There is a concerted effort from the Government of India to identify large projects like weather modelling to core scientific research,” said Srinivaschary.
 

Consumer applications like in-car analytics were rapidly growing in India too, he said. “Some startups have started to reach out to us to build a foundational model like ChatGPT from base and not necessarily leverage whatever is existing outside. That has kick-started a huge spurt in behaviour.” 
 

HCLTech said recently it was expanding its collaboration with Microsoft to provide HPC solutions for clients in various industries. 
 

“High-performance computing is becoming increasingly important as organisations look to gain insights from vast amounts of data,” said Nidhi Chappell, general manager, Azure HPC AI, Microsoft.