High-performance computing (HPC) uses clusters of powerful processors, working in parallel, to process big data and solve complex problems at high speeds. HPC is believed to perform at a speed which is more than one million times faster. Experts believe with the use of GPU as a Service (GPUs) HPC can upgrade their performance. Also, a graphic processing unit (GPU) can offer computing power through a subscription-based model, which might be a cheaper software-as-a-service (SaaS) model in comparison to others.

Sign Up to get access to the Financial Express Exclusive and Premium Stories.Register NowAlready have a account? Sign in

The global GPU as a service market size was valued at $2.39 billion in 2022 and is projected to grow from $3.16 billion in 2023 to $25.53 billion by 2030, as per insights from Fortune Business Insights, a market research firm. Industry experts believe that the growth of GPUs might contribute to the adoption of artificial intelligence and machine learning in India. In conversation with FE-TransformX, Sunil Gupta, CEO, Yotta, a data-centric platform, is contributing to this. (Edited Excerpts)

Yotta provides HPC as-a-service, driven by state-of-the-art GPUs, which can offer supercomputing performance, extensive storage, optimised networking and scalability, among others. We provide our services at a reduced cost compared to other on-premise high-computing environments, keeping in mind the Indian market. Yotta provides a kind of DIY (do-it-yourself) toolkit for business to create their computing infrastructure as per their need.

What is Yotta’s USP?

Yotta has introduced Shakti Cloud which provides a consistent, full-stack solution for developing, training and deploying your model on a GPU-powered cloud instance. Yotta aims to democratize access to GPU resources, fostering innovation and competitiveness across various sectors. Yotta’s Shakti Cloud will deliver GPUs and various associated artificial intelligence (AI) and Platform-as-a-service (PaaS) services in a highly cost-effective manner on a per-hour usage model, with options for long-term reservations. We also have a transparent cost list, which is available on our website. Our price listings start as low as $1.20 per hour for 48GB to $38.08 per hour for 640GB of GPU memory.

With the adaptation of new tech comes the requirement of awareness. How is Yotta contributing to this?

Yotta also provides virtual pro workstations to understand the workings of GPUs. This not only helps to train engineers but can be a cost-effective method when compared to hands-on training. With the collaboration of artificial intelligence and virtual reality, we create training modules that allow users to understand the function behind GPUs. Our other services include high-performance computing – as a service, for compiling big data in aerodynamics. We also have customers from the health, BFSI (banking, financial services and insurance) and content-creating segments. We plan to use an NVIDIA-powered reference architecture with NVIDIA InfiniBand networking that will allow GPU clusters to deliver performance at large scale for large AI training and inferencing workloads, as well as HPC workloads.

Recently, you launched North India’s hyperscale data center, Yotta D1. How is that expected to change the game for businesses?

We see the potential for high-quality data centres in India’s infrastructure development, and we have doubled down on our Pan-India investment plans to Rs 30,000 crore. We have also signed an MoU (memorandum of understanding) with the government of Uttar Pradesh to invest Rs 39,000 crore in a phased manner over five to seven years . This investment will go towards constructing the data centre campus, and procuring IT equipment and other hardware by Yotta and its customers.

What are your plans to bring more GPUs into the Indian market?

Yotta has already placed an order for NVIDIA H100 Tensor Core GPUs, a powerful GPU for AI and HPC workloads. We plan to go operational with 4096 GPUs by January 2024 and 16,384 GPUs by June 2024. With complete support from NVIDIA and an aim to develop India’s sovereign AI landscape, Yotta also plans to scale up its GPU stable to 32,768 by the end of 2025. This is expected to directly address the huge demand for high-performance GPUs by research labs, enterprises, and startups for HPC and AI workloads.

Follow us on Twitter, Facebook, LinkedIn