Everything About NVIDIA-Yotta Deal, An India's Biggest AI Deal Ever

In December 2022, Yotta Data Services, an Indian data center startup, announced a multi-million dollar partnership with Nvidia to provide high-performance computing capabilities from data centers in India. The partnership includes Nvidia's 16,000 chips, and Yotta plans to double its Nvidia AI chip order to $1 billion.

The partnership is considered India's biggest AI bet yet. The existing plans are worth $500 million and will include the purchase of Nvidia GPUs for Yotta's data centers.

Mumbai-headquartered Yotta Data Services designs, builds, and operates data center parks in India. It offers services such as colocation, connectivity, cloud services, security, and managed IT. Yotta has data centers in Mumbai and Noida, and plans to expand to more locations in India and overseas. 

Yotta is aiming to offer high-performance computing capabilities from data centers in India. The company's AI chips will enable computer vision, speech, natural language processing (NLP), generative AI, recommender systems, and more.

In January this year, the company's CEO, Sunil Gupta, told Reuters that its latest AI purchase will be worth $500 million, and consist of 16,000 H100 and GH200 GPUs by March 2025. By the end of 2025, Yotta aims to expand its GPU inventory to 32,768 units. 

This collaboration involves Yotta Data Services investing in NVIDIA's GPUs, specifically the H100 Tensor Core GPUs, to bolster high-performance computing from their data centers in India. The deal is part of a broader initiative to support the development of AI services for corporations, startups, and researchers within the country.

The partnership is noteworthy for its scale, with Yotta Data Services is reportedly planning to increase its NVIDIA GPU inventory to about 20,000 by June 2024. These GPUs, known as Hoppers, are valued between $30,000 to $40,000 each and are integral to training large language models and building applications akin to OpenAI's ChatGPT and Microsoft GitHub Copilot.

This move by Yotta Data Services is seen as one of the biggest bets on AI in India, reflecting the company's ambition to be at the forefront of AI technology and infrastructure development. The total order book with NVIDIA is expected to reach $1 billion, marking a significant investment in AI cloud services.

Yotta's deal with NVIDIA is part it's strategy to support India's Al transformation, providing the necessary infrastructure for corporations, startups, and researchers to develop Al services. The GPUs will be used to train large language models (LLMs) and build applications similar to OpenAl's ChatGPT and Microsoft GitHub Copilot. It's a significant step towards making advanced Al capabilities more accessible within the country. 

What is the H100 Tensor Core GPU?

The NVIDIA H100 Tensor Core GPU is a cutting-edge processor designed for accelerated computing, part of NVIDIA's Hopper architecture. It's built to deliver exceptional performance, scalability, and security for data center workloads, particularly in AI and high-performance computing (HPC). Here are some of its standout features:
  • Performance: The H100 offers a significant performance boost for accelerated computing, capable of handling exascale workloads and large language models efficiently.
  • Scalability: It can connect up to 256 H100 GPUs using the NVIDIA NVLink® Switch System, enabling the acceleration of massive workloads.
  • Transformer Engine: This specialized engine is optimized for processing trillion-parameter language models, making it ideal for advanced AI tasks.
  • Memory: Equipped with 80 GB of HBM2e memory and a 50 MB L2 cache, it supports high-bandwidth and low-latency data access.
  • Tensor Cores: The fourth-generation Tensor Cores in the H100 provide up to 4X faster training for AI models compared to the previous generation.
  • Connectivity: It supports PCIe Gen5 and NDR Quantum-2 InfiniBand networking, ensuring high-speed connections.
These capabilities make the H100 a powerhouse for AI research and enterprise AI solutions.

Last December, Yotta launched "Shakti-Cloud", India’s Largest Super Computer of 16 Exaflops AI computing power with NVIDIA H100 Tensor Core GPUs to drive mass-scale AI innovation in India.

Yotta has partnered with India's business groups, including Tata and Reliance, to set up data center facilities in India. The partnership is intended to help Nvidia develop more country-specific artificial intelligence innovations and products.

Yotta has made GPUs available to customers on a cost-effective usage model.

Yotta's Founders & Backers

Yotta Data Services was co-founded by Sunil Gupta, who serves as the Chief Executive Officer. The company was established in 2019 with the backing of real-estate billionaire Niranjan Hiranandani. Additionally, Darshan Hiranandani is mentioned as the Co-founder & Chairman of Yotta. This team has been instrumental in Yotta's significant partnership with NVIDIA, aiming to bolster India's AI capabilities and infrastructure.

Yotta is part of the Hiranandani Group and operates its cloud regions at its two hyperscale data center parks in Panvel (Navi Mumbai) and Greater Noida (Delhi). Yotta has more than 800 enterprise and government customers, including including multinational hyperscale cloud operators, OTTs, governments at all levels, and enterprises in various tech-enabled sectors.

Yotta Data Services has invested in R&D on new-age technologies, onboarding of industry veterans, and has a vast pool of IT resources. It also has a deep partnership with OEMs and Start-ups.

Advertisements

Post a Comment

Comment

Previous Post Next Post