The recent acquisition by Yotta Data Services, a part of the Hiranandani Group, is gaining attention due to the global shortage of Nvidia H100 GPUs, causing delays in major projects worldwide. While countries like the United States and China have surged ahead in AI model innovation, Yotta aims to propel India into the forefront by constructing Shakti Cloud, envisioned as India’s largest supercomputer and potentially ranking as the 10th largest globally.
In pursuit of this ambitious goal, Yotta is set to acquire 16,000 of Nvidia’s H100 Tensor GPUs, a significant move considering Nvidia’s dominance with a 90% share in the GPU market. This acquisition is particularly noteworthy amid the global shortage of Nvidia H100 GPUs, leading to project delays globally. However, Yotta’s breakthrough in securing these GPUs followed a crucial meeting between Prime Minister Narendra Modi and Nvidia CEO Jensen Huang.
According to Sunil Gupta, Co-founder & CEO of Yotta, during Huang’s visit to India in September, the Prime Minister expressed his desire to bridge the innovation gap in AI, seeking Huang’s assistance. The Nvidia CEO pledged to support India’s AI ambitions by offering technology and support to local providers. Gupta states, “As somebody has to invest and take a step forward to build this infrastructure in India, Yotta took the first step and placed the order.”
Yotta plans to deploy the acquired Nvidia GPUs initially in its Mumbai data center, followed by centers in greater Noida and Gift City. However, the installation of the GPUs involves substantial preparation and backend work. Gupta highlights the need for upgrading the engineering of data centers, including the adoption of liquid cooling and immersion cooling capable of handling 40-50 kilowatts per rack. The deployment will commence with 4,096 Nvidia GPUs in February, with plans to scale up to 16,384 GPUs by June 2024. Yotta aims to further increase this to 32,768 GPUs by the end of 2025.
Gupta emphasizes the abundance of talent in India, with numerous startups working on AI solutions. However, he notes the critical lack of compute and network capacity for large-scale AI model training in the country. Yotta’s initiative is positioned to address this gap, providing Indian developers with access to the necessary infrastructure domestically, eliminating the reliance on foreign operators.
The Yotta CEO is confident that this initiative will enable Indian developers to create large-scale AI models without limitations, fostering a self-sufficient ecosystem within the country. With a focus on building robust AI infrastructure, Yotta’s endeavor aligns with India’s ambitions to emerge as a key player in the global AI landscape. As the company overcomes challenges posed by the global GPU shortage, its strategic move aims to contribute significantly to India’s technological advancement and innovation in the AI domain.
