Indian government to build AI computing infrastructure, report says
The Ministry of Electronics and Information Technology (MeitY) in India has recently unveiled a strategic AI vision document that outlines their plans to advance AI infrastructure within the country. India’s goal is to establish a computing infrastructure capable of facilitating a wide range of AI applications.
The infrastructure is structured into three layers – high-end compute, mid-range arm, and edge compute. Each layer has been positioned to address the computational needs of users, and this distributed architecture aims to ensure that users can transition between these resources, according to MeitY.
Given the increasing demand for generative AI and the training of large language models, it has become imperative for nations to have the computing power necessary to support these computationally intensive tasks. The high-end compute layer has been designed to cater to these requirements. As per the vision outlined in the AI document, this category of computing offers 40 exaFLOPs of AI processing power, with each cluster consisting of 10,000 GPUs.
The mid-range compute infrastructure, which falls under the second category, is dispersed across four different geographical centers within India. This segment has a capacity of 3 exaFLOPs of AI computing power, accompanied by 750 GPUs dedicated to AI training. Its primary purpose is to support both AI training and inference tasks specific to their respective regions, with the primary goal of reducing latency and enhancing the accessibility of AI resources.
Of equal significance is the third tier of AI computing infrastructure, known as edge compute, which is designed to cater to users with limited resources and smaller datasets. This component has 500 petaFLOPS of AI computational power and is equipped with 750 GPUs for AI training. These centers have been established across 12 geographic locations throughout India, ensuring widespread availability of AI capabilities.
As outlined in the vision AI document, India is committed to a substantial investment of 80 exaFLOPS, underscoring its dedication to advancing AI capabilities. In a similar vein, China has also unveiled plans to leverage even greater computing power, targeting 150 exaFLOPS.
Furthermore, the document highlights another initiative by the Indian Government, the India Datasets Platform (IDP). This platform provides access to a diverse range of datasets, enabling the development of data-driven AI models, accelerating AI research and development efforts, and facilitating the government’s transformation into a data-driven entity.
Generative AI: What it is and how it’s shaping edge infrastructure
Article Topics
AI | AI infrastructure | edge compute | MeitY
Comments