India is on the path to becoming a global leader in AI, but realising this vision requires a robust and scalable infrastructure. A distributed network of micro data centres (MDCs), designed to handle critical CPU and GPU workloads while occupying significantly less space and demanding lower upfront investments, represents the future of accessible, scalable, and cost-effective AI infrastructure in India.
This paper delves into the key components and requirements for establishing micro data centres, drawing on our research to define their scope and functionality. We categorise a micro data centre as one with a capacity of 25-300 kW, typically occupying around 800 to 3000 square feet. These data centres are crucial for expanding AI capabilities to the edge, enabling sustainable development through the integration of renewable energy. The flexible CPU-GPU ratio allows MDCs to scale efficiently, providing resilient and self-sufficient compute power necessary for India's growing AI use cases.
Currently, India hosts fewer than 10 micro data centres. Although MDCs lack a standardised definition or size, they are emerging as a critical segment in the hosting infrastructure landscape. Existing MDCs in India serve both domestic and international clients, but there is a pressing need to expand beyond mega data centres to include more micro facilities. The paper explores their use cases in sectors such as healthcare, banking, financial services, insurance (BFSI), and large-scale government operations. The increasing demand at the edge, driven by population growth in Tier II and Tier III cities and the rise of engineering universities focused on deep learning, highlights the importance of these centres.
The paper also examines the capital expenditure (CAPEX) and operational expenditure (OPEX) models associated with micro data centres. Our analysis suggests that building green MDCs can enhance cost-effectiveness, providing a compelling economic model. We predict that an investment of INR 60 crores in a MDC could yield a return of up to 3 times that of a larger data centre. Financing strategies and the potential impact on overall economic growth are also discussed.
Finally, we review the current policies in India that support the development and deployment of micro data centres. Major costs associated with software licenses and certifications (e.g., Uptime and TIA-942) need to be addressed through standardization and improved policy frameworks. Government initiatives, similar to the Udaan scheme, are necessary to foster the growth of smaller players in the MDC market.
This endeavour requires collaboration across various stakeholders, including compute users in both the private and public sectors, government ministries and states responsible for policy creation and enforcement, and investors. The concept of Open Cloud Compute suggests that if hundreds of smaller players can operate collectively like a large cloud provider, a network of micro players can function like a mega network, driving the next phase of AI infrastructure development in India.
