New Delhi, Sep 19 Mumbai has emerged as India’s data centre capital, accounting for 40 per cent of the nation’s total capacity and 44 per cent of live IT capacity, a report said on Friday.
In the first half of 2025, Mumbai’s capacity jumped 14.3 per cent to exceed the 4 GW milestone, comprising 591 MW operational, 185 MW under construction and 3.2 GW in the pipeline, according to a report from real estate services firm Knight Frank.
This growth builds on India’s data centre market, surpassing 10GW in H2 2024, supported by 1.4GW live and 400MW under construction, the report added.
With 97.6 MW of take-up over six months, the city’s vacancy rate at data centres is at 5.4 per cent -- less than half India’s overall vacancy of 12.3 per cent. Two-thirds of under-construction space is already pre-leased to companies.
Three live Mumbai sites can support hyperscale deployments above 2.5 MW, and just one offers more than 10 MW available. Hence, the fragmented market of smaller 1 MW–2 MW facilities has encouraged global players and joint ventures to plan high-capacity campuses, such as NTT’s 500 MW NAV2 project and Blackstone-Panchshil Realty’s 500 MW AI facility.
Hyderabad, India’s second-largest data-centre market at 2.1 GW, is positioning itself for hyperscale growth with over 500 MW in new capacity via two major projects, the report noted.
“With over 3GW of capacity in the pipeline and strong policy support for green data centre parks, Mumbai is attracting sustained global investment," said Shishir Baijal, Chairman and Managing Director, Knight Frank India.
Its robust subsea cable connectivity, scalable power infrastructure, proximity to enterprise hubs, and progressive state policies make it South Asia’s gateway for cloud, AI, and enterprise workloads, he added.
“Mumbai has firmly established itself as the epicentre of India’s digital infrastructure growth. While other metros like Chennai, Hyderabad, and Bengaluru are gaining traction, none match Mumbai’s scale, speed, and ability to serve as South Asia’s gateway for cloud, AI, and enterprise workloads.
Disclaimer: This post has been auto-published from an agency feed without any modifications to the text and has not been reviewed by an editor