In late August, Reliance Industries Chairman and Managing Director, Mukesh Ambani, shocked India’s data center industry by announcing ambitious plans to build a 1 GW, AI-ready data center in Jamnagar, located in the western Indian state of Gujarat. This would be truly phenomenal in a country where the total IT load capacity has just about touched 1 GW nationwide, with Mumbai alone accounting for at least half of it.
Today, Indians are consuming online services and generating data like never before. This, in turn, has generated an unprecedented demand for digital infrastructure and compute power. Now, add artificial intelligence and high-performance computing into this mix, and it all becomes truly mind-boggling!
While there are hyperscale data centers in all major Indian markets like Mumbai, Chennai, Hyderabad, Bengaluru, and Delhi/NOIDA, their IT load capacities are nowhere near GW-scale. It is also a fact that the nation of 1.4 billion tech- savvy people remains grossly underserved when it comes to data centers.
Are GW-scale data centers on the horizon?
“It is inevitable,” says data center industry expert Sandeep Dandekar. In fact, unlike others who see only edge data centers or facilities with smaller IT capacities being built in smaller towns across India, Dandekar feels even hyperscale data centers would find their way into India’s interiors. “With the rise in total demand, I see a rise in density levels and a need for latency- free high-speed high-volume data flow right up to deeper interior locations of the country in the near future,” he says. Dandekar expects the market to grow at a CAGR of at least 20 percent, and that this would lead to greater adoption of new technologies in the areas of automation, as well as new age cooling technologies that will save power. However, he cautions, “In case the total capacity availability does not grow at that pace, the steam will be wasted, and that will constrain growth achieved in various sectors.”
In its report titled Is India Building Enough to Power its Digital Transformation, Cushman and Wakefield found that India’s under-construction Colocation data center capacity addition stands at 1.03 GW for 2024-2028. An additional 1.29 GW is planned, and operators have been building land banks with 3-4 GW development potential over the medium to long term. When it comes to hyperscale data centers, the report found that the total operational cloud self-build capacity stood at 68 MW as of end-2023, with under-construction and planned capacity of 96 MW and 226 MW, respectively. “Hyperscalers have also moved ahead on land banking, with AWS and MS Azure acquiring land parcels across Mumbai, Thane, Hyderabad, and Pune for future development,” found the report.
“At the rate at which India’s data center industry is growing, GW- scale data center campuses are no longer a luxury but the only option that we have,” says Rohan Sheth, Head – Colocation and Data Center Services, Yotta Data Centers. Yotta is a major player in India’s data center market that already operates hyperscale data centers across India, including the world’s second largest Uptime Institute Tier IV-certified facility in Navi Mumbai. “Yotta’s Navi Mumbai DC 1 has 45 MW of IT power with a 48-hour backup. Our DC 2, which is core and shell ready, has an IT power of 60 MW. In addition to these 2 buildings, Yotta’s facilities offer strategic scalability with the capacity to expand up to 1 GW in Panvel,” he shares.
What is driving the demand for Hyperscale data centers in India?
Just taking a closer look at internet and mobile phone usage figures should be enough to understand the demand-supply gap surrounding digital infrastructure. According to the Telecom Regulatory Authority of India (TRAI), at the end of December 2023, India had 936.16 million internet subscribers and 1.15 billion wireless telecom subscribers.
Moreover, given the government’s push towards cashless transactions, in the wake of the 2016 demonetisation initiative, and the impact of the COVID-19 pandemic, Indians have taken to digital banking and payment services in a big way. Unified Payments Interface (UPI), which is a mobile-based payment system in India developed by the National Payments Corporation of India (NPCI) has made real time, round-the-clock digital transactions convenient. So much so that it is now even used by autorickshaw drivers, street-side food vendors, and neighbourhood mom-and-pop stores to accept payments from customers!
Meanwhile, e-commerce and OTT platforms are thriving, as are apps offering services like food and grocery delivery, car hire and taxi booking, and flight and rail reservations, not to mention the ubiquitous social apps.
According to the 2023-2024 Annual Report of the Reserve Bank of India (RBI), which is India’s central bank and the regulatory body overseeing India’s banking sector, the Department of Payment and Settlement Systems (DPSS) has recorded at growth of 44 percent in terms of transaction volume during 2023-24, on top of the expansion of 57.8 percent in the previous year.
Building GW-scale Data Centers
So, the big question is, can India truly build and operate data centers that will have a capacity equivalent to the current IT capacity of all its data centers put together?
“Why not?” asks Vinod Javur, COO, Digital Edge DC, adding with cautious optimism, “As long as operators can arrange for adequate power, it is possible.”
Yotta’s Sheth concurs saying, “The capital expenditure on setting up these data centers averages between USD 4-10 million per MW, depending on the geography, with the largest component being power, which constitutes about 30 percent of the total capex.” He further says, “Securing clean and green power of such a large quantity and also unlimited bandwidth of fiber are the key challenges, along with cap-ex costs and energy consumption for GW-scale data centers.”
But despite the challenges, one cannot deny that India’s hyperscale journey has just started, and GW-scale data centers will not be a distant dream for long.
***
This piece was originally published in Issue 6 of w.media Cloud & Data Center Magazine
