Experience Superior Performance with Auto Scalable eNlight Managed Cloud
eNlight cloud gives you complete control and power to do more as you achieve superior performance on the World’s First Auto-Scalable cloud.
- Create on-demand Virtual Machines
- Reboot, Start and Shut down VMs, independently
- Deploy & assign resources dynamically
- Intelligent Auto Vertical and Horizontal Scaling
- Know Resource Usage before-hand, through eNlight's wise & Timely Predictions
- Pay Per Consume, eliminate the cost of unused resources
Register for eNlight cloud services by paying a one-time fee of Rs. 100 + GST which will be adjusted against your final cloud package amount once we begin your billing cycle.
How Indian data center operators are mitigating AI power consumption woes
Indian data centre (DC) operators are joining the race to mitigate the impact of humongous power consumption by AI compute infrastructure that has raised concerns among sustainability experts globally, and the impact could also control their operational costs to some extent.
Digital Connexion, the Reliance-Brookfield-Digital Realty joint venture DC operator, said it was using direct liquid-to-chip cooling, immersion cooling and magnetic levitation air-cooled turbocor chillers, curbing energy consumption by up to 15%. Its Chennai data centre has no water dependency.
ESDS Software Solutions said it was using APFC (automatic power factor control) panels provision and metering & monitoring for effective energy management.
CtrlS Datacenters said it was investing in energy storage solutions to balance grid loads and maximize renewable energy utilization.
With the Indian government and private players set to acquire 100,000 GPUs by the end of 2024, Deloitte estimates the power consumption is set to rise by 300MW.
This is a sharp increase from the existing installed capacity of 726MW of the entire industry, data from JLL Research showed.
“However, at a baseline level, the apparent increase in AI power consumption aligns with India’s consumer growth needs, and should not cause any major supply disruptions,” said Viral Thakker, Partner and Sustainability and Climate Leader, Deloitte South Asia.
GPU servers consume 20 times more energy than CPUs and need more cooling in a tropical country like India where indoor temperatures touch 29 degrees.
“An AI data centre can consume up to ten times more energy compared to a traditional CPU data centre with the same server count,” said CB Velayuthan, CEO, Digital Connexion, which recently operationalized its AI-ready data centre in Chennai and is planning another one in Mumbai.
To mitigate the impact of energy consumption, “we are provisioned with advanced liquid cooling solutions such as Direct Liquid to Chip Cooling (DLC) and Immersion cooling for high-density deployments,” he said adding that the company’s Chennai data centre is India’s first to incorporate magnetic levitation air-cooled turbocor chillers, curbing energy consumption by up to 15%.
The data centre, which has allocated two floors for high density AI deployments, also has higher sustainability with no water dependency, Velayuthan said.
Besides, Brookfield’s $3 billion renewable power and transition portfolio comprising 20 GW of wind and solar assets, provides the company a robust source of green power, he explained.
The AI revolution has also raised concerns over water scarcity as data centres directly consume water for cooling purposes and indirectly through nonrenewable electricity generation.
“India, with 18% of the world's population but only 4% of its water resources, faces severe water stress,” said Vinit Mishra, Partner, Technology Consulting, EY India. “Failure to adopt sustainable practices such as waterless cooling technologies, could potentially force data centres to relocate, particularly in water-stressed locations like Bengaluru.”
“We (India) are on track to reach 10 GW within the next seven years. areas like Mumbai, Chennai, and Delhi-NCR are even more challenging with the average temperature crossing 29°C,” said Piyush Somani, founder, CMD and CEO of ESDS, adding that geographies like US, Europe are better off at 17°C.
Therefore, ESDS has constructed sun louvers over the building to ensure minimal sunlight penetration to networked areas.
Indian DC operators are already making exceptional efforts to tap renewables solar power, he said. “Preliminary results are very encouraging, and studies point to a possible 30-40% drop in our overall power footprint."
CtrlS, which also uses liquid cooling and adiabatic cooling systems, said “AI for predictive maintenance and optimization, is continually improving our PUE (power usage efficiency) metrics,” said Vipin Jain, President, Data centre Operations.
However, these efforts do not allay the concerns of ecological impact of AI computing.
Google, which targets reduction of its carbon emissions to net zero by the year 2030, said last week that AI poses a significant threat to this goal. In its 2024 Environmental Report, the search engine giant admitted that it produced 13% more carbon emissions in 2023. This translates to 14.3 million metric tons of carbon dioxide pollution
To curb this, the likes of Amazon Web Services are turning to nuclear power to fuel their projects.
“Environment and sustainability are a global concern, whether a DC operates from Dallas or Chennai,” said Saurabh Rai, CEO of sustainable technologies firm Arahas. “In a single month, the power consumed by AI operations in India could light up the iconic Taj Mahal for 1,000 years. The annual carbon emissions from India’s AI industry could potentially exceed the lifetime carbon footprint of 5 million cars.”
This begs the critical question of whether tech companies should pay for the negative environmental effects of their AI activities, he posits