AI is power-hungry. According to estimates, by 2030, data centres powering artificial intelligence could be consuming an extra 200 terawatt-hours of electricity every year. That’s a massive 160% jump from today’s levels, as per Goldman Sachs research.
That’s not all—AI also demands a lot of water, putting our limited resources under immense strain. Even a small 1 MW data centre can use up to 25.5 million litres of water annually, while large-scale data centres, such as those operated by Google, consume more than a billion litres each year.
To combat some of these issues, Microsoft introduced a new data centre design in August 2024, aimed at reducing water usage and promoting a more sustainable future. This innovative design incorporates chip-level cooling technology, effectively eliminating the need for water evaporation in cooling processes.
Microsoft isn’t alone. Driven by these sustainability concerns, Mysore-based software firm Vigyanlabs Innovations Private Limited took a proactive approach by developing its own micro green data centres.
The Genesis
Vigyanlabs has undertaken the complex task of making data centres more energy-efficient and reducing their carbon footprint. Srinivas Varadarajan, CEO and the visionary behind this innovation-driven company, shared with AIM the journey that led to the development of their energy-saving solutions.
As the former principal architect of Aadhaar, Varadarajan witnessed first-hand the massive energy consumption and carbon footprint associated with building a large-scale platform. It was during this time that he fully realised the significant energy demands of operating thousands of servers, leading to a greater awareness of their environmental impact.
In India, the ₹85,580 crore ($10 billion) data centre industry is primarily concentrated in Mumbai (49%) and Chennai, with other key hubs in Noida, Bengaluru, and Pune, as per reports. Additionally, the demand for hyperscale data centres is growing rapidly, with projections indicating a 35% annual increase until 2040.
In the past five years alone, Hyderabad, Navi Mumbai, and Pune have collectively committed 440 acres to hyperscale development.
Through extensive research, Vigyanlabs’ team discovered that the energy consumption of data centres was comparable to that of the aviation industry. Designing an energy-efficient data centre, however, is no small feat.
“The process of building data centres is highly sophisticated. It involves civil, electrical, cooling systems, network and IT infrastructure, software —all of which add to its complexity,” he explained.
With multiple vendors contributing to different aspects of the infrastructure, addressing this issue required a systematic and comprehensive approach.
Vigyanlabs’ journey began in a startup incubator at SJCE-STEP, a government-backed business incubator in Mysore, where a group of engineers had access to 1,000 computers to test their algorithms. This research led to the development of their flagship application, which provided approximately 42% energy savings on desktop versions.
“We built the product and won the Nasscom Innovation Award for it. That recognition gave us visibility and opened doors to collaborations with major institutions like SBI and other large companies,” Varadarajan recalled. Since then, Vigyanlabs’ technology has saved nearly 100 gigawatt hours of energy.
Innovation that Makes an Impact
At the heart of their innovation is Vigyanlabs’ sensor-based Enterprise Power Management Software, featuring a unique soft energy meter (SoftEM)—an application that measures energy consumption across laptops, desktops, and servers. Its flagship product, Intelligent Power Management (IPM+), supports Windows, Linux, servers, UPS systems, energy meters, and printers.
Expanding beyond desktop computers, Vigyanlabs turned its focus to tackling broader energy inefficiencies in data centres. It found that nearly 50% of the energy consumed by data centers was either wasted or used inefficiently.
Recognising that different users have varying computing needs, the company developed adaptive algorithms that learn and adjust based on individual usage patterns, ensuring more efficient energy management.
Today, Vigyanlabs’ energy optimisation solutions are deployed on millions of machines worldwide, significantly reducing companies’ carbon footprints. Their algorithm actively runs on between 1.5 and 6 million machines globally, making a substantial environmental impact.
Vigyanlabs Green AI Chatbot
Besides, the company has a network of micro data centers, backed by 7+ US patents. It operates on renewable energy and consumes 50% less energy compared to hyperscalers. Vigyanlabs also claims that its green micro data centre network operates with zero water for cooling. By optimising chiller operations, it has successfully reduced PUE (power usage effectiveness) from 2.0 to 1.20.
This data centre relies on an open source large language model (LLM), which has been further optimised, called Mysore.ai—a green AI chatbot.
What makes Mysore.ai special? According to Vigyanlabs, its uniqueness lies in its significantly lower cost and energy footprint, claimed to be the order of magnitude less than the leading AI-supported chatbots such as ChatGPT (OpenAI), Perplexity.AI, Gemini (Google), and CoPilot (Microsoft).
Unlike these models, Mysore.ai can run on a standard computer with no GPU or a low-powered GPU and uses open source tools, allowing free usage. The company shared its passion for building its own AI stack using fewer resources, much like DeepSeek, rather than relying on commercial AI technology.
Unlike traditional AI models designed for general-purpose tasks, Vigyanlabs’ AI system is specifically built to optimise energy use across all operations. It manages the entire data centre, including servers, storage, cooling, electrical, and security systems, automating processes to minimise errors and ensure optimal performance.
Recognising the difficulty of retrofitting existing buildings, the company chose to build its own facility with sustainability as a key focus. To minimise environmental impact, it designed special bricks that reduced cement usage by 70%, allowing the building to stay cool naturally and reduce air conditioning needs.
This design achieved a 30% energy saving of the cooling system, reducing the building temperature by 6 degrees Celsius. “We designed transformers with only 5% energy losses compared to the usual 10-15%.”
Additionally, Varadarajan mentioned that a real-time energy auditing system was introduced to ensure efficient power management. Cooling remains one of the biggest energy consumers in data centres, with traditional chillers relying heavily on water.
The initiative extends beyond their own operations. “We are actually being called around the world to help build national energy policies. We worked in the Philippines, Togo, and several other countries as technology advisors to tackle similar challenges,” Varadarajan added.
