Data Centers – Todays version of Rail Roads and Highways

Just like the railroads in the 1800s and highways in the 1950s catapulted economic development in the US, Data Centers, powering the world’s information superhighways, are bringing in rapid global economic and societal change. The global installed capacity of Data Centers is ~85-90 GW. Generative AI, Cloud, SAAS, 5G, Edge computing, video streaming, cloud gaming, IoT and Industrial data, the key drivers of compute, are expected to drive a 15% CAGR to reach 127 GW by 2028 and 150-160GW by 2030. India represents a very small portion of this overall pie with an installed capacity of ~1-15GW going to ~2.5-3GW by 2030.

Power and efficiency evolution over the decades

The data centers of 15 years ago resembled an air-conditioned server room with bare bones hardware. Today, data centers resemble a mini-power station packed with CPU and GPU clusters. A modern-day “AI-ready” data center is planned for 50-100MW, and a mega-campus like Microsoft Fort Worth or Google Stargate targets 1GW of an eventual draw. Data centers are also power guzzlers where massive amounts of energy are required for the compute as well as cooling. About 60-70% of the power runs the servers, storage, networking and about 30 – 40% is consumed in cooling, power conversion, lights, security cameras, etc. The efficiency of data centers has dramatically increased over the last 15 years. A critical metric defining efficiency is PUE – Power Usage Efficiency, which is the ratio of Total Facility Power / IT Equipment Power. A PUE of 2.0 means that the site consumes one extra watt of power for cooling and other applications for every watt of computing. The average PUE of the industry has evolved from 1.92 in 2010 to 1.56 in 2024. This efficiency improvement has been driven by advances in efficiency of server grade silicon, virtualization and containerization, and advances in cooling. Some of the best hyperscalers operate at a PUE of 1.2.

Cooling becomes more critical; substantial advances in cooling technologies

Thanks to a multi fold leap in server efficiency and AI accelerators, operators have been able to cram 3-4x more compute into the same rack footprint. With this increase in compute density, effective cooling has emerged as a critical parameter in data center operations. Cooling technologies have become more advanced and energy efficient. Several of these such as hot-aisle / cold-aisle containment, VFD fans, free-air economizers, precision cooling and the first wave of liquid cooling shift heat 3-5x more efficiently and are pushing the PUE closer to the theoretical ideal of 1.

2 °C cooler burns a hole in the pocket

Yet, even the most cutting-edge hardware stays safe only if the air entering it is cool enough. Airflow can be blocked by an errant cable, a blanking panel can fall out, or a rack can recirculate its own exhaust – creating hidden hot-spots the room’s average temperature does not reveal. In order to tackle this problem, operators keep the entire data-hall a couple of degrees cooler than necessary “just in case”. Keeping the whole hall just 2 °C centigrade colder than necessary can typically add 14-18% to the cooling plant’s energy draw. An illustrative example is a data center with a 10MW of load at a PUE of 1.55 consuming 15.5 MW in total. Overcooling by 2 °C can add up to 0.5-0.8MW nonstop, or 4-4.7 GWh annually, adding about INR 3 to 5 crore to energy costs.

Syrma SGS’s solution to unlock the energy savings potential

Syrma SGS, a leading Indian EMS and RFID company, by leveraging decades of in-house expertise in high volume RFID design and manufacturing, has developed a solution based on an ultra-thin, battery free temperature sensing label. This label simply peels onto the front and rear of every rack. Each tag harvests power from a standard UHF reader and instantly streams inlet and exhaust temperatures into any DCIM platform. Accurate, real time and rack level temperature data lets operators see hidden hot-spots, safely increase thermostat set-points, and fine-tune liquid loop flow without fear of throttling GPUs or tripping alarms. By pairing its tags with pre-calibrated reader kits and open APIs, Syrma SGS turns granular thermal telemetry into an easy retrofit – unlocking the full energy savings promised by next generation cooling while keeping mission critical hardware in safe thermal zone.

Reach out to us. Our engineers are eager to speak to you and help drive the next level of energy savings.