Cool stuff
Driven by the steep rise of AI, gigantic data centers are booming. Measurement technology from Endress+Hauser helps their cooling systems operate safely and efficiently.
Viewed from outside, you would think they were giant warehouses. But inside, rather than stacks of shipping crates, there are rows upon rows of high-performance servers. Data centers are currently springing up all over the world to meet the immense demand for computing power to drive AI and cloud services. US tech giants – including Meta, Amazon, Apple, Google and Microsoft – and other companies are collectively investing hundreds of billions of dollars in their infrastructure. One example is Meta’s facility currently under construction in the US state of Louisiana, which will cover an area roughly the size of 70 soccer pitches.
As the computing power in data centers grows, so do the challenges. Most people are familiar with how loud a personal computer can be under heavy use, when fans accelerate from a whine to a roar as they blow more cool air into the system. However, a high-performance graphics processing unit (GPU) requires 10 times more energy than the CPU chip in a typical home computer or laptop. When up to 100,000 GPUs are packed together in a tight space, temperatures rise, and so does the need for cooling.
Cooling can account for up to 40 percent of a data center’s energy consumption. In search of a sustainable solution, most operators today are turning to energy-efficient liquid cooling instead of air cooling. This involves absorbing heat directly from the processors and dissipating it via a central heat exchanger. “We have been involved in the development of this advanced technology from the very beginning. Today, Endress+Hauser is a leading industrial instrumentation provider for the AI revolution in the country,” says Liza Kelso, National Business Development Manager for Semiconductors and Data Centers in the US.
Uncharted territory
When liquid cooling was introduced in facilities a few years ago, there was no off-the-shelf solution. So Endress+Hauser partnered with plant engineers and system integrators to specify, size and integrate the right instruments. Cooling in modern data centers is run from central utility buildings (CUBs) that house systems of interconnected modular process skids. Inside every CUB are massive chillers, heat exchangers, control cabinets and other equipment that use real-time data to maintain optimal temperatures for the servers. The cooling system in just one CUB may contain several dozen measuring instruments to monitor flow, temperature, pressure and water quality. One data center often combines four or more skids, so on a campus there can be thousands of measuring instruments working together.
“Our high-precision process measurements can help optimize cooling system performance and manage resources effectively and sustainably,” says Kelso. In addition to minimizing energy and water consumption, system reliability is a top priority for providers: Data center downtime due to overheating would cost around 150,000 US dollars – per minute. Beyond this sobering fact, operators generally are paying more attention to the PUE (power usage effectiveness) value. This compares the total energy consumed to the energy consumption of the IT infrastructure. The closer the value is to the ideal 1.0, the more efficient the data center, as almost all the energy supplied is converted into computing power. In Europe, new data centers will not be allowed to exceed a PUE value of 1.2 beginning July 2026.
Annual production increased fivefold
Endress+Hauser impressed the system providers and facility operators not only with its comprehensive portfolio of measurementtechnology for flow, liquid analysis, level and temperature, but also with its agility and customer focus. Because when it comes to building data centers, every day counts. “For the first major order, the Group had to deliver five times the normal annual production of one type of instrument – and the responsible product center immediately increased its capacity accordingly,” recalls Kelso. She is convinced that, thanks to their extensive expertise, the teams at Endress+Hauser can help AI data centers overcome other challengeson the path to sustainable growth: “The centers will increasingly generate their own electricity on site, from sources including renewable energy. In addition, water consumption is to be further reduced – for example, by treating wastewater from surrounding communities. Here, too, we can provide support with measurement technology and application knowledge.”
The heat is on
The rise of digitalization, cloud services and AI is leading to more and bigger data centers. In 2024 alone, over 500 billion US dollars was invested globally, according to the International Energy Agency (IEA). The IEA expects these centers’ power needs to double to 945 terawatt-hours over the next five years – that’s the equivalent of Japan’s electricity consumption. Today’s large AI data centers already consume on average as much electricity as 100,000 households. Future facilities will consume up to 50 times that amount. Operators must therefore find sustainable solutions for power generation and increased energy efficiency.
Downloads
Explore additional resources
