Green Revolution Cooling (GRC) designs and manufactures single-phase liquid immersion cooling systems for edge data centers with powerful Graphics Processing Units (GPU) use, for applications such as AI and high-performance computing.
Can you explain what GPU as a service is and how it relates to your business?
GPU-as-a-service is what customers use instead of building their own data centers for intensive applications such as AI learning, training, inference and other high-performance computing applications like video animation and weather forecasting. For example, hyperscalers such as Microsoft, AWS and Google build massive data centers and “rent” their compute power as a service to other companies. Another example is the Texas Advanced Computer Center (TACC). TACC allocates a GRC system to the National Oceanic and Atmospheric Administration for eight hours a day during hurricane season to run weather models that predict hurricanes and provide warnings to millions of people.
What are the main types of cooling and how does immersion compare?
Data centers generate tremendous amounts of heat. In a typical data center, for every watt used to power servers you may need another half watt, and sometimes a full watt, just for cooling. Air cooling, the standard approach for decades, blows air through servers like blowing on hot soup. However, GPUs and processors have become too hot for air to cool effectively. A newer approach is direct liquid cooling, or liquid-to-chip, where cold water flows through plates attached to processors. But other components such as memory and network interface cards still require air cooling. The cold water flowing through these plates is conductive, so if it leaks, it damages the most expensive assets in the data center–the servers. Manufacturers are improving things, but leaking risk is a genuine concern.
Our approach is immersion cooling, which submerges the entire server in non-conductive, non-corrosive fluid, typically a refined synthetic hydrocarbon oil also used in cosmetics and motor vehicle lubricants. If liquid-to-chip is like putting an ice pack on your forehead. Immersion is like getting in a nice tub bath. This complete cooling solution is energy-efficient, protects servers from humidity, oxidation, airborne contaminants and sea mist and can extend server life while reducing costs. Generally, the fluids last the life of the data center, which is about 15 years. And recycling and disposal systems already exist for the hydrocarbon solution.
What is the ICEraQ™ Nano?
The ICEraQ™ Nano began as a demonstration platform for trade shows. Typically, once our liquids have captured heat, you pass that heat in a water loop through plumbing to a cooling tower on the roof. This setup is difficult to replicate in an exhibition so we built a smaller liquid-to-air system to show customers how a live server operates. Cisco asked us to scale this for small data centers in retail stores, hospitals, banks and similar environments. The Nano uses immersion cooling with a liquid-to-air heat exchanger, operates quietly, works with existing HVAC systems and eliminates the need for extensive plumbing. Impacts can already be seen in defense, telecommunications and retail applications with customers who have up to 4,500 locations where these systems could go.
How do immersion systems cope with changing conditions?
Servers do not respond well to changing conditions like thermal shock associated with rapid temperature changes. GRC systems are highly instrumented and monitor fluid temperatures in real time. If the fluid temperature rises, the software signals the pumps to move faster in the tanks, which removes the heat faster. If temperatures drop below target, the software slows the pumps to save energy. The result is a continuously optimized system that maintains stable temperatures while minimizing energy consumption. This system absorbs heat about 1,200 times more effectively than air, maintaining stable conditions and reliability. Even in extreme situations, modern servers have built-in throttling mechanisms that reduce performance before overheating becomes a problem. Aside from these controls, the analogy I give for why liquid-cooled servers do not overheat to the extent that they might with other technologies is: imagine going outside when it is 50 degrees Fahrenheit outside. You could stand there for quite a long time before feeling like you had to go back indoors. If you jumped into a swimming pool whose water was the same temperature, you would be hypothermic within two minutes. That is the power of the fluid system and its ability to absorb energy.
What is the ElectroSafe program?
Early on, we used a fluid from ConocoPhillips and labeled it ElectroSafe. As the market matured, fluids became standard off-the-shelf products. Today, ElectroSafe is a certification program pioneered by Green Revolution Cooling to test and qualify third-party fluids for immersion systems. We do this to protect our customers' investment in the expensive servers they use to run these powerful applications.
We have our own Hippocratic oath: do no harm to the servers. The program also benefits customers because large energy companies with massive R&D resources compete on performance and price.
Where is immersion cooling seeing the greatest uptake?
The United States represents the largest market and from a market acceptance standpoint, Asia and the Middle East stand out. Asia tends to be a little less risk-averse and more willing to experiment with new technologies. In the Middle East, operators face unique challenges because water is scarce and temperatures are extremely high. Europe is more progressive on sustainability policy but we have surprisingly not seen that translate into adoption of immersion cooling at scale.
How are you working with industry groups and policymakers?
AI factories are placing significantly more demand on the world’s energy and water resources, so we are seeing rapidly evolving policies that are trying to protect these finite resources. There is more public resistance to data centers because of concerns about the impacts of local energy and water usage on energy prices. GRC’s technology is a unique solution to reducing energy and water consumption. Fortunately, we are partnered with the biggest brands in the data center industry, which work directly with policy makers. GRC focuses on working with industry associations to advocate for policies beneficial to the industry and its customers.
What trends do you see shaping the future of the industry?
One factor is the emergence of “neo clouds,” new providers building large GPU-as-a-service data centers because hyperscalers cannot meet all the demand. Another major trend is edge computing, closer to where the action really is. These facilities are often located in harsh environments. Immersion protects them from surrounding conditions, making them more reliable and reducing downtime and servicing needs. Due to the market's demand to deploy AI as fast as possible, modular data centers 10 feet wide and 50 feet long that can be deployed in six months are experiencing significant growth. Immersion cooling can lower operating costs by reducing energy and water consumption for all of these units.
