A Startup Wants to Power Data Centers With ‘Supercritical’ Carbon Dioxide

A new company has a plan to make the dirty gas turbines powering data centers more efficient: liquid carbon dioxide. American Supercritical came out of stealth Wednesday, announcing $8 million in funding. It wants to retrofit inefficient gas turbines that many data centers rely on for power with units that can generate more power, without adding more emissions (though the gas-fired turbines will continue to emit carbon pollution). The technology can also theoretically be used on a wide variety of energy sources at a time when power demand is skyrocketing. “We want to start with gas turbines but eventually expand beyond that,” says cofounder Simon Shuham. Most large gas-fired power plants in the United States use an array of heat engines in what’s known as a combined-cycle process: First, turbines generate electricity from burning compressed air and natural gas, then a separate engine uses the hot exhaust to make steam and create additional energy. But for a variety of reasons, data centers across the US have opted to power their operations with what are known as simple-cycle turbines, and exclude the steam component. These turbines are much less efficient than combined-cycle plants. Usually, only about 35 percent of the energy from simple-cycle turbines is converted to electricity, while the rest escapes as exhaust. (In combined-cycle plants, that figure hovers closer to 60 to 65 percent.) That exhaust includes greenhouse gases, making plants that run on simple-cycle turbines a much worse choice for the environment than combined-cycle plants. The size of some of these plants combined with their inefficiency is a recipe for climate disaster. A massive data-center power plant in Texas that Amazon is building with just simple-cycle turbines, for instance, is permitted to emit more than 33 million tons of greenhouse gases per year—more than the annual total of some small countries. But all these small, inefficient turbines could be a great match for supercritical CO2 technology, American Supercritical’s founders say. Carbon dioxide becomes supercritical when it’s pressurized and held at a certain temperature. In this state, it gets the density of liquid but still behaves like a gas, meaning it can move energy more efficiently through much smaller amounts of equipment. American Supercritical wants to attach its units to small gas turbines and help generate more energy. While the turbines themselves would still use gas, the supercritical CO2 unit can use the hot exhaust generated from those turbines to create additional electricity. Instead of using that heat to boil water and create steam, the heat is transferred directly by the pressurized CO2 to generate additional energy with no additional emissions. “We’re essentially building miniature combined-cycle plants,” says Shuham. Using supercritical CO2 also can eliminate or greatly reduce water use in the power generation process—something that’s drawn intense scrutiny when it comes to data centers. Importantly, the CO2 involved operates in a closed-loop system, meaning that it doesn’t have to be refilled. Cofounder Matthew Carlson, who researched supercritical CO2 for more than a decade, likens it to refrigeration systems that circulate CO2 to facilitate cooling. The company plans to initially sell a 10-megawatt unit—relatively small by gas plant standards. By American Supercritical’s math, adding supercritical CO2 units could increase turbines’ efficiency by up to 50 percent without adding any more greenhouse gas emissions or water use. But retrofitting all of the country’s inefficient gas turbines powering data centers with supercritical CO2 units won’t reduce emissions to zero. Natural gas production and use, even with efficient turbines, warms the planet. Data centers also represent a massive new source of pollution—right at a time when the world needs to be scaling back on its use. Using supercritical CO2 for power has been studied for more than 50 years in US national labs. Over the past two decades, research has moved away from theory toward trying out practical applications, assisted by technological breakthroughs that help with the pressurization process. Doug Hofer, an adviser to American Supercritical who spent two decades at GE as a turbine engineer, says that one important innovation has been the development of more compact and more affordable heat exchangers—devices that help both heat and cool the CO2. American Supercritical isn’t the only company trying to bring this technology to market, says Subith Vasu, a professor of engineering at the University of Central Florida, who runs a lab at the school’s Center for Advanced Turbomachinery and Energy Research. The companies still have hurdles to overcome to achieve commercial success. “Whenever you are venturing into new technologies, you need to have all the associated components, the supply chain,” in place, he says. Technical challenges also abound, he adds, “and cost has been a huge issue.” Hofer says that part of the delay has also simply been because larger companies are satisfied with the performance of traditional power plants and unwilling to invest in different technologies—and turbine makers are not interested in funding potential competitors. “It’s the innovator’s dilemma, right?” he says. “Getting this [technology] out there needs to have an outside influence.” The $7 trillion set to be invested in data centers globally by 2030 has helped open up a whole new potential market, especially for an industry racing to get as much power on the grid as it possibly can to come out ahead in the AI race. “There’s this incredible willingness to pay [for power], but they’re bottlenecked,” says Shuham. “The goal is to come in here, retrofit their existing systems, and give them the ability to build additional data centers—or get rid of some of the gas turbines.” Earlier this year, China announced that operations started at the world’s first geothermal plant using supercritical CO2. Carlson says that in his career, he’s worked on supercritical CO2 applications in nuclear, geothermal, and solar energy. He and Shuham say that their units could be converted to work with small modular reactors, a technology that’s enjoyed strong support from the Trump administration. (The company says it has already signed an initial agreement with one company, though it didn’t name it). It can work with other sources of heat as well. “We just need to get something hot, and then we can convert that into electricity,” says Carlson. Comments Back to top