Artificial intelligence is all over the media—and so is its energy consumption, which is spurring something of a new tech race. That race, however, besides innovation, includes some simpler moves, such as a return to a kind of power generator that had been consigned to history.
Single-cycle gas turbines used to be popular before combined-cycle technology emerged. It was much more efficient and a lot cleaner, Bloomberg reported this week in a story looking into the latest developments in AI and securing the energy for its continued growth. Single-cycle gas turbines work by burning gas and using the heat to generate power. Combined cycle turbines work pretty much the same way, but they also utilize the excess heat to turn it into steam, which also generates power (or is used for district heating).
Yet, while Bloomberg focuses on the fact that single-cycle turbines are less efficient and emit more carbon dioxide, they have certain advantages that may well explain why AI developers have become fans. For starters, gas plants with this type of turbine can ramp up more quickly than combined-cycle ones. Fast response times for power generators are bound to be important for tech majors. But single-cycle gas turbine facilities are also cheaper and faster to build than combined-cycle power plants. And this is even more important for AI developers.
Meta just inked a 20-year power supply deal with Constellation Energy, which operates a nuclear power plant in Illinois. Interestingly, the plant was supposed to be shut down back in 2017. It wasn’t, because Illinois legislators decided to prioritize both reliability and emission reduction and extended the life of the facility. This is the latest in a string of such deals in which Big Tech demonstrates a newfound appetite for nuclear power because of its AI ambitions. But this appetite is suffering certain constraints: new nuclear power plants take a whole lot of time to get built.
“There is a high level of urgency in the industry to get power fast,” the chief operating officer of Crusoe, a data center-building company, told Bloomberg. “We have tried to be creative about the energy component of data centers,” Cully Cavness also said. Crusoe is building the data centers for the Stargate project, a joint venture between OpenAI, Oracle, SoftBank, and MGX that is eyeing an international network of such centers. It is by far not the only one.
“To have the energy we need for the grid, it’s going to take an all of the above approach for a period of time,” Kevin Miller, Amazon vice president of global data centers, said earlier this year. “We’re not surprised by the fact that we’re going to need to add some thermal generation to meet the needs in the short term.”
Indeed, no one should be surprised that the tech industry would need to add a lot of thermal generation in short order to satisfy its energy demand. And if anyone thinks that it will be for the short term only, then they are deeply mistaken. Whether it’s a single-cycle gas power plant that could be built in a few months or a nuclear facility that will take years, AI needs reliable electricity—lots of it. Ensuring the supply of such electricity is a long-term investment, not something you pay for, and then shut down five years later because of emission regulations.
It is rather ironic that none other than Big Tech, a most vocal industry backer of the energy transition, is bringing the issue of supply reliability—and the importance of hydrocarbons—back into the spotlight of the energy stage. Perhaps even more ironically, there is a shortage of gas turbines already, so those single-cycle turbine power plants may take a bit longer than a few months to build.
By Irina Slav for Oilprice.com