Gas Turbines: The Heart of AI Data Centers
Natural Gas Demand Set to Surge Soon
“A Choice Between AI and Consumers”
Editor’s NoteWe break down the essential yet often unfamiliar stories behind the technologies that power our lives—including AI, semiconductors, telecommunications, and biotech.
“By 2028, there will be a shortage of natural gas, and by 2030, the United States may run out of storage for gas.”
These were the recent words of Matt Smith, an American energy resource analyst, during his appearance on the popular podcast “Invest Like the Best.” Having analyzed fossil fuel market trends for decades, he has recently detected an unprecedented level of demand for natural gas. The reason, he says, is the worldwide boom in artificial intelligence (AI) data centers.
Today, among the many options for power generation, data centers are particularly well-suited to gas turbines. With the number of data centers using natural gas as fuel increasing, there is a possibility that ordinary consumers will face a shortage of gas for heating.
AI Boom Drives U.S. Natural Gas Demand to Soar
Drilling facility in Texas, the largest shale oil production area in the United States. Photo by Yonhap News Agency
At Henry Hub, the world’s largest market for natural gas and liquefied natural gas (LNG) trading in the United States, natural gas prices remain in the $2.70s range per million BTUs—a level considered very low compared to the historical average. However, Smith warned that natural gas demand has surged rapidly and that starting in 2028, bottlenecks could drive price increases.
The main driver behind this spike in natural gas demand is data centers built for AI. This was already pointed out by the International Energy Agency (IEA) in a report published last May. The IEA noted that gas turbine generators are proliferating to supply power to AI data centers. As a result, the United States’ annual investment in fossil-fuel power generation could surpass China’s for the first time in decades. The total power consumption from data centers scheduled to be constructed in the U.S. by 2028 is estimated to be between 130 and 150 gigawatts (GW).
The Heart of Today’s Data Centers: Gas-Powered Turbines
Industrial gas turbine from General Electric (GE) installed in an artificial intelligence (AI) data center. CNBC YouTube
Data centers, packed with computer chips and cooling devices for those chips, are already notorious for consuming vast amounts of electricity. Even the leaders of several major AI companies have identified electricity supply—not chips—as the real bottleneck for AI progress. What’s notable is that among all types of power generation equipment, data centers work especially well with industrial gas turbines.
Why is this? First, data centers are required to operate 24 hours a day, 365 days a year, with zero downtime. Renewable energy sources, which fluctuate with weather conditions, can pose significant risks. In contrast, gas turbines offer much higher output than other types of generators and are easily adjustable to changing demand. Above all, gas turbines can reach operating capacity in a very short time—meaning their warm-up period is extremely brief. If there is an issue with the power grid and a data center is cut off, an auxiliary gas turbine can quickly restore stability.
Second, extending the existing power transmission and distribution grid to supply gigawatt-scale data centers is a time-consuming process. It involves passing strict regulatory reviews and addressing local opposition. Relative to this, installing gas turbines is much simpler.
According to the U.S. tech outlet Wired, Texas—where data centers have proliferated rapidly—is now permitting the construction of “stand-alone gas turbine power plants” exclusively serving data centers. Elon Musk, the founder of Tesla, is building the “Colossus” AI data center in Memphis, Tennessee, which includes a dedicated power complex with as many as 69 large gas turbines. Construction was completed in just 122 days—a testament to the pace of such developments.
“A Choice Between AI or Higher Electricity Bills”
Storage facility of Henry Hub, the world’s largest natural gas trading hub located in Louisiana, USA. CME Group website
Some have proposed supplying data centers with power from small modular reactors (SMRs), but SMRs are still an emerging technology that only a handful of countries have managed to prototype. For now, it appears that gas turbines will remain the heart of data center infrastructure in the current AI race—at least through 2030.
If the “AI-induced gas crisis” becomes a reality, its ripple effects could spread worldwide. The United States is both the world’s largest consumer and exporter of natural gas. If Henry Hub gas prices rise, prices for export LNG will increase as well, pushing up global fuel prices altogether. Smith warned, “At the end of the day, it’s consumers who will bear the brunt. When electricity bills go up, we’ll face a choice between using our power for AI computing or for keeping household electricity costs stable. Either way, it’s going to be unpleasant.”
“Not Even Daiso, but Prices from 1,000 to 7,000 Won” Crowds Flock… This Place Becomes Millennial & Gen Z Shopping Hotspot
In U.S. political circles, the AI-driven strain on energy supplies is becoming a hot-button issue. On July 27 (local time), the Trump Administration unveiled what is called the “Ratepayer Protection Pledge.” The gist of the pledge is to prevent household electricity costs from rising due to AI data centers. From now on, big tech firms seeking to build ultra-large data centers will need to invest in and operate their own power plants to supply electricity directly.
This content was produced with the assistance of AI translation services.
© The Asia Business Daily. All rights reserved. Unauthorized AI training and use prohibited.