Tech companies need power solutions that are fast and effective. Mark Zuckerberg summed it up over a year ago, “We would probably build out bigger clusters than we currently can if we could get the energy to do it.”
It just works: These organizations have had a hard time finding the power they need with local grids and only renewable solutions at the scale and speed they require, so they’re turning to ‘ol Reliable: Natural gas.
Moar power: Meta’s Socrates project in Ohio has grown from a $1.6-billion project that will provide 400 megawatts of gas-fired power to a $2-billion project providing 750 megawatts, according to Morgan Stanley analysts.
Unexpected beneficiaries: Companies that stand to benefit from bringing in natural gas include Williams Cos., which is in the natural gas infrastructure business, but also companies like Caterpillar thanks to its gas turbine business.
Turbine diversity: Caterpillar’s turbines are already part of the power team for OpenAI’s Stargate data center project in Abilene, TX, but the type of turbine being used and the type of capacity are important.
Bottleneck alert: There is a gas turbine shortage, the price for gas-fired power plants has tripled, and the lead times can be five years or more.
Specifics matter: The turbine shortage requires context, though, because it is most applicable to large turbines that are used in combined cycle systems for power plants that serve the grid, meaning their scales and power outputs are substantial.
Small, but mighty: Data centers, while requiring vast amounts of power, have specific needs, and the behind-the-meter solutions they are seeking can be smaller, simple-cycle turbines: The turbines for Meta’s Ohio data center range in output from about 14 to 23 megawatts.
Big grid comparison: The GE Vernova 7HA combined-cycle turbines that utility Duke Energy buys, however, range in output from 290 to 430 megawatts.
Fast delivery: Renewable solutions like combined solar and battery banks have been sold as fast and cost-effective solutions, with timelines of 12 to 18 months, but simple-cycle gas turbines are proving to have at least equally fast delivery times.
Building steam: Williams is under contract to install six gigawatts of behind-the-meter power for their combined projects by the first half of 2027, and it took xAI’s Colossus data center in Memphis only six months to go from lease to training its LLMs.
CO2 factories: Efficiency is still a concern, though. BloombergNEF analysts estimate 1,400 pounds of carbon per megawatt-hour for single-cycle turbines, which is nearly double the more than 800 pounds of output by combined cycle.
Viable tradeoff: When data centers need electrons, though, carbon emissions seem like something companies are willing to slate as a “tomorrow” problem, or maybe as an excuse for more work and problems for the AI models to solve.
To learn more about the way AI companies are going “all in” on solving their problem needs, go here.