The old dogs being, in this case, old(er) jet engines and more-or-less purpose-built jet engines.
There is a move afoot to convert commercial aircraft jet engines to produce electricity for AI-centered data centers. The conversion is relatively straightforward: replacing the fuel nozzles to utilize natural gas instead of jet fuel, and replacing the large fan on the front of the flight engine with a much smaller fan that is better suited for power generation.
FTAI has said it expects to be able to deliver about 100 turbines, or 2.5 gigawatts, a year. Boom Supersonic said its goal is to have 4GW of manufacturing capacity or more annually by 2030.
If jet engines can do this—and they can—they also can be used, or ganged together to be used, as electricity generators for localized needs other than AI centers in much the same way small modular reactors are planned for localized electricity needs.
One GW is enough electricity to power a city with a population of 1.8 million people. That works out to enough electricity for towns of 18,000 for each of FTAI’s turbines. They’ll gang together and scale for this, just as they will for AI centers, and just as SMRs will for either purpose.
Two points, which are related:
1/ One is none, two is one. Redundant systems will be required.
2/ At a grid level, distributed systems are arguably more resilient (systemically). They need not be fully meshed, but what network engineers call “meshy” – enough interconnections to preserve the network’s functions in the event of a unit loss of function, or maybe two.
The redundancy also will allow jet engine turbines to be taken offline for maintenance or replacement without interrupting the node’s electricity generation.
Meshy vs fully messed: the one also simplifies the control software with no loss to the grid’s resilience while reducing the likelihood of software bugs in the complexity and increasing the ease of bug fixes in the (relatively) reduced software complexity and size.
Eric Hines