DAVID BLACKMON: AI Faces Chicken-And-Egg Conundrum
Which came first, the chicken or the egg? It’s one of the grand conundrums of history, and it is one that is impacting the rapidly expanding AI datacenter industry related to feeding its voracious electricity needs …


Which came first, the chicken or the egg? It’s one of the grand conundrums of history, and it is one that is impacting the rapidly expanding AI datacenter industry related to feeding its voracious electricity needs.
Which comes first, the datacenters or the electricity required to make them go? Without the power, nothing works. It must exist first, or the datacenter won’t go. Without the datacenter, the AI tech doesn’t go, either.
Logic would dictate that datacenter developers who plan to source their power needs with proprietary generation would build it first, before the datacenter is completed. But logic is never simple when billions in capital investment is at risk, along with the need to generate profits as quickly as possible.
Building a power plant is a multi-year project, which itself involves heavy capital investment, and few developers have years to wait. The competition with China to win the race to become the global standard setters in the AI realm is happening now, not in 2027, when a new natural gas plant might be ready to go, or in 2035, the soonest you can reasonably hope to have a new nuclear plant in operation.
Some developers still virtue signal about wind and solar, but the industry’s 99.999% uptime requirement renders them impractical for this role. Besides, with the IRA subsidies on their way out, the economics no longer work.
So, if the datacenter is the chicken in this analogy and the electricity is the egg, real-world considerations dictate that, in most cases, the chicken must come first. That currently leaves many datacenter developers little choice but to force their big demand loads onto the local grid, often straining available capacity and causing utility rates to rise for all customers in the process.



