but AI datacenters do not actually use that much energy compared to other industries,
To counter that statement:
The state that made itself America’s data center capital just realized the data centers want more electricity than five entire Texases could ever produce. Less than a year ago, Governor Greg Abbott declared Texas the epicenter of AI development. On August 3, 2026, he froze all new data center power connections to the Texas grid and called for a full audit of every project in the queue. The reason was a number so large it barely feels real. ERCOT is currently tracking more than 1,800 projects in the queue, representing over 474 gigawatts of electricity, or more than five times the grid’s record for peak demand. Approximately 90 percent of the new power requests are data centers. To put that in perspective, Texas set its all-time electricity demand record in August 2023 at approximately 85.5 gigawatts, during one of the most brutal heat waves in state history.
The fun part of that entire statement is the word “Queue.” Lots of projects WANT to be approved because they will make money. That’s how investing works. Are they all going to be approved? Nope. Obviously there isn’t enough power available, but there aren’t even enough chips to put into that many factories if they were to get approved, Nvidia is only producing about 16-18GW worth of chips per year globally.
As for energy use, one clarification, you’re confusing “Energy” and “Electricity” and those two things are not the same. Electricity is only one part of Energy. The vast majority of energy use in the state is not as electricity, it’s fossil fuels burned directly in industrial processes and in transportation (and specifically excluding the fossil fuels used in power plants). Direct use makes up around 70% of total use, with electricity only making up the remaining 30%. It’s worth keeping that in mind when discussing the global scale of AI.
Canada is thankfully not Texas and is not proposing hundreds of large data centers in dry arid water stressed climate zones. A reason there are so many proposals for datacenters in the desert is to be able to use evaporative cooling which is not applicable in our situation.
To counter that statement:
Source 1
Source 2
Source 3
The fun part of that entire statement is the word “Queue.” Lots of projects WANT to be approved because they will make money. That’s how investing works. Are they all going to be approved? Nope. Obviously there isn’t enough power available, but there aren’t even enough chips to put into that many factories if they were to get approved, Nvidia is only producing about 16-18GW worth of chips per year globally.
As for energy use, one clarification, you’re confusing “Energy” and “Electricity” and those two things are not the same. Electricity is only one part of Energy. The vast majority of energy use in the state is not as electricity, it’s fossil fuels burned directly in industrial processes and in transportation (and specifically excluding the fossil fuels used in power plants). Direct use makes up around 70% of total use, with electricity only making up the remaining 30%. It’s worth keeping that in mind when discussing the global scale of AI.
Canada is thankfully not Texas and is not proposing hundreds of large data centers in dry arid water stressed climate zones. A reason there are so many proposals for datacenters in the desert is to be able to use evaporative cooling which is not applicable in our situation.