I'd say 400 US Data Centers using a fifth to a quarter of the energy of Texas would also not be a "nothingburger". That's a huge amount of power for a manipulative tech whose owners are trying to take over the businesses that use it by convincing them they're "increasing productivity" while destroying the open web.
If the major LLMs were built in good faith, with consensual training sets, no rentals, and built without the constant patting on your head and telling you what a good boy you are, we might be able to have a discussion about whether the benefits are worth spending more than a fifth of a Texas on. But we're nowhere near that.
BTW Michigan is extremely cold half the year, and is one of the US's major industrial centers. Looking here: https://www.eia.gov/electricit... I'm not surprised to see that despite Texas being "bigger", Michigan is actually one of the highest users of electricity, beaten out only by Indiana, Tennessee, Illinois, North Carolina, Virginia, New York, Pennsylvania, Georgia, Ohio, California, Florida, and Texas. Put another way, it's easily in the top 15% of electricity users in the US, with most states using well under Michigan's 98,991,014MWh/year.
I say this because you're acting like it's a trivial figure, it most definitely isn't. Especially when it's being compared to just the top 400 data centers. According to at least one source, there are over 3,400 active running standalone data centers in the US right now. (Michigan has 54 of these, Texas 571...) Indiana and Tennessee both use less than 5% more energy than Michigan. How many more data centers do you include to beat them? Beating Ohio's 153,707,376KWh/yr would take finding another 50-60% in power usage, is it reasonable to suggest it's likely that include all those 3,400 standalone data centers, and they beat Ohio (and by implication Indiana, Tennessee, Illinois, North Carolina, Virginia, New York, and Pennyslvania) Leaving only four states with power usage that's higher than the combined 3,400 data centers?
Clearly without the raw numbers, it's not easy to know, it could be most data centers sip power compared to the top 8%, but it seems credible, and to me I suspect they picked MI as the cut off rather than Indiana or Tennessee because it's a state most people know uses huge amounts of power, rather than because the 401th data center is actually a raspberry pi running off of a solar panel.