I have been writing all year about the unbelievable size of the AI builtout in terms of spending and corporate capex. This is a great chart to put it into perspective and note that he dates are much more compressed than anything previously.
What still puzzles me is exactly where all the money is going.
This isn’t a buildout like railroads where you need steel mills and enormous amounts of workers. Instead, estimates on some data centers are that 80% of the costs will go straight to Nvidia with a good portion of the rest going to rack components and whatnot. The actual building part — the part that distills into the economy — might be 5% or less of the spending.
That makes me skeptical that some of the historical multiples apply.
On the other hand, these are tremendously energy intensive operations and the building of power generation is a classic broadly-dispersed exense that will flow through factories and raw materials. Some of the numbers around the power generation are truly mind-blowing and what scares me is that I don’t think there is the capacity for it to keep up with data center demand.
This is from ZeroHedge:
How mathematically impossible is the energy math behind the data center rollout? Just Texas alone is facing 474GW of interconnection requests (ERCOT), of which 90% is data centers. Which is why gov Abbott froze rollout of new data centers in Texas.
For context, ERCOT all-time peak demand record is 91.1GW on July 22, 2026, with normal demand between 40 and 80GW.
Of course, that’s peak. In a more realistic scenario, Morgan Stanley sees the entire US short about 80-100GW through 2030, or a deficit of about 100 nuclear power plants.
That sounds like a big problem and there is a very real scenario that if these data centers actually get built on time, they won’t be allowed to connect to the grid for years.
