Annotation by Maya Okafor on How to Make AI Companies Pay Your Energy Bill | Ari Matusiak | TED

The 'on your bill' line at 3:10 skips the denominator. Price per kWh is roughly what the system costs divided by the kWh it sells, and a data center brings a mountain of new kWh.

“So in fact, I think the utilities and the data centers are doing a pretty good job at the local level, when any given new data center comes to town, at making sure that that data center is not going to raise your rates, and in fact should lower your rates a bit, right?”

LDo data centers really increase electricity prices?latitudemedia.com

@priyavenkat that's a lot of weight on 'should'. The Oklahoma case in this piece is an 'illustrative' $1.83 a month off by 2027, and the article says flat out those are projections, not promised reductions. Electric is the line right after payroll, rent and flour at the bakery. I'll believe it when it shows up there.

“Direct testimony filed with the Corporation Commission projects that spreading existing system costs across Google’s additional electricity demand could produce an illustrative residential benefit of about $1.83 a month by 2027.”

OUtilities contend data centers could lead to savings for ratepayersokenergytoday.com

@priyavenkat that's the local half of the episode. the other half is at 21:22: the new costs are global and there's 'basically nothing that your friendly local utility or data center developer can do about' them with a special tariff. then conductor at about 2x, transformers at 2x plus. so @mayaokafor i think the tariff framing undersells it.

@devmalhotra fair, I'd have led with that half, and transformer prices are the line in my budgets that keeps going the wrong way. But if wires cost 2x, his home-upgrade fund looks better on paper, so it gets the same test. At 7:13 it's 50,000 households standing in for a 200 MW data center. That's 4 kW from every house at the peak hour, whenever it's called. He says 'we estimate.' I want to see the estimate.

@mayaokafor your padding point cuts the other way, though. Canary's Ohio story, the same one that has developers shopping projects around, leads with this. Forecasts push up capacity prices and get new lines built, and PJM capacity charges are up more than tenfold since May 2025. Has anyone charted announced vs. energized load? That's the chart I want.

“Many of those planned facilities may never get built, but their projected electricity needs will drive up energy bills for Ohioans anyway.”

CQuestionable data center forecasts are driving up Ohio power billscanarymedia.com

Same Canary piece, last paragraph: Belden, from the Ohio manufacturers' energy group, doesn't want existing customers turned into 'an insurance policy for projects that may never materialize.' Insurance is the right word for it. The risk on load that never shows up doesn't go away, it moves to whoever is already paying bills.

@elliebrennan half agree. The capacity-price half runs through PJM's auction, and no utility's tariff sets that price. The wires half is fixable, and the fix is dull: the large-load customer contracts for the capacity and pays for it whether or not the building ever gets finished. Then the padding is their risk, not the residential class's.

On who writes those tariffs: the House passed the Ratepayer Protection Act 417-3, and per ABC all it does is make states consider federal rules that would have data centers pay for grid upgrades. It doesn't force them, and it's already hit a snag in the Senate, where the objection was that it doesn't go far enough. So it lands back on state utility commissions, whose hearings get about as much coverage as the council meetings I sit through.

The interstate line at 3:16 is about size, but the funding is the more useful comparison. Those roads were built mostly out of federal fuel taxes through the Highway Trust Fund, so drivers paid at the pump, roughly per mile. A user fee. Maya's tariff is that idea for wires. His community fund (8:08) is something else: a subsidy to the neighbors in place of a power plant, not a charge for what the load costs the system.

52 of the 100 biggest projects are from companies that have never built one. 87 GW of expansion teams. I'm taking the under on 270.

“We found that 52 of the largest 100 projects in the US are announced by companies that have never built one before, accounting for 87GW.”

ASix Things to Know About BNEF’s New US Data Center Capacity Outlookabout.bnef.com

@kwameasante pjm isn't taking the under.

““The Ohio Manufacturers’ Association has been saying that, effectively, load growth is not real. I think that is both naive and irresponsible,” said Asim Haque, PJM’s executive vice president for government and member services.”

CQuestionable data center forecasts are driving up Ohio power billscanarymedia.com