How could it possibly be that using more electricity lowers costs?

Also, here is average price for retail electricity over the last 10 years by whether the state voted for Trump or Harris. I think this is some evidence Democratic policies are implicated in higher prices.

Good overview of the intuition. Our recent paper provides causal evidence that data centers lowered electricity prices modestly through 2024 (using an IV approach to account for the siting challenge you raised):

Thank you. I'm ashamed to say I should have cited your paper but did not. I read it yesterday while thinking about this. When recording this morning I tried to explain the intuition but completely failed to acknowledge where I learned it from.

You do a much better job with the explanation than I would! I'm always tempted to bust out this bad boy, which I have been (repeated) told that most people outside of #EconTwitter don't understand.

The push for much more data centers seems to have only happened in the last 2-3 years, so looking at stats for the last 10 years seems a weak argument

Since we can't look at future data (yet) we could instead just look at change in total electricity consumption. If More Perfect Union is right, then more usage should be associated with more costs. Actually, it's the reverse, as my model predicts.

but lower cost only assumes going from, with your example, 50%->80% capacity. What does the model say if data centers start pushing power plants over 100% capacity?

In that case it's possible for costs to increase because everyone will bear the cost of building a new power plant or upgrading one or improving the grid etc. It's also possible for costs to decrease because the newer power plants may be more efficient. As before, it's an empirical question what will happen, not just something you can assert. Also, here are capacity factors by generation type in the US.

I'm pro-AI and agree the water usage argument is BS, but electricity costs are a concern If current power plants can't meet demand of new data centers, prices for Americans will most likely increase in the next few years. It's something that needs to be discussed more

If current plants can't meet demand, then one of two things will happen: 1) developers won't build data centers in that market or 2) the data centers will build onsite capacity to make up the deficit (this is already happening). Another under appreciated development in this debate: grids are starting to mandate that data center developers pay for transmission and related infrastructure for these projects, a cost that historically the utilities would bear (and by extension, all other rate payers, whose costs might increase to subsidize the upgrades). ERCOT (TX's grid authority) is even going so far as to say that data centers may have to pay for grid maintenance, a cost typically shouldered by the entire customer base.

@lthlnkso Ty for response. I think pushing the grid cost onto data center developers is the most obvious solution. If this is in fact happening, and will continue in the future, I see no issue with building as many as we need

Pretty bad take from you to. Expecting these companies to reduce prices since they have become more efficient.

While you seem like you would be a financially responsible person, I’m still questioning this bank’s lending standards to give you a $100 million based on a good credit score.

I generally agree with you. One caveat I would be interested in is that not all data centers are not the same. A data center that holds all my old photos dormant probably consumes electricity differently than a compute hungry AI data center.

Even if a DC requires the building of additional plants, it could still theoretically lower costs, since a lot of energy infrastructure is aging or dated and need plants might be more efficient.

“If there is sufficient capacity” “IF” My man that is the entire point. That “if” invalidates your entire rant.

lol…lmao even😂

example

😂😂😂😂😂😂😂😂😂😂😂😂

There's a little thing called economies of scale

In addition, AI allowed electricity to be distributed for more efficiently from both the plant and distributed energy units (solar etc), which could massively drive down prices especially during peak load time (Utilities consultant here)

Is there data on # of data centers in a regional interconnection and the consumer and wholesale price change over time? Some regions (e.g. PJM) have a capacity shortage, so the economics may not work out like your example. Just curious about this angle

Just wanted to say I'm glad I found your account and hope you continue on with something new once this data center drama is over.

I live in Wv every opportunity to raise the electrical prices Allegheny power takes it also known as first energy Potomac Edison among other names if you know these names you know that's a rotation between Maryland West Virginia Virginia and Pennsylvania To avoid being monopolies

Well, granted, it's a debate, but consider two additional factors against this: Fuel prices increase when power plants needs to buy more to serve the data centers. Real utilization of power plants are usually better than that. If datacenters used the underutilized capacity only, this would be great, but their on-demand nature means they're usually most active when people are - at peak power consumption hours.

The more energy you have to build, the more it's worth it to scale up industry to actually build it, thus, energy becomes cheaper. Building the very first Nuclear Power Plant is very cheap. Building 1000 is way cheaper per unit.

this is nonsense and assumes uniform draw. this is exceedingly basic supply and demand. there’s a maximum output the plant is capable of at a given moment. increased demand means increased prices, allowing for budgeting of power by pricing people out.

The thing is tho. Is that if there is even one instance of a data centers raising the cost of residential energy. Then communities have good enough reason to not want data centers. Cause once the data centers are built commitunites have zero power to do anything about it.

Really interesting. But why wouldn't they keep the rate the same and keep the profit? Or raise it with the expectation that consumers aren't going to blame them, but blame the datacenter?

Finite supply. Increasing complexity curve. Increased demand. Price goes up. This isn’t super hard stuff.

You're a real asset to X.

Yeah but your forgetting profit incentive on the companies end. If sales of power go up profit goes up and they have no incentive to lower prices.

I view data centers like nuclear energy. They provide a lot of economic good, but people are too fearful of the POTENTIAL for downsides to support them even though they are a net good.

Every part of this video is either a logical fallacy or just pure retarded

have you never heard of supply and demand? the cost of fuel for the power stations will go up if more is being used

Here’s my concern. If a data center is on the grid & eventually builds out its own power (like we’re told they are), the utility company will 100% seek (& more than likely receive) a rate increase to make up for the lost demand.

I have not studied this topic. But my intuition says your logic was right at the beginning. Now utilization should be high. I think I saw such news. This means new data centers should raise electricity prices unless new power plants are built in parallel.

The capacity question is also base load vs peak load. A plant may run at 50% capacity throughout the year in order to have enough capacity for the 100 degree days in July. What happens when the base capacity jumps to 75%?

Why does electricity costs more during peak times if using more electricity reduces prices

It also depends on the ratio of grid costs to marginal costs and when they use energy. Someone using large amounts of energy at night is likely to reduce costs for other users etc.

He’s right about it being a debate. His simplistic example while good isn’t taking into account new infrastructure requirements and the scale of the new data centers.

This is rather ridiculous. The question is still that they raise costs. Having a buyer in the market doesn't result in cheaper electricity for the rest. Utilisation as a % has a best efficiency but that is for the generators to balance out. This is a ridiculous video.

I usually really respect you and your perspective but this is not an economically sound argument. Price is determined by supply and demand, not this equation. If they sell more electricity it doesn’t mean they decrease price. That’s ridiculous. This is an embarrassing argument.

Data centers have become the next nuclear power fearmonger. A bunch of idiots and midwits, who don’t know the first thing about the burgeoning tech, start spread misinformation and fear and advocate for regulations and prohibition.

Over time it will always lower costs because increased demand for electricity means suppliers have more incentive to increase capacity. In the short term it could go either way depending on if the demand goes beyond current capacity or not. Very simple stuff.

My biggest complaint about data centers is they are building them in places that the overall noise is hurting people around them and they really should be building them in colder regions to moderate cooling costs. Be it I know there is a trade off in needs for proper internet and power structures that lead to them putting them where they are for less cost now rather than building infrastructure for a better future.