I accept that data centers generate more load for a system. Which will make the overall system need more maintenance, which is something that others paying into the system will also have to support. But, I'm not clear on why this is a hidden cost.
Consider, if people get the new housing developments that they want, that would also add load to the system. This larger energy system will be more expensive to run, which will lead to higher costs. Adding houses would probably be even more expensive in the transmission maintenance costs associated.
Any model you do that tries to prevent this is essentially rent stabilization for early members. And that has a pretty good track record of not being a good idea.
Homes also generate property taxes and sales taxes (from the occupants inside of them). Cities nearly always make money selling to new homes -- low density suburbs are highly profitable for municipalities, even on utilities alone, both initially and over a 40 year time period.
Data Centers do not work like this. They don't generate any new sales taxes, they don't really generate much in the way of new jobs, and they often don't even pay property taxes at all (our biggest data center here, for example, got a sweetheart deal on a massive property tax exemption -- they literally don't have to pay any property tax at all)
Data Centers also don't pay standard price for their power -- they get 'industrial' power rates (locally here, our industrial power rate is much lower than what a home would pay for equivalent kwh usage, even after factoring in transmission differences).
If you just charged equivalent access (if industrial users had to pay to-the-penny exactly the same prices as a residential user, identical transmission fees, identical per-kwh prices, identical time-of-day usage surcharges, etc), it would go a long way to making the data center setup more fair for everyone.
> If you just charged equivalent access (if industrial users had to pay to-the-penny exactly the same prices as a residential user, identical transmission fees, identical per-kwh prices, identical time-of-day usage surcharges, etc), it would go a long way to making the data center setup more fair for everyone.
Why? An industrial feed to a data centre is the company dealing with a single customer contact, to a single location, probably bulk-buying up front. That's very different to serving 500 houses with very variable demand.
Most data centers almost certainly pay property taxes, as well. It is still a deeded plot of land, after all? I'm curious what data center you know of that doesn't have to pay any. I know incentives are common, but they are usually tied to some other growth metric.
I'd go further than common, I don't know any major data center built in the last 6 years that didn't get them. 36 out of 50 US states give away public money to privately owned data centers right now (see https://www.naiop.org/research-and-publications/magazine/202... )
It's not always property taxes (sales tax and/or use tax and/or waived utilities costs are also common). But property taxes are also waived in various cases.
> Most data centers almost certainly pay property taxes, as well. It is still a deeded plot of land, after all?
Not always -- they often waive property taxes too.
I'm told Nevada, West Virginia, and Wyoming all have waivers on property taxes for data centers specifically, and at least 12 other states (including Illinois, New York, Texas, and many others) also waive property taxes for data centers through indirect means.
Locally here, data centers get to skip out on paying 100% of their property taxes for 10 to 20 years. They do this by getting a county to label their property as a 'Renaissance Zone' (more commonly known as an 'Enterprise Zone').
Rules for that also vary. As one example, Connecticut has 'Enterprise Zone' distinction as well, but theirs is only an 80% tax abatement, and only for 5 years.
As you might imagine, this quickly becomes a race to the bottom, on which state is willing to give away the highest amount of public money to these private companies. See https://goodjobsfirst.org/enterprise-zones/ for more details.
I don't know if I can get behind rhetoric like "further than common." It is just common. Period. And you can be opposed to it, just on principle. It would help me oppose if you showed places that were not seeing benefits.
I'm reminded of places like Alabama that would do massive incentives to Mercedes to build a production plant in the state. Again, I know many that opposed on principle, but they were a touch off balance when it came to the benefits that the state was getting from the plants.
And I agree it can be a bit of a race to the bottom, as it were. As things are, there is little to no evidence that that is the case. Building any large industrial building is a large construction project that alone will generate plenty of tax revenue to justify pushing off many taxes. Would it be even better for the localities to not? Yeah, but only if they could still get the build.
doesn't really match what I'm seeing in Chelan County, WA... the county is now in a fight with the port because the port wants to create a Tax Increment Area to capture the taxes from the two data center buildouts... ~$100m over 25 years. they also estimate $150m sales tax revenue and $70m in property tax revenue for those data centers. one relevant stat is that Microsoft will become the biggest taxpayer in the county.
They're getting that rate because of the reduced cost to support their connection to the grid per kWh. It's essentially the cost of the "packaging". If this is resulting in a loss of revenue for the utility, the blame for that falls on the utility for not properly measuring costs.
Is that also factoring in demand charges? I find this gets forgotten in a lot of discussion of industrial rates, when it is often a large factor in the effective $/kwh.
These larger power-generation systems tend to be more efficient than smaller power-generation systems, not less, which should result in a cost decrease, not an increase.
Tennessee (for example) has fairly cheap electricity because the TVA uses a lot of hydroelectric, and since we have a ridiculous amount of rain and violent thunderstorms each year, every decade or two they build another hydroelectric dam and create a new lake, which generates more hydroelectric power (and a moderate increase in tourism/recreation). We don't have buried power lines (excepting in a very few places) but we've got a ton of redundant power substations and multiple transmission paths (because storms). The TVA and Corps of Engineers are kinda hardcore here otherwise the valley would flood about a quarter of the year and be sitting around in the dark for another quarter of the year.
Maintenance of the power transmission lines is paid for by the electrical customer as a part of paying for the electricity itself. This actually scales just fine. If your local electrical utility is not doing it this way, someone needs to explain to them how proper accounting works.
Calling a "hidden cost" is just a convenient way to say "We're making this up because we feel like it's right and we don't intend to show any proof."
Now let's consider a different form of govt/shared charges.
Taxes, why are individual and corporate taxes and structures different?
Both are doing work and generating income?
Why do corporations get to deductions and do so much tax magic that individuals don't?
why don't we charge them both with the same laws, and structures?
Cause corporates generate jobs? Isn't that unfair to the people born individuals...
I don't think anyone is saying there should be stabilization of electricity prices.
But costs for grid improvements for industrial, data center or AI usage should be on the said companies.
They are using that resource to generate a profit.
While people in residential homes are just living their lives, you are comparing cost of essential commodities to production inputs...
We make taxes on poorer people less as well for the same reason.
But yes if someone is saying electricity prices should be stabilized for early consumers that's definitely unfair. But I didn't see or read that here.
The easy way to justify why corporations get to deduct their spending, is it encourages corporations to spend. Something that a for profit company would not necessarily do otherwise. With the obvious note that spend sent to people is taxed as income for that person.
Now, I agree that this gets super odd when people also make an odd "corporations are people" argument.
As for my assertion that this is effectively arguing for price stabilization. The entire thing hinges on the complaint that costs have been going up. Which, of course the price of a good that has increasing demand is going up. I'm not sure how to read that other than an appeal to price stabilization for early consumers.
>Why do corporations get to deductions and do so much tax magic that individuals don't?
Individuals can do deductions if they're a sole proprietorship. The requirement in general is that the expense is used for a business purpose. Buying a laptop for personal consumption isn't tax deductible, but buying it for your consulting business is.
Make the user who requested the extra power pay for it (upfront then monthly cost)? That's what we do in the UK, then if you stop paying the monthly cost for the 'commit' you surrender it and someone else is allowed to use it without the upfront cost (but with the monthly cost). Why is that hard please?
> Consider, if people get the new housing developments that they want, that would also add load to the system.
It's a matter of scale and efficiency. Housing also adds (full-price) customers new taxpayers.
A major problem with housing development is that low density (e.g. single-family zoned suburbs) are net-negative contributors. (But this effect is primarily seen with infrastructure like roads)
> Any model you do that tries to prevent this is essentially rent stabilization for early members.
The problem with datacenters is twofold:
1) It's a demand-shock. Because of the rapid rollout of datacenters, energy production and transmission capacity simply can't scale up. This causes prices to spike locally.
2) 'Industrial' electricity like this tends to pay very cheap rates, leading to residential electricity customers subsidizing them. Especially as such industry tends to be given high tax & other incentives.
> Any model you do that tries to prevent this is essentially rent stabilization for early members. And that has a pretty good track record of not being a good idea.
A question here is whether or not we're in a datacenter construction bubble. (To anyone who says we're not: Nvidia's stock price has soared more than Cisco's during the dotcom bubble.)
If you're in charge of a major electricity company, are you going to sign off on major expansion investments right now, knowing that it will take 5-10 years?
A lot of these datacenters are not owned and operated by Big Tech itself, but instead disposable companies like CoreWeave. If there's a bubble, it'll pop, and these companies will just go bankrupt. The power purchase agreement'll be worthless then.
My understanding, the last time rising electric costs came up, is that a large part of the cost increase was finally paying for the infrastructure maintenance that companies put off for a time. Is that no longer the case?
That is, it is specifically the transmission and distribution network that is driving costs increases. Per the crappy AI answer "Electricity distribution costs now often exceed the cost of electricity generation for U.S. utilities, with transmission and distribution (T&D) costs potentially comprising over half of a customer's energy bill, up from about one-third previously."
I should say, I have no problem thinking data centers may need to pay more. If only because they are probably able to. My assertion is just that it is misleading to pin them as THE reason costs have gone up. It is far more complicated than that.
Generally (but there are exceptions) in the US they have not gotten better.
It's just that datacenter demand has eclipsed the effect of that.
> My assertion is just that it is misleading to pin them as THE reason costs have gone up. It is far more complicated than that.
It's not so much that they are the sole cause, but that their effects (especially on local scales) are 1-2 orders of magnitude larger than all other causes, so people focus on this matter.
Do you have numbers showing that data centers have eclipsed this? Quickly looking at search results on this it seems transmission/delivery is estimated to be half of a consumer's bill. Even the high estimate for increased costs from data centers has the increase at less than a quarter of the bill.
Not nothing. And if we can do things to reduce this, I'm game to do that. But I can't get behind the rhetoric that they are eclipsing other cost drivers.
One particular datapoint to watch is auction prices for production capacity as these are the most immediate market. Those have jumped, with PJM's +800% in 2024 and +22% (22% from 2024->2025, $28.92 -> $329.17 or +1038% total over 2023 -> 2025)
https://www.reuters.com/business/energy/biggest-us-power-gri...
We (Well, the article does it for us as well) can identify that this market jump is driven primarily by increased demand rather than lower capacity, as the amount of capacity sold has increased. (Covering ~50% of the new demand)
These prices do not immediately translate into residential fees, there's a lot of financial engineering in between the auction and customer contracts (F in chat for the counterparty of the hedge), but at the end of the day you can't sell $500 of power for $5 so the increased cost will worm their way through to residential consumers eventually. One particularly annoying thing is that this isn't a consistent amount of time by region; Some people are seeing 3x electricity bills already, for others their power company may have more strongly hedged against price increases.
I confess I don't know how to contextualize this. The article says that PJM claims this should be a 1.5% to 5% increase in bills for folks. And it notes this is the first time capacity has been added in the last four auctions.
Is your contention that we a) not believe this will only have an "at most 5% increase" or b) predict that it will be more dramatic in the future?
I'm also curious where you are getting the 3x electric bills claim. The largest bill jumps I'm aware of were in CA and were largely related to fires. That and general grid upgrades.
I'm in favor of them only if they are extended to renters. I don't know how to do this, but everyone should get a lower cost for their first home, even if it is an apartment. If you want a second home, apartment, campsite, or whatever additional place to dwell I'm fine with you paying a little extra for that luxury, but for your first home we should make it affordable.
Same way I feel about any rules. Worth trying and seeing how they can get the desired outcomes.
I should have said that rent stabilization may work out some day. Just it is not my expectation that it will, as it has a record of not. As such, I would expect some inflationary pressure on utility rates.
Americans are discovering the massive underinvestment in the electrical grid.
When you live in San Francisco, you can experience several power outages during the summer, and it's perfectly normal to have a generator at home.
If you live like me in France, you can go several years without a power outage.
You don't maintain your infrastructure, so you pay ridiculously low costs. And then, you're caught up in reality, here, dominated by data centers. And it's the fault of the GAFAMs ;)
This ignores that the French have massively underinvested in the replacement for the nuclear fleet, and the maintenance of the existing one.
Should we all pretend that half the French nuclear fleet was not offline at the height of the energy crisis? A large portion of the european energy crisis came from the French nuclear power not delivering.
A large renewable buildout should be ongoing in France but wasting money on dead-end subsidies for new built nuclear power seems to be what the French people and politicans wants.
We're seeing a huge number of closures coming in the next decades and the politicians can't even agree on the bonkers large subsidy program for the first 6 EPR2 reactors. Maybe targetting some new power on the grid by 2040? If we're lucky with the construction timeline, at a cool cost of ~20 cents per kWh excluding subsidies.
Americans have to deal with the cost of living near data centres, and everyone on the planet has to deal with global warming. At least the AI companies are providing a free tier that almost anyone with internet can use.
I was looking into energy markets and how they work and it is truly a cluster of moving parts all along the Eastern Interconnect. The question is when is the shoe going to really drop? You can only keep prices going up on an inelastic good before something really bad happens, and this doesn't even touch climate tail risks like heat sagging tx lines across the grid.
Energy markets are deeply complicated markets. They are heavily regulated and have a lot of challenges outside of strictly cost of power. Worth remembering they have safety requirements, uptime, national security implications, truly the backbone of our economic welfare.
Only upside I see out of this huge demand for electricity -- hopefully nuclear will clear the deck on using coal, gas and diesel. If we can build and operate nuclear again we can level our long term cost of power. Combined with renewables its a good combo.
Other than that - power prices will always be derivative to the price of natural gas.
If there is one good thing the admin might actually do, it's gut the incredibly suffocating regulations on nuclear.
Nuclear is only a fraction as expensive as the regulations around it. My company has a division that deals with the nuclear industry, and I really cannot overstate how incredibly intense they are.
Imagine your company issued laptops and required 14 different "live scanner" security apps running with twice daily full system scans. Now be a productive dev on that system. Would you be surprised if every project runs far over budget and far over timeline?
Vogtle is bad example. It was the first of its kind after a long stagnation of nuclear builds (US no longer has the ability to build and operate new nuclear) - that and the regulatory/licensing costs blew through the roof.
Vogtle is the wake-up call needed to get nuclear manufacturing/talent going again. I hear you but I think you are taking the wrong lesson.
I will add -- if indeed prices stay at Vogtle then yes Nuclear is dead. That said there's no way thats the new pricing going forward.
Vogtle is an excellent example - it literally had a negative learning rate from construction of reactor 3 to reactor 4 (Reactor 4 cost more and took longer even though we ostensibly learned from building Reactor 3). The "West" has lost the ability to build mega projects, and until we regain that ability, nuclear is indeed dead. Some minuscule part of the issue is the licensing and regulatory regime, but it's a complete cop-out to lay the entirety of the blame there.
The UK is in the exact same boat with Hinkley C -- initially licensed in 2012 with a budget of £18 billion, construction starting in 2016 and a completion date in 2025. Now we're looking at £50 billion in cost with 'best case' start dates in 2030. All of that to generate electricity at over $0.20/kwh wholesale.
Utilities are dragging their feet on interconnect. Tariffs are squeezing solar supply and probably will impact availability of materials and products (reshoring will take a lot of time if it even makes sense).
Nuclear regulatory environment is overly prescriptive. You need nuclear safety but you also can't use regulatory as a mechanism to shutdown all development - which has been the defacto case for the last 40 odd years.
Agree faster timelines are necessary -- we also need to see if this AI thing is real real or if its in over its skis a little bit. At this particular moment it feels like it might be selling more than it is offering - and thereby the energy demand won't materialize in the way that they assume...
Noting -- that the financial backers for energy products aren't all cash cows like tech companies -- they need to see real return on projects before they put up the necessary large amount of money and public-private partnerships given the risk on putting projects together like that.
Good news is that line replacement is definitely occurring. Especially for higher power lines, the modern conductors that are used for rewiring lines have lower loses and less sag.
I think it’s coming soon. The data center demand is just one aspect. In New York, a combination of half baked policy to go carbon free by 2030, and 100% EV a few years later, combined with insane political decisions (Gov Cuomo decommissioning the nuclear plant that supplied 20% of NYC), and the black swan event of Trump starting a trade war with Canada (Quebec is the best source of electricity) is making electric delivery rates alone increase 30%.
On the gas side, we’re both forcing ratepayers to spend billions on conversion from low to high pressure gas mains AND prohibiting new gas service.
The environmental lobby is so dumb and ineffective - the public will be demanding that coal come back.
The policy is intended to fail. There will be another wave of failed policy under the guise of “getting it right”. Because if you set the dates for 2050/2060 whenever the state is able to convert over, then businesses won’t do anything until 2050/2060. The most effective method forward is going to be continuously failed but enforceable policy to legally push industries and people into cleaner energy.
Agree that Cuomo decommissioning the plant was DUMB.
> On the gas side, we’re both forcing ratepayers to spend billions on conversion from low to high pressure gas mains AND prohibiting new gas service.
On the gas side, Trump's one BBB greatly increased the caps on natural gas exports.
The reality is that the companies extracting NG are not going to be giving it to Americans, they're going to sell it abroad and rake in way more money.
Which would maybe be fine... if we also weren't currently (and severely) artificially limiting the supply of renewables. Um, oops. There's nothing left.
This is a tangent, but one untapped source of energy savings that seems to be invisible to climate activists is Microsoft Windows' constant drain on resources relative to Linux and MacOS. It's shocking how energy inefficient Windows is even when it's doing absolutely nothing noticable for a user.
This seems unlikely to me. Most of these machines are laptops nowadays. While my Linux system idles at 4W, the plug is only capable of going up to 60W. So even if Windows brought it to the point of nearly overheating the power brick (I’m no Windows fan, but I’m pretty skeptical there!), it would only be 1/10’th of a typical 600W window air conditioner.
The untapped energy savings is, IMO, getting people to run their climate control less. We should toughen up a bit, tell our brains that 50F-80F is the comfortable range. (Depending on humidity and your workload).
I compared to air conditioners though, which would also be multiplied by that several hundred million factor (well, maybe the laptops get an additional factor of, like, 4 because a room can fit multiple people with laptops, and you don’t need an air conditioner in every season).
If we’re looking at choices a person can make, every choice is multiplied by millions when applied to the entire population of a country, so the 1W differences are swamped by the equally scaled 10W differences.
It doesn’t make sense to talk about “adding up” unless we define what we’re aggregating over. Residential energy consumption, for sure, is a significant chunk of energy consumption.
If we’re looking at the things we can do to reduce our individual consumption, it absolutely makes sense to prioritize the things which are large relative to our other individual contributions, first.
Fine, but I have already upgraded to the most efficient heat pump, and so I'm at the end of where I can go there. Likewise my insulation cannot be improved without a major remodel (My house is not built to modern standards so there is a lot of improvement, but those require thicker walls and a higher attic ceiling) and so while I'd like to do them they are not possible. What is left is the smaller things.
It's true. There are more Windows desktops used in enterprise environments and households than linux and there are so many insane design choices like making the taskbar clock slider thing be written in react. How many useless clocks does that waste? That is a very easy optimization that can be remediated instead of being lazy and waiting for the next node from Intel or AMD.
Computers use up extremely small amounts of electricity. Most people don't actually know this. Your washing machine is a good ~100 computers. A lightbulb? If it's not LED, that's a few computers. For one lightbulb.
Of course it depends on the computer, but if we're talking laptops or corpo computers it's like 25 - 50 watts. Supercomputers, like those used for AI, are different of course.
I can pretty much guarantee you that the typical Windows desktop machine uses much less energy than the typical Linux desktop machine of similar capabilities. Linux power saving basically never works out of the box unless the user has carefully selected the platform to avoid the Linux kernel's numerous defects. By contrast every computer that comes with Windows has working energy-saving features from the factory because that's how Dell and HP get those Energy Star ratings.
I wholeheartedly agree. Software bloat is a hidden energy drain on corporate America as well as the general public no different than how landlords only leech rents from firms and individuals while providing nothing of value. There is no need to politicize by yelling at climate activists though.
Because it is the exact opposite? Power management and drivers are a shitshow on Linux. When I want my Lenovo (RIP IBM) Thinkpad to last more than 2 hours, I have to boot Windows.
But is is no question: the Apple Silicon chips sip power. If you're looking to minimize watts consumed, a MacBook Air is still on top (or bottom), even if OS X has many pain points too.
There's quite a bit of flim-flam in the video, but the key falsifiable claim is that residential electricity demand is not growing. I am looking at the EIA Monthly Energy Review and it appears that residential electricity grows at ~1% CAGR this century. Which raises the question of whether the bills are now due for past underinvestment in generating capacity.
Great video on how the public is getting screwed on energy deals.
Basically large tech companies have the deep pockets to push up prices at electricity auctions. But why bid in public when you can do those deals in private. That's the first problem. All that needs to be out in the open.
What really irks me is that the market is so manipulated that we can't do anything about it. Think about NEM 3.0 vs 2.0. Putting data centers in their own rate class does make sense as the first step.
>Basically large tech companies have the deep pockets to push up prices at electricity auctions. But why bid in public when you can do those deals in private.
Public utilities can't do the same? Moreover if the implication is that large tech companies are somehow getting great prices at the expense of residential users, what does that mean for the electric generators on the other end of this transaction? Why are they leaving money on the table by selling to large tech companies for cheap?
These companies are regulated and can only charge for the costs they incur plus a flat profit on top of that of 10% or so.
The datacenters give allow them to justify building a lot more capacity to serve them. That increases costs, which means that 10% added for profit is now a bigger number and they can give bigger returns to their shareholders. But those profits are extracted from the existing customers who now see higher bills to cover the costs of expanding capacity to serve the datacenters.
The whole capped profit creates the distortions you illustrate.
The effect has a name: the Averch–Johnson effect, named after the Harvey Averch and Leland Johnson paper: "Behavior of the Firm Under Regulatory Constraint"
In the US people are complaining about electricity like it’s a scarce resource while China has drastically overbuilt energy capacity (solar and nuclear) and gives 0 shits about AI or crypto. One of these approaches seems better thought through.
Right - because their nuclear power plant construction mechanisms are just starting to scale up and they need something to backfill baseload when solar doesn't cut it. Coal growth is still shrinking and will peak in 2030 at which point it will get squeezed out by all other energy sources.
> […] Applying a new data set for country-level energy prices since 1960, this study evaluates the effects of energy prices on economic growth in 18 OECD countries by controlling for other important macroeconomic conditions that shape economic activity. Mean-group estimates that control for cross-country correlations are used to emphasize average responses across nations. Averaged across all nations, results suggest that a 10% increase in energy prices dampened economic growth by about 0.15%. Moreover, some evidence exists that this response may be larger for more energy-intensive economies.
Most water systems have similar deals. Customers will pay a fixed rate infrastructure base fee plus a usage fee, that usage fee for residential is higher than the discounted rate industry receives.
Much like corporations venue shopping municipalities for tax discounts, we need a Federal ban on these corporate subsidies at a Federal level to put us all back on a fair playing field.
Actually it's because there isn't enough solar and generation because of previous policies. Check out how Alex Albrecht got screwed doing solar in Los Angeles.
Yes it does. Prices are based off supply and demand. Demand is increasing but the government is making it prohibitively costly to expand supply so prices are going up. The solution is to generate more electricity, not stifle demand.
"Loudoun County is home to one of the largest concentrations of data centers in the world through which much of the world's internet traffic passes every day. The industry has helped to diversify Loudoun County's economy, which has included bringing new jobs and revenue streams to the county over the past 14 years that have helped to keep real property tax rates relatively low."
In the same county, maybe.
But electricity is a tradable commodity. The electricity price goes up for the neighbouring counties without the tax benefits.
The energy costs of certain parts of Maryland are about to skyrocket, or have already risen dramatically, driven by datacenter demand, primarily but not exclusively from Northern VA.
I am an engineer at a large utility in the Pacific Northwest. My ratepayers pay very little for power, We generate with hydro which is very, very cheap. There are very few opportunities for more hydro. Every new MW we add is going to cost two orders of magnitude more per MWh compared with the existing stuff. We would not need to add capacity like this, except that we are being absolutely deluged with requests for enormous new data centers, which will require purchasing power from these very expensive sources. This will drive up prices for everyone. There's a lot of people in this thread that are confidently claiming things will happen in systems they barely understand. I'm here to tell you this will create a power affordability crisis in my region, and it hasn't even started yet. The effects won't be fully seen for years.
I feel like the solution is for the state owned power company to add capacity to scoop up the additional revenue. This is quite literally an opportunity to advance society while generating more tax revenue and *improving* public services.
Blaming tech for using power is regressive, the real issue here is bad energy policy.
> add capacity to scoop up the additional revenue.
they are; but Big Tech isn't paying the cost to add that capacity, because it's getting volume discount prices to attract them in the first place. So the utility has to raise prices on everyone else to cover the cost of adding the capacity.
so essentially, all the other consumers are subsidizing the extra power needed for the data centers (instead of the data centers paying for it)
Adding capacity isn't free either, in particular on environmental costs and/or additional risks. We haven't been trying to reduce consumption for a few decades now just to be cute, the tradeoff just wasn't worth it.
Maybe it is? The US has ample land and money to build out clean energy, there’s zero excuse not to be adding capacity.
It’s literally the governments job to build out the infrastructure for society to thrive. Don’t blame technology for getting better, blame the government for not doing their job.
The money part doesn't ring true given the current US political climate. That's not the president you elect when your future looks bright.
> society to thrive
Does society thrive on AI though ? And is the technology better ?
Those are real questions. My answers would be no, we have better use for our resources. If we need to sacrifice for something, at least be it worth it.
Clear example of privatization of everything... until its the big corporations who need to get something done - then its socialism in disguise. And of course they hide all the deals behind closed doors so their customers can't see who all, and how much they're subsidizing.
If there is a sudden increase in demand/supply issues due to large corporations they should be the ones to subsidize the bill. I'd argue in the communities this affects they should be forced to cover for everyone who lives in the area for materially making their lives worse.
Not to mention utility companies should be forced to be transparent in the pricing/deals they make with tech companies. They are likely just luring them in with a deal and passing off the rest to their customers so they still don't lose out in the short term. They want the best of both worlds - still making all the same profits short term by passing on to customers while getting the long term lock-in for these data centers.
Clarity on the wording here. They should be the ones _paying_ the bill, not subsidising it. It's not subsidising if they're the consumer of the marginal product.
If an industrial customer pays for the entire power plant, but then also has to outbid everyone in 13 states to access the energy it paid for, then the industrial customer has clearly subsidized it.
It's another variation of what all the MBAs are trained to do: build your fortune using someone else's money. The difference here, of course, is that that money is being used without consent.
I really dislike this kind of rhetoric. This has nothing to do with socialism. Corporations profiting from externalities and pushing costs onto regular workers is just capitalism. If you have a problem with it, maybe you have a problem with the inevitable concentrations of wealth and power which result from capitalism.
Well, what if the corporations would pay the bills - then they would increase the price of the product so the people would pay for it again. Now you can say that you do not use Metas or Googles products, but their business models is ads - and you DO pay the products advertised for.
In scenario 1, a corporation externalizes some of its costs. Those costs are then paid by people who may or may not actually use the corporation's product — people who never chose to be part of any transaction. This is coercive because the people paying for the corporation's externalities are forced to: they may not use the product, or do so to different degrees not proportional to the price they pay for the externality.
In scenario 2, the corporation does not externalize costs and raises their prices, offsetting costs by passing them on to their customers. The people paying the additional cost are those who know the price of what they are buying and willingly engage in the transaction for the good or service.
Do you understand why scenario 2 is bad and scenario 1 is less bad?
That is just a very simplified and incomplete model. I never owned a car, so should I advocate to stop all fundings for streets? Well I consume products build by other people that use cars to go to work. Now, if I don't consume drinks of the Coca Cola Company, what if my cleaning lady enjoys those in her break? Direct vs. Indirect is not a good measure of value, PRICE is.
> I never owned a car, so should I advocate to stop all fundings for streets?
Streets are generally paid for by taxes, which are categorically different than corporate profits. In theory taxes are under democratic control. If you don't want to pay for streets you don't use, you can vote for a politician who passes that law. You have no control over the governance of a private corporation, but it can still pass its costs on to you via externalities (in the absence of regulations preventing it from doing so).
> Now, if I don't consume drinks of the Coca Cola Company, what if my cleaning lady enjoys those in her break?
What are you even talking about? What is the externality here? The wages you presumably pay your cleaning lady are hers to do with as she wishes.
> don't want to pay for streets you don't use, you can vote for a politician who passes that law
Just as you can build your own power plant (solar, wind) if you don't like the electricity prices of your provider... That works in theory, in practice you will have to pay...
That's a strange definition of capitalism you're using.
Most people use capitalism to describe a system where people trade goods and services with as little interference from government as possible.
In this case, the government has written laws in a way that indirectly transfers wealth from consumers to large corporations.
Saying that such laws are "inevitable" in any conceivable capitalist system is unfalsifiable and adds little to the discussion.
I think calling it "crony capitalism" to make it clear that it's the undue influence of capital on government specifically causing the problem here lends more clarity to the discussion.
> Most people use capitalism to describe a system where people trade goods and services with as little interference from government as possible.
This is not a real definition. Saying "most people" use the term to mean what you want it to mean in this argument is ridiculous.
Capitalism is a system of private ownership of capital. We live under that system. Anything you see that happens now is the result of that system because it's the one that exists.
There is no such thing as "crony capitalism". What came before crony capitalism? Was it regular capitalism? Did regular capitalism turn into crony capitalism? Or — more likely — is it all one continuous process and system of accumulation?
> Most people use capitalism to describe a system where people trade goods and services with as little interference from government as possible.
“Capitalism” is a term coined for the dominant system of the industrialized West in the mid-19th Century and is defined by the specific orientation of property rights in that system around the private and marketable ownership of the non-financial means of production (the “capital” in “capitalism”), and the way in which the system was fundamentally structured around—and institutions within it, including government, invariably served the interests of—the owners of that capital, who formed its ruling class, displacing the landed hereditary aristocracy of the preceding systems.
While the dominant politico-economic systems of the developed world have evolved somewhat since then, with modern mixed economies having structures in place mitigate some of the adverse impacts the original system for which thr label “capitalism” was created for has on the vast majority of the population that is not major capital owners, it retains the basic property structure and resulting class heirarchy of the original “capitalism”, and neither it nor the original tended tof eatite annabsence of regulation of commerce.
> I think calling it "crony capitalism" to make it clear that it's the undue influence of capital on government specifically causing the problem here lends more clarity to the discussion.
The commanding and undue influence of capital on all of society, government and otherwise, is literally the feature for which critics of the then-dominant system coined the term “capitalism” to refer to that system. It doesn’t need an extra qualifier for that.
This video is unconvincing. The thesis of the video basically boils down to "AI companies are using electricity, driving up prices, and residential consumers are paying for it". However it neglects that most of the power usage is caused by the same residential users, who are getting something out of it. 80–90% of AI energy usage is estimated for inference (eg. generating responses on chatgpt), not training[1]. Moreover despite AI companies being unprofitable as a whole, they're making fat margins on inference[2]. Therefore it's reasonable to assume that most of AI electricity consumption that the video complains about is as a result of ordinary people using AI. Taking this into account, blaming AI companies for jacking up electricity prices in this context makes as much sense as blaming airlines for jacking up oil prices. They're only doing so because people are buying their services, and presumably deriving some sort of utility from it.
You are not taking into account the fact that data centers tend to be pretty concentrated geographically.
If I live in an area that is trying to attract data centers by giving them sweetheart deals, my electricity bill is going to go up by more than my share in inference usage since those data centers will be used by people all over the globe.
Worldwide, maybe. But in the "western world"? Couple weeks ago my mother asked me to compare the main LLM providers. And then for tips how to best use it. She is over 80.
And there was a letter in the local paper this week about someone complaining they councils park in town because they didn’t have a debit card and they certainly don’t have any “app phones”, a far more common approach amongst older people
How is this different than any other form of industry?
If I live in an area that is trying to attract [steel mills] by giving them sweetheart deals, my electricity bill is going to go up by more than my share in [steel] usage since [that steel] will be used by people all over the globe.
I think if you found the average person on the street and asked them if their electric bill would go up when a steel mill opens nearby they'd only say yes because why else are you asking them.
This sounds like a generic NIMBY argument to me? eg. "why should we build a datacenter in my backyard, when it'll raise MY electricity prices?" Of course, if everyone has this attitude, no datecenters, power plants, homeless shelters will ever be built
I am very YIMBY, but it's totally reasonable that you don't want industrial work done in your backyard. This is exactly what zoning and city planning were designed for.
It seems odd to me to try to redefine Y/NIMBY depending on the individual project. I wouldn’t want a sewage plant or a datacenter in my neighborhood, absolutely no one would. Why would one be NIMBY and one not?
Very anecdotal, but a friend of mine works for a fuel cell company that basically converts natural gas and/or propane to grid electricity a huge part of their work right now is powering new datacenters for AI growth. I don't really know the technology, so I may have the details wrong, but the fact companies willing to take on cost and complexity does hint towards real limits on energy supply.
> Taking this into account, blaming AI companies for jacking up electricity prices in this context makes as much sense as blaming airlines for jacking up oil prices.
I'm not sure if "blame" and "jacking up" are the right words here. AI technology is demanding power quicker than it is being added to the grid and add so there is going to be tension. Now if tech companies start buying the grid operators and power, I'm more inclined to start blaming the tech companies themselves, but for now I still place broad energy failures squarely with the government.
in Chelan County where I live they came up with a Large Load policy specifically to protect the rates of other local users... the large load customer gets short-term access to power and then has to source it elsewhere (with the PUD in the middle). Microsoft is here and apparently plans to source from Helion long-term which is building their first generator just down the street. they also have to pay for their own grid upgrades as needed.
The best way to make the grid fair is to remove fixed rate accounting altogether. The reality is that all power must be delivered instantaneously. Any meaningful arbitrage (i.e., adding actual value to grid services) must occur within milliseconds. The ability to purchase contracts on power that hasn't been generated yet is what is driving all the volatility and exploitation.
There is a reason that Griddy (variable rate wholesaler for consumers) had to be killed in Texas. The popular narrative is that they "fucked" their customers over in 2021. The reality is that anyone who had been using their variable rate system for more than 12 months (and made any attempt whatsoever to respond to demand signals) had easily compensated for the $9/kWH crunch that lasted merely ~72 hours. Most of their other customers who I talked to had no problem with the situation. There was a deliberate PR campaign (likely by the other retail electricity providers in the market) to make it seem like an unbelievably impossible situation for a normal person to cope with. Now it's illegal to have wholesale rates for consumers in Texas. We've gone two steps in the wrong direction.
> The ability to purchase contracts on power that hasn't been generated yet is what is driving all the volatility and exploitation.
That is completely opposite of the truth. The ability to buy a contract is what allows power generators to get loans to build at good rates. Everyone wants someone else to take on the risk, when you buy a contract for power at some rate in the future you are taking on the risk that prices could go down. The generator then goes to the bank and can show that even if competition forces prices down they still are going to get that much from the contract and thus can pay off the loans.
Power is complex and I just touched the surface of the truth.
Ah now I see why electric costs are increasing at double the rate of inflation. But hey I got to make some images of the Muppets in various movies using AI.
Unless the big data centers are entering into mandatory load reduction during high strain on the system, then theres going to be an increased cost to the little buyers.
The tradeoff is that Utilities can upgrade their systems and improve uptime thanks to the higher base load. It's up to the utils to negotiate well against the big data players.. good luck on that with politicians breathing down your neck to secure a big win.
GW is already a measure of rate of energy use. You can talk about GWh per day (which is really just another way of saying 0.041 GW), but GW per day is only sensible in the context of a ramp in power consumption.
Assuming the original source of that number didn't incorrectly conflate GWh into GW, that puts your comparison as underestimating the energy usage 24-fold.
I've heard 5GW touted as the capacity (so like 120GWH per day if flat loaded at 100%). However that is clearly B.S as I've also heard 2GW and another much more sensible number
Just to give perspective 5GW would power about 10 Million H100s or about 5 million G200. Completely farcical, won't happen.
There's a 7-million-sqft data center near Reno that claims 650MW capacity. Zuck says his will be 4 million sqft. So that doesn't really get us to 5GW unless we're going vertical.
> Yet increasing the power consumption of an area by 25X over a few years requires a significant amount of transmission and distribution infrastructure upgrades to make that happen
Not if you put your data-center right next to the power station.
Like any other industry, I don't care what the electricity cost of data centres is if they create actual value. This is no different than a factory that makes car parts - you need energy to make stuff.
I think the part the bugs me is how over-hyped AI is, and there's a fear that all this electricity is being used as part of fundamentally unsustainable business models that will not prove sustainable long term. If that's the case, and we're overbuilding data centres what's going to happen to those buildings once the bubble pops?
Because of Trump, and just like Trump, this is a perfect storm. Unrestrained AI and crypto are causing a surge in demand at the same time Trump is spitefully and recklessly killing wind and solar installations and subsidies.
Trump-licans - the party of mean and stupid and self-destruction.
Some of us call this "capitalism". Prices are driven by offer and demand, efficient market allocation to who can extract most value from a resource, we like that on HN.
Prices are actually driven by value. How much do you want the thing? Do data centers value power? I would think so, since they cannot operate without it. Utilities should be charging them a premium.
How do you define what is an essential good here? For electricity I would place most residential usage near the bottom in priority and definitely below a datacenter. People complaining about facing a tradeoff between setting the thermostat a couple degrees higher this summer or facing a higher power bill just doesn't get much sympathy from me.
The real reason is that fossil fuel plants have been offlined and decommissioned, and the reason this occurred is because they were made unprofitable by punitive regulation far more so than market forces.
1. The bill shown for July 2024 was for about $100, which includes $65 of supply charges.
2. The bill for July 2025 was for about $129, with supply charges of $84.
3. Despite actual usage being lower for July 2025, the total cost increased by 29%.
The point being raised is that energy costs are higher for individuals because the cost of building infrastructure to support data centers is being subsidized by individuals. Even if her bill was for $65, it is inconsequential to the main point. I don’t understand why people make comments like yours other than to muddy the waters and obfuscate the issue. Or truly failing to understand what is being communicated due to some bias or ideology. A sad microcosm of what passes for discourse in our time.
Does anyone here care though? Everyone using LLMs are directly supporting lower quality of life and high cost-of-living across the country.
No person can do anything except fund Zuckerberg's next island purchase and laugh about how they’re using LLMs to vibe code their next unreleased hobby app.
It is truly sad that we are selling out our quality of life for short term gain for tech corporations.
This feels misleading to me.
I accept that data centers generate more load for a system. Which will make the overall system need more maintenance, which is something that others paying into the system will also have to support. But, I'm not clear on why this is a hidden cost.
Consider, if people get the new housing developments that they want, that would also add load to the system. This larger energy system will be more expensive to run, which will lead to higher costs. Adding houses would probably be even more expensive in the transmission maintenance costs associated.
Any model you do that tries to prevent this is essentially rent stabilization for early members. And that has a pretty good track record of not being a good idea.
Homes also generate property taxes and sales taxes (from the occupants inside of them). Cities nearly always make money selling to new homes -- low density suburbs are highly profitable for municipalities, even on utilities alone, both initially and over a 40 year time period.
Data Centers do not work like this. They don't generate any new sales taxes, they don't really generate much in the way of new jobs, and they often don't even pay property taxes at all (our biggest data center here, for example, got a sweetheart deal on a massive property tax exemption -- they literally don't have to pay any property tax at all)
Data Centers also don't pay standard price for their power -- they get 'industrial' power rates (locally here, our industrial power rate is much lower than what a home would pay for equivalent kwh usage, even after factoring in transmission differences).
If you just charged equivalent access (if industrial users had to pay to-the-penny exactly the same prices as a residential user, identical transmission fees, identical per-kwh prices, identical time-of-day usage surcharges, etc), it would go a long way to making the data center setup more fair for everyone.
> If you just charged equivalent access (if industrial users had to pay to-the-penny exactly the same prices as a residential user, identical transmission fees, identical per-kwh prices, identical time-of-day usage surcharges, etc), it would go a long way to making the data center setup more fair for everyone.
Why? An industrial feed to a data centre is the company dealing with a single customer contact, to a single location, probably bulk-buying up front. That's very different to serving 500 houses with very variable demand.
Most data centers almost certainly pay property taxes, as well. It is still a deeded plot of land, after all? I'm curious what data center you know of that doesn't have to pay any. I know incentives are common, but they are usually tied to some other growth metric.
> I know incentives are common
I'd go further than common, I don't know any major data center built in the last 6 years that didn't get them. 36 out of 50 US states give away public money to privately owned data centers right now (see https://www.naiop.org/research-and-publications/magazine/202... )
It's not always property taxes (sales tax and/or use tax and/or waived utilities costs are also common). But property taxes are also waived in various cases.
> Most data centers almost certainly pay property taxes, as well. It is still a deeded plot of land, after all?
Not always -- they often waive property taxes too.
I'm told Nevada, West Virginia, and Wyoming all have waivers on property taxes for data centers specifically, and at least 12 other states (including Illinois, New York, Texas, and many others) also waive property taxes for data centers through indirect means.
Locally here, data centers get to skip out on paying 100% of their property taxes for 10 to 20 years. They do this by getting a county to label their property as a 'Renaissance Zone' (more commonly known as an 'Enterprise Zone').
Rules for that also vary. As one example, Connecticut has 'Enterprise Zone' distinction as well, but theirs is only an 80% tax abatement, and only for 5 years.
As you might imagine, this quickly becomes a race to the bottom, on which state is willing to give away the highest amount of public money to these private companies. See https://goodjobsfirst.org/enterprise-zones/ for more details.
I don't know if I can get behind rhetoric like "further than common." It is just common. Period. And you can be opposed to it, just on principle. It would help me oppose if you showed places that were not seeing benefits.
I'm reminded of places like Alabama that would do massive incentives to Mercedes to build a production plant in the state. Again, I know many that opposed on principle, but they were a touch off balance when it came to the benefits that the state was getting from the plants.
And I agree it can be a bit of a race to the bottom, as it were. As things are, there is little to no evidence that that is the case. Building any large industrial building is a large construction project that alone will generate plenty of tax revenue to justify pushing off many taxes. Would it be even better for the localities to not? Yeah, but only if they could still get the build.
doesn't really match what I'm seeing in Chelan County, WA... the county is now in a fight with the port because the port wants to create a Tax Increment Area to capture the taxes from the two data center buildouts... ~$100m over 25 years. they also estimate $150m sales tax revenue and $70m in property tax revenue for those data centers. one relevant stat is that Microsoft will become the biggest taxpayer in the county.
This is wrong, the data center supply chain is very large with hundreds of thousands of components all generating tax.
How much of that tax goes to the city or county that the data center is in?
A recurring use of energy generates tax.
They're getting that rate because of the reduced cost to support their connection to the grid per kWh. It's essentially the cost of the "packaging". If this is resulting in a loss of revenue for the utility, the blame for that falls on the utility for not properly measuring costs.
Is that also factoring in demand charges? I find this gets forgotten in a lot of discussion of industrial rates, when it is often a large factor in the effective $/kwh.
These larger power-generation systems tend to be more efficient than smaller power-generation systems, not less, which should result in a cost decrease, not an increase.
Tennessee (for example) has fairly cheap electricity because the TVA uses a lot of hydroelectric, and since we have a ridiculous amount of rain and violent thunderstorms each year, every decade or two they build another hydroelectric dam and create a new lake, which generates more hydroelectric power (and a moderate increase in tourism/recreation). We don't have buried power lines (excepting in a very few places) but we've got a ton of redundant power substations and multiple transmission paths (because storms). The TVA and Corps of Engineers are kinda hardcore here otherwise the valley would flood about a quarter of the year and be sitting around in the dark for another quarter of the year.
Maintenance of the power transmission lines is paid for by the electrical customer as a part of paying for the electricity itself. This actually scales just fine. If your local electrical utility is not doing it this way, someone needs to explain to them how proper accounting works.
Calling a "hidden cost" is just a convenient way to say "We're making this up because we feel like it's right and we don't intend to show any proof."
Now let's consider a different form of govt/shared charges.
Taxes, why are individual and corporate taxes and structures different?
Both are doing work and generating income?
Why do corporations get to deductions and do so much tax magic that individuals don't?
why don't we charge them both with the same laws, and structures?
Cause corporates generate jobs? Isn't that unfair to the people born individuals...
I don't think anyone is saying there should be stabilization of electricity prices. But costs for grid improvements for industrial, data center or AI usage should be on the said companies.
They are using that resource to generate a profit.
While people in residential homes are just living their lives, you are comparing cost of essential commodities to production inputs...
We make taxes on poorer people less as well for the same reason.
But yes if someone is saying electricity prices should be stabilized for early consumers that's definitely unfair. But I didn't see or read that here.
Apologies, I misread some of this yesterday.
The easy way to justify why corporations get to deduct their spending, is it encourages corporations to spend. Something that a for profit company would not necessarily do otherwise. With the obvious note that spend sent to people is taxed as income for that person.
Now, I agree that this gets super odd when people also make an odd "corporations are people" argument.
As for my assertion that this is effectively arguing for price stabilization. The entire thing hinges on the complaint that costs have been going up. Which, of course the price of a good that has increasing demand is going up. I'm not sure how to read that other than an appeal to price stabilization for early consumers.
>Why do corporations get to deductions and do so much tax magic that individuals don't?
Individuals can do deductions if they're a sole proprietorship. The requirement in general is that the expense is used for a business purpose. Buying a laptop for personal consumption isn't tax deductible, but buying it for your consulting business is.
Isn't that proving my point you change laws based on what you do with your money and where you get it from.
If so data centers should also be separately booked for their electricity consumption.
Make the user who requested the extra power pay for it (upfront then monthly cost)? That's what we do in the UK, then if you stop paying the monthly cost for the 'commit' you surrender it and someone else is allowed to use it without the upfront cost (but with the monthly cost). Why is that hard please?
TBH, most articles like this read like industry propaganda intended to promote rate increases.
It seems to work.
In most of the US new housing has to pay (sometimes large) “impact fees” to cover infrastructure upgrades.
> Consider, if people get the new housing developments that they want, that would also add load to the system.
It's a matter of scale and efficiency. Housing also adds (full-price) customers new taxpayers. A major problem with housing development is that low density (e.g. single-family zoned suburbs) are net-negative contributors. (But this effect is primarily seen with infrastructure like roads)
> Any model you do that tries to prevent this is essentially rent stabilization for early members.
The problem with datacenters is twofold:
1) It's a demand-shock. Because of the rapid rollout of datacenters, energy production and transmission capacity simply can't scale up. This causes prices to spike locally.
2) 'Industrial' electricity like this tends to pay very cheap rates, leading to residential electricity customers subsidizing them. Especially as such industry tends to be given high tax & other incentives.
> Any model you do that tries to prevent this is essentially rent stabilization for early members. And that has a pretty good track record of not being a good idea.
A question here is whether or not we're in a datacenter construction bubble. (To anyone who says we're not: Nvidia's stock price has soared more than Cisco's during the dotcom bubble.)
If you're in charge of a major electricity company, are you going to sign off on major expansion investments right now, knowing that it will take 5-10 years?
A lot of these datacenters are not owned and operated by Big Tech itself, but instead disposable companies like CoreWeave. If there's a bubble, it'll pop, and these companies will just go bankrupt. The power purchase agreement'll be worthless then.
My understanding, the last time rising electric costs came up, is that a large part of the cost increase was finally paying for the infrastructure maintenance that companies put off for a time. Is that no longer the case?
That is, it is specifically the transmission and distribution network that is driving costs increases. Per the crappy AI answer "Electricity distribution costs now often exceed the cost of electricity generation for U.S. utilities, with transmission and distribution (T&D) costs potentially comprising over half of a customer's energy bill, up from about one-third previously."
I should say, I have no problem thinking data centers may need to pay more. If only because they are probably able to. My assertion is just that it is misleading to pin them as THE reason costs have gone up. It is far more complicated than that.
> Is that no longer the case?
Generally (but there are exceptions) in the US they have not gotten better. It's just that datacenter demand has eclipsed the effect of that.
> My assertion is just that it is misleading to pin them as THE reason costs have gone up. It is far more complicated than that.
It's not so much that they are the sole cause, but that their effects (especially on local scales) are 1-2 orders of magnitude larger than all other causes, so people focus on this matter.
Do you have numbers showing that data centers have eclipsed this? Quickly looking at search results on this it seems transmission/delivery is estimated to be half of a consumer's bill. Even the high estimate for increased costs from data centers has the increase at less than a quarter of the bill.
Not nothing. And if we can do things to reduce this, I'm game to do that. But I can't get behind the rhetoric that they are eclipsing other cost drivers.
One particular datapoint to watch is auction prices for production capacity as these are the most immediate market. Those have jumped, with PJM's +800% in 2024 and +22% (22% from 2024->2025, $28.92 -> $329.17 or +1038% total over 2023 -> 2025) https://www.reuters.com/business/energy/biggest-us-power-gri...
We (Well, the article does it for us as well) can identify that this market jump is driven primarily by increased demand rather than lower capacity, as the amount of capacity sold has increased. (Covering ~50% of the new demand)
These prices do not immediately translate into residential fees, there's a lot of financial engineering in between the auction and customer contracts (F in chat for the counterparty of the hedge), but at the end of the day you can't sell $500 of power for $5 so the increased cost will worm their way through to residential consumers eventually. One particularly annoying thing is that this isn't a consistent amount of time by region; Some people are seeing 3x electricity bills already, for others their power company may have more strongly hedged against price increases.
I confess I don't know how to contextualize this. The article says that PJM claims this should be a 1.5% to 5% increase in bills for folks. And it notes this is the first time capacity has been added in the last four auctions.
Is your contention that we a) not believe this will only have an "at most 5% increase" or b) predict that it will be more dramatic in the future?
I'm also curious where you are getting the 3x electric bills claim. The largest bill jumps I'm aware of were in CA and were largely related to fires. That and general grid upgrades.
Sounds more like externalized cost that data centers should pay more for but it is spread evenly across all customers.
How do you feel about homestead exemptions in general?
I'm in favor of them only if they are extended to renters. I don't know how to do this, but everyone should get a lower cost for their first home, even if it is an apartment. If you want a second home, apartment, campsite, or whatever additional place to dwell I'm fine with you paying a little extra for that luxury, but for your first home we should make it affordable.
Same way I feel about any rules. Worth trying and seeing how they can get the desired outcomes.
I should have said that rent stabilization may work out some day. Just it is not my expectation that it will, as it has a record of not. As such, I would expect some inflationary pressure on utility rates.
[dead]
Americans are discovering the massive underinvestment in the electrical grid. When you live in San Francisco, you can experience several power outages during the summer, and it's perfectly normal to have a generator at home. If you live like me in France, you can go several years without a power outage. You don't maintain your infrastructure, so you pay ridiculously low costs. And then, you're caught up in reality, here, dominated by data centers. And it's the fault of the GAFAMs ;)
This ignores that the French have massively underinvested in the replacement for the nuclear fleet, and the maintenance of the existing one.
Should we all pretend that half the French nuclear fleet was not offline at the height of the energy crisis? A large portion of the european energy crisis came from the French nuclear power not delivering.
A large renewable buildout should be ongoing in France but wasting money on dead-end subsidies for new built nuclear power seems to be what the French people and politicans wants.
We're seeing a huge number of closures coming in the next decades and the politicians can't even agree on the bonkers large subsidy program for the first 6 EPR2 reactors. Maybe targetting some new power on the grid by 2040? If we're lucky with the construction timeline, at a cool cost of ~20 cents per kWh excluding subsidies.
Americans have to deal with the cost of living near data centres, and everyone on the planet has to deal with global warming. At least the AI companies are providing a free tier that almost anyone with internet can use.
I was looking into energy markets and how they work and it is truly a cluster of moving parts all along the Eastern Interconnect. The question is when is the shoe going to really drop? You can only keep prices going up on an inelastic good before something really bad happens, and this doesn't even touch climate tail risks like heat sagging tx lines across the grid.
Energy markets are deeply complicated markets. They are heavily regulated and have a lot of challenges outside of strictly cost of power. Worth remembering they have safety requirements, uptime, national security implications, truly the backbone of our economic welfare.
Only upside I see out of this huge demand for electricity -- hopefully nuclear will clear the deck on using coal, gas and diesel. If we can build and operate nuclear again we can level our long term cost of power. Combined with renewables its a good combo.
Other than that - power prices will always be derivative to the price of natural gas.
After looking at Vogtle 3/4 I highly doubt any admin including this one is willing to bear the cost.
If there is one good thing the admin might actually do, it's gut the incredibly suffocating regulations on nuclear.
Nuclear is only a fraction as expensive as the regulations around it. My company has a division that deals with the nuclear industry, and I really cannot overstate how incredibly intense they are.
Imagine your company issued laptops and required 14 different "live scanner" security apps running with twice daily full system scans. Now be a productive dev on that system. Would you be surprised if every project runs far over budget and far over timeline?
Vogtle is bad example. It was the first of its kind after a long stagnation of nuclear builds (US no longer has the ability to build and operate new nuclear) - that and the regulatory/licensing costs blew through the roof.
Vogtle is the wake-up call needed to get nuclear manufacturing/talent going again. I hear you but I think you are taking the wrong lesson.
I will add -- if indeed prices stay at Vogtle then yes Nuclear is dead. That said there's no way thats the new pricing going forward.
Vogtle is an excellent example - it literally had a negative learning rate from construction of reactor 3 to reactor 4 (Reactor 4 cost more and took longer even though we ostensibly learned from building Reactor 3). The "West" has lost the ability to build mega projects, and until we regain that ability, nuclear is indeed dead. Some minuscule part of the issue is the licensing and regulatory regime, but it's a complete cop-out to lay the entirety of the blame there.
The UK is in the exact same boat with Hinkley C -- initially licensed in 2012 with a budget of £18 billion, construction starting in 2016 and a completion date in 2025. Now we're looking at £50 billion in cost with 'best case' start dates in 2030. All of that to generate electricity at over $0.20/kwh wholesale.
I agree with you, nuclear and renewables are the future but surely there has to be a better way to speed these deployments to production right?
Utilities are dragging their feet on interconnect. Tariffs are squeezing solar supply and probably will impact availability of materials and products (reshoring will take a lot of time if it even makes sense).
Nuclear regulatory environment is overly prescriptive. You need nuclear safety but you also can't use regulatory as a mechanism to shutdown all development - which has been the defacto case for the last 40 odd years.
Agree faster timelines are necessary -- we also need to see if this AI thing is real real or if its in over its skis a little bit. At this particular moment it feels like it might be selling more than it is offering - and thereby the energy demand won't materialize in the way that they assume...
Noting -- that the financial backers for energy products aren't all cash cows like tech companies -- they need to see real return on projects before they put up the necessary large amount of money and public-private partnerships given the risk on putting projects together like that.
Good news is that line replacement is definitely occurring. Especially for higher power lines, the modern conductors that are used for rewiring lines have lower loses and less sag.
PJM is a shitshow though.
Do you have more info on this? What's the deal with PJM's reluctance on upgrades?
I think it’s coming soon. The data center demand is just one aspect. In New York, a combination of half baked policy to go carbon free by 2030, and 100% EV a few years later, combined with insane political decisions (Gov Cuomo decommissioning the nuclear plant that supplied 20% of NYC), and the black swan event of Trump starting a trade war with Canada (Quebec is the best source of electricity) is making electric delivery rates alone increase 30%.
On the gas side, we’re both forcing ratepayers to spend billions on conversion from low to high pressure gas mains AND prohibiting new gas service.
The environmental lobby is so dumb and ineffective - the public will be demanding that coal come back.
The policy is intended to fail. There will be another wave of failed policy under the guise of “getting it right”. Because if you set the dates for 2050/2060 whenever the state is able to convert over, then businesses won’t do anything until 2050/2060. The most effective method forward is going to be continuously failed but enforceable policy to legally push industries and people into cleaner energy.
Agree that Cuomo decommissioning the plant was DUMB.
> On the gas side, we’re both forcing ratepayers to spend billions on conversion from low to high pressure gas mains AND prohibiting new gas service.
On the gas side, Trump's one BBB greatly increased the caps on natural gas exports.
The reality is that the companies extracting NG are not going to be giving it to Americans, they're going to sell it abroad and rake in way more money.
Which would maybe be fine... if we also weren't currently (and severely) artificially limiting the supply of renewables. Um, oops. There's nothing left.
All of our lobbies are fucking stupid.
Why coal and not shale? Is this not cope?
This is a tangent, but one untapped source of energy savings that seems to be invisible to climate activists is Microsoft Windows' constant drain on resources relative to Linux and MacOS. It's shocking how energy inefficient Windows is even when it's doing absolutely nothing noticable for a user.
This seems unlikely to me. Most of these machines are laptops nowadays. While my Linux system idles at 4W, the plug is only capable of going up to 60W. So even if Windows brought it to the point of nearly overheating the power brick (I’m no Windows fan, but I’m pretty skeptical there!), it would only be 1/10’th of a typical 600W window air conditioner.
The untapped energy savings is, IMO, getting people to run their climate control less. We should toughen up a bit, tell our brains that 50F-80F is the comfortable range. (Depending on humidity and your workload).
For any individual the difference is not noticeable. But across the several hundred million laptops in the US even 1 watt adds up.
I compared to air conditioners though, which would also be multiplied by that several hundred million factor (well, maybe the laptops get an additional factor of, like, 4 because a room can fit multiple people with laptops, and you don’t need an air conditioner in every season).
If we’re looking at choices a person can make, every choice is multiplied by millions when applied to the entire population of a country, so the 1W differences are swamped by the equally scaled 10W differences.
Every watt into laptops means a watt the air conditioner needs to move out.
you cannot ignore small things in the grand scheme as they add up.
It doesn’t make sense to talk about “adding up” unless we define what we’re aggregating over. Residential energy consumption, for sure, is a significant chunk of energy consumption.
If we’re looking at the things we can do to reduce our individual consumption, it absolutely makes sense to prioritize the things which are large relative to our other individual contributions, first.
Fine, but I have already upgraded to the most efficient heat pump, and so I'm at the end of where I can go there. Likewise my insulation cannot be improved without a major remodel (My house is not built to modern standards so there is a lot of improvement, but those require thicker walls and a higher attic ceiling) and so while I'd like to do them they are not possible. What is left is the smaller things.
Way to bring up Linux power management into an unrelated discussion x_x
It's true. There are more Windows desktops used in enterprise environments and households than linux and there are so many insane design choices like making the taskbar clock slider thing be written in react. How many useless clocks does that waste? That is a very easy optimization that can be remediated instead of being lazy and waiting for the next node from Intel or AMD.
Computers use up extremely small amounts of electricity. Most people don't actually know this. Your washing machine is a good ~100 computers. A lightbulb? If it's not LED, that's a few computers. For one lightbulb.
Of course it depends on the computer, but if we're talking laptops or corpo computers it's like 25 - 50 watts. Supercomputers, like those used for AI, are different of course.
I can pretty much guarantee you that the typical Windows desktop machine uses much less energy than the typical Linux desktop machine of similar capabilities. Linux power saving basically never works out of the box unless the user has carefully selected the platform to avoid the Linux kernel's numerous defects. By contrast every computer that comes with Windows has working energy-saving features from the factory because that's how Dell and HP get those Energy Star ratings.
The average windows box is full of so much bloatware, it likely evens out.
I wholeheartedly agree. Software bloat is a hidden energy drain on corporate America as well as the general public no different than how landlords only leech rents from firms and individuals while providing nothing of value. There is no need to politicize by yelling at climate activists though.
You are being funny, considering what a shit-show is Linux laptop power management.
As a desktop and laptop Linux user for many years, I wholeheartedly agree.
Because it is the exact opposite? Power management and drivers are a shitshow on Linux. When I want my Lenovo (RIP IBM) Thinkpad to last more than 2 hours, I have to boot Windows.
But is is no question: the Apple Silicon chips sip power. If you're looking to minimize watts consumed, a MacBook Air is still on top (or bottom), even if OS X has many pain points too.
There's quite a bit of flim-flam in the video, but the key falsifiable claim is that residential electricity demand is not growing. I am looking at the EIA Monthly Energy Review and it appears that residential electricity grows at ~1% CAGR this century. Which raises the question of whether the bills are now due for past underinvestment in generating capacity.
This has been obvious for several years now thanks to the crypto boom https://www.technologyreview.com/2022/04/18/1049331/bitcoin-...
Great video on how the public is getting screwed on energy deals.
Basically large tech companies have the deep pockets to push up prices at electricity auctions. But why bid in public when you can do those deals in private. That's the first problem. All that needs to be out in the open.
What really irks me is that the market is so manipulated that we can't do anything about it. Think about NEM 3.0 vs 2.0. Putting data centers in their own rate class does make sense as the first step.
>Basically large tech companies have the deep pockets to push up prices at electricity auctions. But why bid in public when you can do those deals in private.
Public utilities can't do the same? Moreover if the implication is that large tech companies are somehow getting great prices at the expense of residential users, what does that mean for the electric generators on the other end of this transaction? Why are they leaving money on the table by selling to large tech companies for cheap?
Watch the video.
These companies are regulated and can only charge for the costs they incur plus a flat profit on top of that of 10% or so.
The datacenters give allow them to justify building a lot more capacity to serve them. That increases costs, which means that 10% added for profit is now a bigger number and they can give bigger returns to their shareholders. But those profits are extracted from the existing customers who now see higher bills to cover the costs of expanding capacity to serve the datacenters.
It's a question of aligning incentives.
(Why) are the datacenters not (also) charged (pun intended) for this?
Because they are given a sweetheart deal to attract them to that specific area. And that deal happens behind closed doors.
Can you link the .PDF with this evidence?
The whole capped profit creates the distortions you illustrate.
The effect has a name: the Averch–Johnson effect, named after the Harvey Averch and Leland Johnson paper: "Behavior of the Firm Under Regulatory Constraint"
Fun stuff: https://en.wikipedia.org/wiki/Averch%E2%80%93Johnson_effect
Private deals - do you mean like a Power Purchase agreement? That doesn't fall out to the public cost domain and certainly shouldn't be a public good.
NEM 2.0 was completely unsustainable and it was extremely regressive, punishing people who couldnt afford panels. 3.0 is a much better system
In the US people are complaining about electricity like it’s a scarce resource while China has drastically overbuilt energy capacity (solar and nuclear) and gives 0 shits about AI or crypto. One of these approaches seems better thought through.
Fact- China is the number 2 bitcoin miner, and cares a great deal about AI and is only constrained by chip import restrictions
> solar and nuclear
And coal: https://www.carbonbrief.org/chinas-construction-of-new-coal-...
Right - because their nuclear power plant construction mechanisms are just starting to scale up and they need something to backfill baseload when solar doesn't cut it. Coal growth is still shrinking and will peak in 2030 at which point it will get squeezed out by all other energy sources.
Also, crypto helps!
Texas paid bitcoin miner Riot $31.7 million to shut down during heat wave in August: https://www.cnbc.com/2023/09/06/texas-paid-bitcoin-miner-rio...
Two of the biggest bitcoin mining companies in the world are battling it out in a Texas town of 5,600 people: https://www.cnbc.com/2021/10/31/bitcoin-mining-giants-bitdee...
Cryptocurrency Mining in Texas: https://earthjustice.org/feature/cryptocurrency-mining-texas Discussion: https://news.ycombinator.com/item?id=37516498
Texas had to pay a bitcoin miner to shave load... that hurts, not helps? Helping would be a mandatory brownout for excessive users.
From 2022:
> […] Applying a new data set for country-level energy prices since 1960, this study evaluates the effects of energy prices on economic growth in 18 OECD countries by controlling for other important macroeconomic conditions that shape economic activity. Mean-group estimates that control for cross-country correlations are used to emphasize average responses across nations. Averaged across all nations, results suggest that a 10% increase in energy prices dampened economic growth by about 0.15%. Moreover, some evidence exists that this response may be larger for more energy-intensive economies.
* https://www.sciencedirect.com/science/article/abs/pii/S01409...
Though an interesting graph for Sweden on GDP growth and energy use:
* https://ourworldindata.org/grapher/energy-use-gdp-decoupling
Texas even subsidizes the energy bill of crytominers directly, they receive discounts and subsidies while consumers pay more and more:
https://earthjustice.org/experts/mandy-deroche/how-much-do-w...
Most water systems have similar deals. Customers will pay a fixed rate infrastructure base fee plus a usage fee, that usage fee for residential is higher than the discounted rate industry receives.
Much like corporations venue shopping municipalities for tax discounts, we need a Federal ban on these corporate subsidies at a Federal level to put us all back on a fair playing field.
Actually it's because there isn't enough solar and generation because of previous policies. Check out how Alex Albrecht got screwed doing solar in Los Angeles.
That doesn't excuse gouging consumers now to subsidize data centers for large corporations.
Yes it does. Prices are based off supply and demand. Demand is increasing but the government is making it prohibitively costly to expand supply so prices are going up. The solution is to generate more electricity, not stifle demand.
Is the increased electric bill offset by reductions in taxes? The answer looks like "probably".
https://www.loudoun.gov/6188/Data-Centers-in-Loudoun-County
"Loudoun County is home to one of the largest concentrations of data centers in the world through which much of the world's internet traffic passes every day. The industry has helped to diversify Loudoun County's economy, which has included bringing new jobs and revenue streams to the county over the past 14 years that have helped to keep real property tax rates relatively low."
In the same county, maybe. But electricity is a tradable commodity. The electricity price goes up for the neighbouring counties without the tax benefits.
The energy costs of certain parts of Maryland are about to skyrocket, or have already risen dramatically, driven by datacenter demand, primarily but not exclusively from Northern VA.
https://www.wusa9.com/article/tech/science/environment/data-...
https://content.govdelivery.com/accounts/MDOPC/bulletins/3ea...
https://www.thebanner.com/economy/maryland-energy-cost-virgi...
I am an engineer at a large utility in the Pacific Northwest. My ratepayers pay very little for power, We generate with hydro which is very, very cheap. There are very few opportunities for more hydro. Every new MW we add is going to cost two orders of magnitude more per MWh compared with the existing stuff. We would not need to add capacity like this, except that we are being absolutely deluged with requests for enormous new data centers, which will require purchasing power from these very expensive sources. This will drive up prices for everyone. There's a lot of people in this thread that are confidently claiming things will happen in systems they barely understand. I'm here to tell you this will create a power affordability crisis in my region, and it hasn't even started yet. The effects won't be fully seen for years.
I feel like the solution is for the state owned power company to add capacity to scoop up the additional revenue. This is quite literally an opportunity to advance society while generating more tax revenue and *improving* public services.
Blaming tech for using power is regressive, the real issue here is bad energy policy.
> add capacity to scoop up the additional revenue.
they are; but Big Tech isn't paying the cost to add that capacity, because it's getting volume discount prices to attract them in the first place. So the utility has to raise prices on everyone else to cover the cost of adding the capacity.
so essentially, all the other consumers are subsidizing the extra power needed for the data centers (instead of the data centers paying for it)
This sounds like "lines goes up" view of it.
Adding capacity isn't free either, in particular on environmental costs and/or additional risks. We haven't been trying to reduce consumption for a few decades now just to be cute, the tradeoff just wasn't worth it.
Maybe it is? The US has ample land and money to build out clean energy, there’s zero excuse not to be adding capacity.
It’s literally the governments job to build out the infrastructure for society to thrive. Don’t blame technology for getting better, blame the government for not doing their job.
> The US has ample land and money
The money part doesn't ring true given the current US political climate. That's not the president you elect when your future looks bright.
> society to thrive
Does society thrive on AI though ? And is the technology better ?
Those are real questions. My answers would be no, we have better use for our resources. If we need to sacrifice for something, at least be it worth it.
Why let a state owned power company scoop of the revenue if we could privatize it and then let rich people scoop it up?
Related:
Electricity prices are climbing more than twice as fast as inflation
https://news.ycombinator.com/item?id=44931763
Big Tech's A.I. Data Centers Are Driving Up Electricity Bills for Everyone
https://news.ycombinator.com/item?id=44905595
The U.S. grid is so weak, the AI race may be over
https://news.ycombinator.com/item?id=44910562
AI is booming so are household utility bills
https://news.ycombinator.com/item?id=44931763
Clear example of privatization of everything... until its the big corporations who need to get something done - then its socialism in disguise. And of course they hide all the deals behind closed doors so their customers can't see who all, and how much they're subsidizing.
If there is a sudden increase in demand/supply issues due to large corporations they should be the ones to subsidize the bill. I'd argue in the communities this affects they should be forced to cover for everyone who lives in the area for materially making their lives worse.
Not to mention utility companies should be forced to be transparent in the pricing/deals they make with tech companies. They are likely just luring them in with a deal and passing off the rest to their customers so they still don't lose out in the short term. They want the best of both worlds - still making all the same profits short term by passing on to customers while getting the long term lock-in for these data centers.
>should be the ones to subsidize the bill
Clarity on the wording here. They should be the ones _paying_ the bill, not subsidising it. It's not subsidising if they're the consumer of the marginal product.
If an industrial customer pays for the entire power plant, but then also has to outbid everyone in 13 states to access the energy it paid for, then the industrial customer has clearly subsidized it.
They can build a dedicated plant off-grid, or BTM.
> should be the ones to subsidize the bill
It's another variation of what all the MBAs are trained to do: build your fortune using someone else's money. The difference here, of course, is that that money is being used without consent.
> socialism in disguise
I really dislike this kind of rhetoric. This has nothing to do with socialism. Corporations profiting from externalities and pushing costs onto regular workers is just capitalism. If you have a problem with it, maybe you have a problem with the inevitable concentrations of wealth and power which result from capitalism.
Well, what if the corporations would pay the bills - then they would increase the price of the product so the people would pay for it again. Now you can say that you do not use Metas or Googles products, but their business models is ads - and you DO pay the products advertised for.
In scenario 1, a corporation externalizes some of its costs. Those costs are then paid by people who may or may not actually use the corporation's product — people who never chose to be part of any transaction. This is coercive because the people paying for the corporation's externalities are forced to: they may not use the product, or do so to different degrees not proportional to the price they pay for the externality.
In scenario 2, the corporation does not externalize costs and raises their prices, offsetting costs by passing them on to their customers. The people paying the additional cost are those who know the price of what they are buying and willingly engage in the transaction for the good or service.
Do you understand why scenario 2 is bad and scenario 1 is less bad?
That is just a very simplified and incomplete model. I never owned a car, so should I advocate to stop all fundings for streets? Well I consume products build by other people that use cars to go to work. Now, if I don't consume drinks of the Coca Cola Company, what if my cleaning lady enjoys those in her break? Direct vs. Indirect is not a good measure of value, PRICE is.
> I never owned a car, so should I advocate to stop all fundings for streets?
Streets are generally paid for by taxes, which are categorically different than corporate profits. In theory taxes are under democratic control. If you don't want to pay for streets you don't use, you can vote for a politician who passes that law. You have no control over the governance of a private corporation, but it can still pass its costs on to you via externalities (in the absence of regulations preventing it from doing so).
> Now, if I don't consume drinks of the Coca Cola Company, what if my cleaning lady enjoys those in her break?
What are you even talking about? What is the externality here? The wages you presumably pay your cleaning lady are hers to do with as she wishes.
> don't want to pay for streets you don't use, you can vote for a politician who passes that law
Just as you can build your own power plant (solar, wind) if you don't like the electricity prices of your provider... That works in theory, in practice you will have to pay...
That's a strange definition of capitalism you're using.
Most people use capitalism to describe a system where people trade goods and services with as little interference from government as possible.
In this case, the government has written laws in a way that indirectly transfers wealth from consumers to large corporations.
Saying that such laws are "inevitable" in any conceivable capitalist system is unfalsifiable and adds little to the discussion.
I think calling it "crony capitalism" to make it clear that it's the undue influence of capital on government specifically causing the problem here lends more clarity to the discussion.
> Most people use capitalism to describe a system where people trade goods and services with as little interference from government as possible.
This is not a real definition. Saying "most people" use the term to mean what you want it to mean in this argument is ridiculous.
Capitalism is a system of private ownership of capital. We live under that system. Anything you see that happens now is the result of that system because it's the one that exists.
There is no such thing as "crony capitalism". What came before crony capitalism? Was it regular capitalism? Did regular capitalism turn into crony capitalism? Or — more likely — is it all one continuous process and system of accumulation?
> Most people use capitalism to describe a system where people trade goods and services with as little interference from government as possible.
“Capitalism” is a term coined for the dominant system of the industrialized West in the mid-19th Century and is defined by the specific orientation of property rights in that system around the private and marketable ownership of the non-financial means of production (the “capital” in “capitalism”), and the way in which the system was fundamentally structured around—and institutions within it, including government, invariably served the interests of—the owners of that capital, who formed its ruling class, displacing the landed hereditary aristocracy of the preceding systems.
While the dominant politico-economic systems of the developed world have evolved somewhat since then, with modern mixed economies having structures in place mitigate some of the adverse impacts the original system for which thr label “capitalism” was created for has on the vast majority of the population that is not major capital owners, it retains the basic property structure and resulting class heirarchy of the original “capitalism”, and neither it nor the original tended tof eatite annabsence of regulation of commerce.
> I think calling it "crony capitalism" to make it clear that it's the undue influence of capital on government specifically causing the problem here lends more clarity to the discussion.
The commanding and undue influence of capital on all of society, government and otherwise, is literally the feature for which critics of the then-dominant system coined the term “capitalism” to refer to that system. It doesn’t need an extra qualifier for that.
This video is unconvincing. The thesis of the video basically boils down to "AI companies are using electricity, driving up prices, and residential consumers are paying for it". However it neglects that most of the power usage is caused by the same residential users, who are getting something out of it. 80–90% of AI energy usage is estimated for inference (eg. generating responses on chatgpt), not training[1]. Moreover despite AI companies being unprofitable as a whole, they're making fat margins on inference[2]. Therefore it's reasonable to assume that most of AI electricity consumption that the video complains about is as a result of ordinary people using AI. Taking this into account, blaming AI companies for jacking up electricity prices in this context makes as much sense as blaming airlines for jacking up oil prices. They're only doing so because people are buying their services, and presumably deriving some sort of utility from it.
[1] https://www.technologyreview.com/2025/05/20/1116327/ai-energ...
[2] https://martinalderson.com/posts/are-openai-and-anthropic-re...
You are not taking into account the fact that data centers tend to be pretty concentrated geographically.
If I live in an area that is trying to attract data centers by giving them sweetheart deals, my electricity bill is going to go up by more than my share in inference usage since those data centers will be used by people all over the globe.
Many, people, probably most, are not using any inference in their lives - certainly not willingly or knowingly.
Worldwide, maybe. But in the "western world"? Couple weeks ago my mother asked me to compare the main LLM providers. And then for tips how to best use it. She is over 80.
And there was a letter in the local paper this week about someone complaining they councils park in town because they didn’t have a debit card and they certainly don’t have any “app phones”, a far more common approach amongst older people
Of course, for example me, the person you are replying to. I am an LLM vegan.
How is this different than any other form of industry?
If I live in an area that is trying to attract [steel mills] by giving them sweetheart deals, my electricity bill is going to go up by more than my share in [steel] usage since [that steel] will be used by people all over the globe.
I think if you found the average person on the street and asked them if their electric bill would go up when a steel mill opens nearby they'd only say yes because why else are you asking them.
Most other industries create (more) local jobs which would offset the localized cost.
This sounds like a generic NIMBY argument to me? eg. "why should we build a datacenter in my backyard, when it'll raise MY electricity prices?" Of course, if everyone has this attitude, no datecenters, power plants, homeless shelters will ever be built
I am very YIMBY, but it's totally reasonable that you don't want industrial work done in your backyard. This is exactly what zoning and city planning were designed for.
It seems odd to me to try to redefine Y/NIMBY depending on the individual project. I wouldn’t want a sewage plant or a datacenter in my neighborhood, absolutely no one would. Why would one be NIMBY and one not?
Then talk to your local politician
Very anecdotal, but a friend of mine works for a fuel cell company that basically converts natural gas and/or propane to grid electricity a huge part of their work right now is powering new datacenters for AI growth. I don't really know the technology, so I may have the details wrong, but the fact companies willing to take on cost and complexity does hint towards real limits on energy supply.
> Taking this into account, blaming AI companies for jacking up electricity prices in this context makes as much sense as blaming airlines for jacking up oil prices.
I'm not sure if "blame" and "jacking up" are the right words here. AI technology is demanding power quicker than it is being added to the grid and add so there is going to be tension. Now if tech companies start buying the grid operators and power, I'm more inclined to start blaming the tech companies themselves, but for now I still place broad energy failures squarely with the government.
I don't use AI.
in Chelan County where I live they came up with a Large Load policy specifically to protect the rates of other local users... the large load customer gets short-term access to power and then has to source it elsewhere (with the PUD in the middle). Microsoft is here and apparently plans to source from Helion long-term which is building their first generator just down the street. they also have to pay for their own grid upgrades as needed.
more information here: https://www.chelanpud.org/about-us/events-calendar/event/202...
The best way to make the grid fair is to remove fixed rate accounting altogether. The reality is that all power must be delivered instantaneously. Any meaningful arbitrage (i.e., adding actual value to grid services) must occur within milliseconds. The ability to purchase contracts on power that hasn't been generated yet is what is driving all the volatility and exploitation.
There is a reason that Griddy (variable rate wholesaler for consumers) had to be killed in Texas. The popular narrative is that they "fucked" their customers over in 2021. The reality is that anyone who had been using their variable rate system for more than 12 months (and made any attempt whatsoever to respond to demand signals) had easily compensated for the $9/kWH crunch that lasted merely ~72 hours. Most of their other customers who I talked to had no problem with the situation. There was a deliberate PR campaign (likely by the other retail electricity providers in the market) to make it seem like an unbelievably impossible situation for a normal person to cope with. Now it's illegal to have wholesale rates for consumers in Texas. We've gone two steps in the wrong direction.
> The ability to purchase contracts on power that hasn't been generated yet is what is driving all the volatility and exploitation.
That is completely opposite of the truth. The ability to buy a contract is what allows power generators to get loans to build at good rates. Everyone wants someone else to take on the risk, when you buy a contract for power at some rate in the future you are taking on the risk that prices could go down. The generator then goes to the bank and can show that even if competition forces prices down they still are going to get that much from the contract and thus can pay off the loans.
Power is complex and I just touched the surface of the truth.
There are capacity markets that are expressly designed to get around these concerns.
Would have liked a text version. And that would save energy too.
Knowledge online rarely lives alone. Here's a similar article: https://theconversation.com/how-your-electric-bill-may-be-pa...
In the description text field click "...more" and then "Show transcript".
There's a transcript of the video in the description nowadays. It's not always great to read or scrub through, but it's a text version.
Absolutely agree. I wonder how many kw get wasted watching YouTube or ticktack?
Go to Google's AI studio. Drop that link in it. Ask it for a transcript or summary. Wait a minute. Then read.
This proves the GPs point about saving energy. Think about how inefficient it would be for each individual to create their own transcript via an LLM.
Warren Buffett is going to be rich!
I understand what you are meaning, but saying that Warren Buffett is going to be rich is like saying that the pope is going to be catholic.
Ah now I see why electric costs are increasing at double the rate of inflation. But hey I got to make some images of the Muppets in various movies using AI.
https://www.npr.org/2025/08/16/nx-s1-5502671/electricity-bil...
https://imgur.com/a/d14oNOP
Yea let's just have China lead in AI... They're building out additional capacity at an incredible pace to meet demand.
Unless the big data centers are entering into mandatory load reduction during high strain on the system, then theres going to be an increased cost to the little buyers.
The tradeoff is that Utilities can upgrade their systems and improve uptime thanks to the higher base load. It's up to the utils to negotiate well against the big data players.. good luck on that with politicians breathing down your neck to secure a big win.
.
>5GW consumption I assume is per day?
GW is already a measure of rate of energy use. You can talk about GWh per day (which is really just another way of saying 0.041 GW), but GW per day is only sensible in the context of a ramp in power consumption.
Assuming the original source of that number didn't incorrectly conflate GWh into GW, that puts your comparison as underestimating the energy usage 24-fold.
I've heard 5GW touted as the capacity (so like 120GWH per day if flat loaded at 100%). However that is clearly B.S as I've also heard 2GW and another much more sensible number
Just to give perspective 5GW would power about 10 Million H100s or about 5 million G200. Completely farcical, won't happen.
Total rewrite of my previous reply:
There's a 7-million-sqft data center near Reno that claims 650MW capacity. Zuck says his will be 4 million sqft. So that doesn't really get us to 5GW unless we're going vertical.
To be fair, datacenter power density has skyrocketed over time: https://www.ramboll.com/en-us/insights/decarbonise-for-net-z...
These are being built at the same time, so any trends over time are not in play.
If they averaged only 60KW per rack that would be about 1.7Million sqft of 'white floor space' or about 3-4 million sqft gross building size.
> Yet increasing the power consumption of an area by 25X over a few years requires a significant amount of transmission and distribution infrastructure upgrades to make that happen
Not if you put your data-center right next to the power station.
> The 5GW consumption I assume is per day? If so, that’s on par with 170,000 average homes… being put in a parish with around 7250 households.
Watts are Joules per second, you'd only measure "watts per day" as a rate of change of power supply, not absolute power supply.
Also, 5GW / 170000 homes ≈ 29kW/home, even the USA isn't that heavily supplied with electricity.
(My home in Germany is very efficient, 0.5 kW average, including heating and cooling as well as all devices).
Like any other industry, I don't care what the electricity cost of data centres is if they create actual value. This is no different than a factory that makes car parts - you need energy to make stuff.
I think the part the bugs me is how over-hyped AI is, and there's a fear that all this electricity is being used as part of fundamentally unsustainable business models that will not prove sustainable long term. If that's the case, and we're overbuilding data centres what's going to happen to those buildings once the bubble pops?
Because of Trump, and just like Trump, this is a perfect storm. Unrestrained AI and crypto are causing a surge in demand at the same time Trump is spitefully and recklessly killing wind and solar installations and subsidies.
Trump-licans - the party of mean and stupid and self-destruction.
Some of us call this "capitalism". Prices are driven by offer and demand, efficient market allocation to who can extract most value from a resource, we like that on HN.
Prices are actually driven by value. How much do you want the thing? Do data centers value power? I would think so, since they cannot operate without it. Utilities should be charging them a premium.
If utilities charge them a premium, why would they sell power to those which do not pay the premium instead of auctioning all power?
Because they're theoretically regulated and ordered to do so. The market for power is not a free market due to the locally monopolistic nature of it.
Things cost exactly what we pay for them.
> Prices are driven by offer and demand, efficient market allocation to who can extract most value from a resource, we like that on HN.
What is the value of the bottom-half of income earners being able to pay for heating and cooking?
This is not capitalism because the laws are distorting the market, favoring large corporations over consumers.
Captialism is really good at efficiency. But it is not so good at allocation, specifically for essential goods, and that's the problem here.
How do you define what is an essential good here? For electricity I would place most residential usage near the bottom in priority and definitely below a datacenter. People complaining about facing a tradeoff between setting the thermostat a couple degrees higher this summer or facing a higher power bill just doesn't get much sympathy from me.
If you view powering people's homes as a lower priority than powering a data center then there is no reasonable discussion to be had here.
Why do you want to subsidize the data centers of some of the largest and most profitable corporations in the world?
And this is precisely why raw, unfettered capitalism isn't really the norm in the developed world.
I don't think it's that hidden by now lol
The real reason is that fossil fuel plants have been offlined and decommissioned, and the reason this occurred is because they were made unprofitable by punitive regulation far more so than market forces.
> fossil fuel plants have been offlined and decommissioned
Coal is down, but there has never been more natural gas generation in the US: https://www.epa.gov/power-sector/power-sector-evolution
Yes, perhaps I should have said coal.
News flash Casandra, a $65/mo power bill is peanuts. Try owning a home with HVAC.
You seem to be missing the point entirely.
1. The bill shown for July 2024 was for about $100, which includes $65 of supply charges.
2. The bill for July 2025 was for about $129, with supply charges of $84.
3. Despite actual usage being lower for July 2025, the total cost increased by 29%.
The point being raised is that energy costs are higher for individuals because the cost of building infrastructure to support data centers is being subsidized by individuals. Even if her bill was for $65, it is inconsequential to the main point. I don’t understand why people make comments like yours other than to muddy the waters and obfuscate the issue. Or truly failing to understand what is being communicated due to some bias or ideology. A sad microcosm of what passes for discourse in our time.
Why should Cassandra be punished for making sensible decisions to keep her energy usage low?
she lives in a tiny apartment though
I've got the solution! Digitize your brain and move into the data center. Can't recommend the show "Pantheon" enough.
This does not have anything to do with the current thread.
If you move into the datacenter, you will benefit from the volume discounts on energy ;)
Does anyone here care though? Everyone using LLMs are directly supporting lower quality of life and high cost-of-living across the country.
No person can do anything except fund Zuckerberg's next island purchase and laugh about how they’re using LLMs to vibe code their next unreleased hobby app.
It is truly sad that we are selling out our quality of life for short term gain for tech corporations.