I think my main concern really more applies to the potentially volatile nature of the industry, much of which is still in the initial part of the boom cycle. I do agree with your premise that these can be a boon for local government budgets and bring subsequent benefits, but hearing that the datacenters fund nearly half of a county's budget gives me significant pause.
The main problem is that, should these datacenters go bust or otherwise radically transform their operations in the future, this threatens to put the city or county in a huge financial hole. The same is true for any other industry, as well. In a lot of ways, I think it's more similar to a resource economy than a typical tech job. And while I know cities argue that you can just use the money to buy extra stuff, like infrastructure, that just becomes a hole to fall into in the future when it comes time to maintain and replace that infrastructure.
All that to say, I think cities need to treat these windfalls very carefully. I personally think the funds from these datacenters should be put into sovereign wealth funds from which the government can live off the interest, not unlike Norway and their oil fund. That way, if the datacenters do go away or change significantly, it doesn't blow a hole in the city budget while still retaining a lot of the benefits, to say nothing of long term stability.
For what it's worth, I voted in favor of a datacenter in my own city, albeit a much smaller scale one oriented more towards internet services (like cloud computing) rather than AI. The only real subsidy we gave them was a limited time deal on their electricity (the city owns its own utility) to pay them back for agreeing to build up the electrical connections they need, for which they don't get the benefit if they don't stick around. That and a couple million dollars of our own infrastructure upgrades, which will largely pay for themselves fairly quickly with such a large electrical customer. We also required them to use a closed-loop cooling system, banned crypto currency mining in the facility, and put strict limits on water consumption. So hopefully it works out for us.
Certainly, there is a risk of data center demand decreasing. In these cases, at least this does not come with a huge economic impact from the loss of income of residents which means there is still a tax base for the city: it just may need to come with resetting property tax rates up.
For instance, when post-industrial cities got hit with a changing economy, they had neither the large factory to tax nor another tax base to compensate, as a large percent of their residents had lost jobs. I think this is less of a concern with data centers. Instead, the tax cuts from the data center existing would need to be reversed, which should not be as bad for the local city (though homeowners would complain).
I think Dustin may have more of the right of it here. Sudden jumps in property taxes for the majority are never going to be politically viable. I think the more likely result of a data center bust in this scenario that is that local governments start facing windfalls or make up the lost revenue with other, less-efficient forms of taxation.
I think the optimal playbook for local govs is to indeed find a way to sequester the cash in a wealth fund or some other financing vehicle that can only be used for capital expenditures or as a populist Citizens' Dividend to help win support for the project. Preferably it would be used as much as possible for capex, with just enough handed out to the public to make it viable.
I missed this somehow. I think we are saying the same thing:
Let's assume a local government does a good job of negotiating (ie. the datacenter is a net positive)
Then, the datacenter through that agreement is providing cash annually to the government.
That cash then can decrease residents property (or other) taxes from the counterfactual of what their taxes would've been with no datacenter.
If datacenter goes bust, that supply of capital is gone.
The city has to return to the previous tax rate.
In the industrial area, the factory collapse also came with a collapse in the local labor economy, which really screwed cities, but here, the labor market would be rather unaffected (except that a bust in datacenters probably has a larger macro impact on the global economy)
Going back to the previous tax rate is not preferred, and residents would see it as a huge tax hike, but would not realize that without the datacenter that is the rate they would've been paying in years past.
Your idea is that a fund could be created to ensure that the city is not over-leveraged on the datacenter paying say annual taxes.
That sounds right to me, and more or less in line with the fund I propose in the piece--there's probably tons of ways something like this could be structured and cities will need to adjust based on their local context.
I tried to write about this (I don't think very successfully, I'm still not well-versed in it) last year during Wisconsin's Port Washington controversy.
I think there's too much asymmetrical information. Yes, most fears are overblown, but at the same time if a corp ignores the noise or light pollution ordinance, what is the city going to do about it? Pull permits?
I saw in DeForest, WI, a relationship between the governing body and the corp. Residents alleged kickbacks. I don't know if there are, but when the decisions are made by so few and the investment is so large, it's hard not to imagine the citizens being bulldozed by the corp interests.
I think that's where I land and agree with y'all. Don't ban the buildings, but be very upfront and a hard negotiator about your ordinances that matter most. Don't give tax breaks (thanks state of Wisconsin!), and if the companies decide that your region (whether for resources or personnel) are needed, they can either meet your demands or go somewhere else.
Too often the corps come in as "build this thing, we need to make money" and not "here's how we will benefit the community and how we will be good stewards".
Yes, and this is where I think data centers can be proactive to prevent backlash.
Data centers should propose all these rules and stipulations up front and propose a fund to bring down electricity costs, etc.
I think we will see more of this as data centers recognize they can cede some of their profits to the local government/residents while still making large amount of money.
Virginian here (former planning commissioner). I agree that the state subsidies should be ended. However, these approvals should not be local land use decisions.
If you look at § 58.1-3295.3 of the state code, it's the computer equipment, not the land, that is the source of the tax, and it doesn't account for the geographic externalities required to run them. It's a nice idea that local governments can condition approval on sourcing the power and the water required for operations, but that still doesn't address the externalities.
Is the battery part of your phone or computer? Can a computer operate without power? Where does that power come from, and why wouldn't the source of power (particularly off site solar panels and batteries dedicated to powering these centers) be considered part of the computer and taxed at the same rate as "computer equipment?" (It is not.) In the event of a severe drought, where local residents and businesses are required to reduce water consumption, will data centers get an exemption? If so, why? Many of these centers use diesel generators as backup power. What about air they pollute?
The benefits are too concentrated in the localities approving these centers while the impacts of the resources needed to operate these server farms are externalized. Final data center approval should be the purview of regional planning district commissions and the state departments of energy and environmental quality. Whatever the arrangement, the benefits should be more evenly distributed and shared with the localities providing the power and water.
Yes, this is a good point. There can be multiple jurisdictional entities at play, and the tide should lift all boats.
Regional planning, or cross-jurisdiction collaboration, feels like a good idea to me. As well as ensuring proper controls that prioritize residents -- water authorities should ensure residents are not most harmed in droughts. I believe Loudoun County services its own water, so Loudoun should have authority there. However, where shared, there should be mechanisms for neighboring jurisdictions to not be impacted.
That probably looks like binding regional rules on cross-boundary impacts (drought allocation, emissions, grid cost-sharing) more than handing final approval authority to regional planning bodies, which I worry could make a much larger regulatory barrier to develop data centers unless done right.
Revisiting this: I have updated my prior and am more sympathetic to a larger authority negotiating — particularly in small cities like those mentioned in the linked article
Sorry but I’m not with you this one. For starters, water usage is actually a concern, especially in dry places already experiencing water stress, and especially because that water stress will only get worse through climate change.
For example for Sydney, Australia, “data centres are set to use up to 20% of total Sydney water usage within five to 10 years.” So your statement that “Data centers don’t threaten national or regional water supply” is just plain incorrect. You’re not wrong that there are plenty of other inefficient activities we could change that also impact water supply, though, like growing corn for fuel.
But the biggest concern right now is climate. AI and data centres are adding absolutely massive energy demand at a time when any added demand slows the energy transition. We can’t afford this. We’re already at unacceptable levels of warming, and the slower we transition the worse it gets.
I’d love to see some hard numbers on this but I suspect that any productivity gains brought by AI are undone by the economic impacts of a warming climate.
If I understand your Sydney claim correctly, that is based on a future-looking forecast. That forecast does not assume closed-loop systems. They say there is a way out:
"Cooling technology choices are the main driver of data centre water use. Water efficient cooling systems exist and are available to be built into data centre designs. International markets are moving towards performance thresholds to ensure efficient cooling is deployed across the market.
These standards should be technology-neutral, fit for context and reflect modern designs
and data. Australian data centres can already achieve Water Use Effectiveness as low as
0.01 "
To me closed-loop water systems should be part of the minimum requirement that local governments negotiate. If it's mandated at the national level, I am probably fine with that.
Current AI technology is fundamentally incapable of providing the benefits that have been promised. The AI bros know that demand is not real, and that it's all about to collapse. The sooner we can force the AI bubble to pop, the better off humanity will be. We need to slow down and get it right.
Instead of trying to negotiate harder with these people, everyone should work to stop the biggest misallocation of capital in history.
I hold that AI already delivers immense value that will be transformative to the economy. Only in the past six month have we started seeing models that are consistently very good, and we are only at the beginning of that technology diffusing through the economy (large enterprises, for instance). Models will keep getting better and so will the scaffolding. I have high confidence in AI fundamentally transforming human life. Still a possibility I am off
Sorry, this generation of AI businesses is collapsing as we speak. There is no value to business because AI output is not reliable or consistent. Read some Gary Marcus.
I think my main concern really more applies to the potentially volatile nature of the industry, much of which is still in the initial part of the boom cycle. I do agree with your premise that these can be a boon for local government budgets and bring subsequent benefits, but hearing that the datacenters fund nearly half of a county's budget gives me significant pause.
The main problem is that, should these datacenters go bust or otherwise radically transform their operations in the future, this threatens to put the city or county in a huge financial hole. The same is true for any other industry, as well. In a lot of ways, I think it's more similar to a resource economy than a typical tech job. And while I know cities argue that you can just use the money to buy extra stuff, like infrastructure, that just becomes a hole to fall into in the future when it comes time to maintain and replace that infrastructure.
All that to say, I think cities need to treat these windfalls very carefully. I personally think the funds from these datacenters should be put into sovereign wealth funds from which the government can live off the interest, not unlike Norway and their oil fund. That way, if the datacenters do go away or change significantly, it doesn't blow a hole in the city budget while still retaining a lot of the benefits, to say nothing of long term stability.
For what it's worth, I voted in favor of a datacenter in my own city, albeit a much smaller scale one oriented more towards internet services (like cloud computing) rather than AI. The only real subsidy we gave them was a limited time deal on their electricity (the city owns its own utility) to pay them back for agreeing to build up the electrical connections they need, for which they don't get the benefit if they don't stick around. That and a couple million dollars of our own infrastructure upgrades, which will largely pay for themselves fairly quickly with such a large electrical customer. We also required them to use a closed-loop cooling system, banned crypto currency mining in the facility, and put strict limits on water consumption. So hopefully it works out for us.
Sounds like your city was smart in its approach.
Certainly, there is a risk of data center demand decreasing. In these cases, at least this does not come with a huge economic impact from the loss of income of residents which means there is still a tax base for the city: it just may need to come with resetting property tax rates up.
For instance, when post-industrial cities got hit with a changing economy, they had neither the large factory to tax nor another tax base to compensate, as a large percent of their residents had lost jobs. I think this is less of a concern with data centers. Instead, the tax cuts from the data center existing would need to be reversed, which should not be as bad for the local city (though homeowners would complain).
I think Dustin may have more of the right of it here. Sudden jumps in property taxes for the majority are never going to be politically viable. I think the more likely result of a data center bust in this scenario that is that local governments start facing windfalls or make up the lost revenue with other, less-efficient forms of taxation.
I think the optimal playbook for local govs is to indeed find a way to sequester the cash in a wealth fund or some other financing vehicle that can only be used for capital expenditures or as a populist Citizens' Dividend to help win support for the project. Preferably it would be used as much as possible for capex, with just enough handed out to the public to make it viable.
I missed this somehow. I think we are saying the same thing:
Let's assume a local government does a good job of negotiating (ie. the datacenter is a net positive)
Then, the datacenter through that agreement is providing cash annually to the government.
That cash then can decrease residents property (or other) taxes from the counterfactual of what their taxes would've been with no datacenter.
If datacenter goes bust, that supply of capital is gone.
The city has to return to the previous tax rate.
In the industrial area, the factory collapse also came with a collapse in the local labor economy, which really screwed cities, but here, the labor market would be rather unaffected (except that a bust in datacenters probably has a larger macro impact on the global economy)
Going back to the previous tax rate is not preferred, and residents would see it as a huge tax hike, but would not realize that without the datacenter that is the rate they would've been paying in years past.
Your idea is that a fund could be created to ensure that the city is not over-leveraged on the datacenter paying say annual taxes.
That sounds right to me, and more or less in line with the fund I propose in the piece--there's probably tons of ways something like this could be structured and cities will need to adjust based on their local context.
I tried to write about this (I don't think very successfully, I'm still not well-versed in it) last year during Wisconsin's Port Washington controversy.
I think there's too much asymmetrical information. Yes, most fears are overblown, but at the same time if a corp ignores the noise or light pollution ordinance, what is the city going to do about it? Pull permits?
I saw in DeForest, WI, a relationship between the governing body and the corp. Residents alleged kickbacks. I don't know if there are, but when the decisions are made by so few and the investment is so large, it's hard not to imagine the citizens being bulldozed by the corp interests.
I think that's where I land and agree with y'all. Don't ban the buildings, but be very upfront and a hard negotiator about your ordinances that matter most. Don't give tax breaks (thanks state of Wisconsin!), and if the companies decide that your region (whether for resources or personnel) are needed, they can either meet your demands or go somewhere else.
Too often the corps come in as "build this thing, we need to make money" and not "here's how we will benefit the community and how we will be good stewards".
Yes, and this is where I think data centers can be proactive to prevent backlash.
Data centers should propose all these rules and stipulations up front and propose a fund to bring down electricity costs, etc.
I think we will see more of this as data centers recognize they can cede some of their profits to the local government/residents while still making large amount of money.
Virginian here (former planning commissioner). I agree that the state subsidies should be ended. However, these approvals should not be local land use decisions.
If you look at § 58.1-3295.3 of the state code, it's the computer equipment, not the land, that is the source of the tax, and it doesn't account for the geographic externalities required to run them. It's a nice idea that local governments can condition approval on sourcing the power and the water required for operations, but that still doesn't address the externalities.
Is the battery part of your phone or computer? Can a computer operate without power? Where does that power come from, and why wouldn't the source of power (particularly off site solar panels and batteries dedicated to powering these centers) be considered part of the computer and taxed at the same rate as "computer equipment?" (It is not.) In the event of a severe drought, where local residents and businesses are required to reduce water consumption, will data centers get an exemption? If so, why? Many of these centers use diesel generators as backup power. What about air they pollute?
The benefits are too concentrated in the localities approving these centers while the impacts of the resources needed to operate these server farms are externalized. Final data center approval should be the purview of regional planning district commissions and the state departments of energy and environmental quality. Whatever the arrangement, the benefits should be more evenly distributed and shared with the localities providing the power and water.
Yes, this is a good point. There can be multiple jurisdictional entities at play, and the tide should lift all boats.
Regional planning, or cross-jurisdiction collaboration, feels like a good idea to me. As well as ensuring proper controls that prioritize residents -- water authorities should ensure residents are not most harmed in droughts. I believe Loudoun County services its own water, so Loudoun should have authority there. However, where shared, there should be mechanisms for neighboring jurisdictions to not be impacted.
That probably looks like binding regional rules on cross-boundary impacts (drought allocation, emissions, grid cost-sharing) more than handing final approval authority to regional planning bodies, which I worry could make a much larger regulatory barrier to develop data centers unless done right.
https://clmarohn.substack.com/p/when-the-decision-is-too-big-for?r=4rhmcp&utm_medium=ios
Revisiting this: I have updated my prior and am more sympathetic to a larger authority negotiating — particularly in small cities like those mentioned in the linked article
To be fair, threatening to reject data-centers entirely is a pretty great opening to a negotiation...
Touche!
Chandler: $10.2m property tax income seems pretty trivial vis a vis a multi-$B city budget?
Sorry but I’m not with you this one. For starters, water usage is actually a concern, especially in dry places already experiencing water stress, and especially because that water stress will only get worse through climate change.
For example for Sydney, Australia, “data centres are set to use up to 20% of total Sydney water usage within five to 10 years.” So your statement that “Data centers don’t threaten national or regional water supply” is just plain incorrect. You’re not wrong that there are plenty of other inefficient activities we could change that also impact water supply, though, like growing corn for fuel.
But the biggest concern right now is climate. AI and data centres are adding absolutely massive energy demand at a time when any added demand slows the energy transition. We can’t afford this. We’re already at unacceptable levels of warming, and the slower we transition the worse it gets.
I’d love to see some hard numbers on this but I suspect that any productivity gains brought by AI are undone by the economic impacts of a warming climate.
If I understand your Sydney claim correctly, that is based on a future-looking forecast. That forecast does not assume closed-loop systems. They say there is a way out:
"Cooling technology choices are the main driver of data centre water use. Water efficient cooling systems exist and are available to be built into data centre designs. International markets are moving towards performance thresholds to ensure efficient cooling is deployed across the market.
These standards should be technology-neutral, fit for context and reflect modern designs
and data. Australian data centres can already achieve Water Use Effectiveness as low as
0.01 "
To me closed-loop water systems should be part of the minimum requirement that local governments negotiate. If it's mandated at the national level, I am probably fine with that.
Ummm, no.
Current AI technology is fundamentally incapable of providing the benefits that have been promised. The AI bros know that demand is not real, and that it's all about to collapse. The sooner we can force the AI bubble to pop, the better off humanity will be. We need to slow down and get it right.
Instead of trying to negotiate harder with these people, everyone should work to stop the biggest misallocation of capital in history.
I hold that AI already delivers immense value that will be transformative to the economy. Only in the past six month have we started seeing models that are consistently very good, and we are only at the beginning of that technology diffusing through the economy (large enterprises, for instance). Models will keep getting better and so will the scaffolding. I have high confidence in AI fundamentally transforming human life. Still a possibility I am off
Sorry, this generation of AI businesses is collapsing as we speak. There is no value to business because AI output is not reliable or consistent. Read some Gary Marcus.