(3). Highly preferred that you are not going to impact cost negatively to households (otherwise why we have this discussion)
based on these assumptions, solution I see is, BigTech subsidising energy costs for 10 years for nearby households (area will be geofenced, e.g. in the radius of 50km), subsidy will be based on the prices outside of that radius. e.g. if you everyone outside of closest DC pays 1$ and in the radius prices become 1.5$, 0.5$ will be covered by BigTech and they're also responsible & pay to setup the system to automatically include everyone in subsidy program, not like you need to apply
Also BigTech is not going to build the power generation plants, it must be built by existing processes to minimize impact on pricing
* power generation is not local in most of the cases. You'd be still fucking energy market to anyone outside it
* power is also used by companies people actually want. Even if household power cost wont chance thanks to that approach, the price for power for every single business around will increase.
* similarly any other manufacturing business will cost more. In essence, the AI boom will reduce profit margin of every single business that has electric power as significant cost in the production. Which is a lot of industries.
"Making AI pay more for its useless power draw" is nice idea but it is pretty hard to realise. Unless we start outright denying connection to power grid but that's pretty dangerous political precedent to set.
> "Making AI pay more for its useless power draw" is nice idea but it is pretty hard to realise.
And should never be a thing. It's not for others to decide something is useless. Let the market figure that out. It should simply be an even playing field, and in theory we already have the regulatory mechanisms to get that done. If the rates are currently able to be abused, that is entirely correctable.
It should not matter if I want to build 1000MW of aluminum smelting capacity, or a facility that digs holes and fills them back in again - so long as I'm paying the regulatory rates that have been set to be fair across all similar power consumers.
The bigger problem by far is that we spent 40-60 years not building anything. Eventually you run out of the previous generation's power capacity. Can only ride inertia and offshoring your industrial energy usage for so long. There are a lot of chickens coming home to roost in the next decade or three on this topic, electric grid was just pulled forward a bit due to the AI bubble.
Power prices have nearly nothing to do with proximity like that, at least in most areas. Local interconnect surcharges are a thing, but not usually material.
It's the entire region where prices will increase. A facility could be built 200 miles away from you and it would have just as much of an impact on your power bill as one built across the road.
> Also BigTech is not going to build the power generation plants, it must be built by existing processes to minimize impact on pricing
The issue with this requirement is that it's not a capital problem. Big tech would be happy to pay someone to build more power generation and just take a power feed, even if it costs marginally more.
What they cannot do is wait 12 years for what they could get done in 2. And I do not blame them at all for this at all - building anything in the US is basically impossible these days due to the NIMBY and general culture of not getting things done. We forgot how to rapidly build infrastructure in an efficient manner, and anyone putting proverbial boot to ass to get this done is a good thing.
It's not a case of a datacenter operator wanting to save a few pennies per kwh. It's a case of the power literally not being there and the utilities being utterly incompetent at building anything in any reasonable timeframe.
Put some company stock in escrow. If they fulfill the promises, they get it back. If not the government keeps it, and uses it to build whatever needs to be built.
A higher tax on DC with the tax revenue invested into public electrical infrastructure and public power stations. With billions powered into AI they can out-price everyone out of resources they use - be it directly (high electricity rates) or indirectly (high prices for everything which is needed to expand power networks). But it will not happen of course.
Build and colocate the needed rewnewable (solar/wind/battery) capacity with the data centers and make them energy efficient by eg choosing cool locations.
1. DCs must be built anyway
2. You can't take away energy from households
(3). Highly preferred that you are not going to impact cost negatively to households (otherwise why we have this discussion)
based on these assumptions, solution I see is, BigTech subsidising energy costs for 10 years for nearby households (area will be geofenced, e.g. in the radius of 50km), subsidy will be based on the prices outside of that radius. e.g. if you everyone outside of closest DC pays 1$ and in the radius prices become 1.5$, 0.5$ will be covered by BigTech and they're also responsible & pay to setup the system to automatically include everyone in subsidy program, not like you need to apply
Also BigTech is not going to build the power generation plants, it must be built by existing processes to minimize impact on pricing