“some fields”, “parking lots”, “maybe some windmills” is a wildly vast under estimation of what would be required to do it.
7.65GW is A LOT.
The USA doesn’t even have an active 1GWh solar farm today. It’s 900MW, and it is 4600-5000 acres and is 1.3 million panels
To run batteries off that, you need to 3-4x that would be a 30GWh generation capacity battery to run in the non sun periods and give some buffer for bad weather. Then you need a massive battery as well to store it all. It would need to be able to run at least for a couple days, which means it needs to store 367.2 GWh (48h storage), so my solar numbers above are actually probably wrong since you won’t fill that up in 1 day, but maybe that’s okay if the size is just an emergency amount. 122GWh battery would probably the be minimum for a perfect summer condition 100% of the time day/night cycle.
The largest upcoming battery installation in the USA is 1.15 GWh, which is not built yet.
I’m not saying don’t do it, but you’ve really glossed over the sheer scale of it, and I doubt they’d be able to easily put the solar field near the datacenter, so there’s extra costs around transmission lines as well.
Edit: For 30GW of solar you’re looking at around 153,333 acres, or about the size of Dallas Texas, before the battery storage
I concede that my grasp of the scale involved was unrealistic. However, I don’t see a viable solution. The issue, I have realized, isn’t how to generate the power, it’s the scale of power needed to run these data center monstrosities.
So if solar and wind (with batteries, of course) would be very challenging, it is at least vaguely possible. Texas is a great state for wind, solar, and geothermal power. It already generates about 30% of it’s power from renewables. Texas has a lot of sunny days and a lot of open spaces for wind farms particularly in the western parts of the state. I can think of worse ideas than covering all those oil fields with solar panels or wind farms. You wouldn’t even need to stop producing oil. You can do both.
Nuclear really isn’t viable. Starting the approval process now, it would easily be 20 years before a new plant would start producing power. No one is waiting two decades to open their data center.
The only other option to power these data centers, the one they are going with, is fossil fuels. That means multiplying the amount of toxic and greenhouse gasses being pumped into our air which is the absolute worst option for everyone on the planet.
(Lemmy failed to send my original reply and I lost it I think, but if another similar one eventually show up… that’s why)
I wouldn’t rule nuclear out in the not so far future if SMRs properly come online. I think those could scale up demand at the pace they could build these monstrosities in the not so distant future, like give it another 15 years and it might start being okay?
And ya, the renewable way IS possible, and Texas would be a good spot for something of this scale, but it would be a huge undertaking. I imagine we would see amazing improvements in solar/battery tech building something like that out with economies of scale and additional research from the huge amount of funding it would create. Also there’s no reason they can’t make them do some % with renewables.
This is aslo why Space based stuff maybe isn’t as far flung as the idea seems. It obviously requires new tech that doesn’t function yet.
e.g Starship needs to get down to its targeted $50-$200/kg) which is not a given.
New chips that can run hotter so the radiator size is reasonable.
Cheaper solar panels as space based panels are very expensive right now.
But with that (and all the other things) you get 7.65GW of power being 7.65GW of power and minimal batteries required. Space solar is more effective as well since there’s no atmosphere, so the panels will be 30% more efficient or something like that than on land. It really does solve some problems, but it obviously creates others.
(edit: I worked it out once, and at the launch cadence of F9, they could put ~1GW of solar up a year, which would be pretty fast compared to land)
All that being said, while I don’t doubt they can launch these satellites, being profitable is an entirely other matter, but being profitable on land is also a huge unknown for something of these scales. Like, do they think there’s enough demand for these large quantities of huge datacenters? I’m not convinced this is a, if you build it, they will come situation for LLMs. True AI absolutely, but LLMs not a chance. So ALL of this IMO is betting these insane amounts of money and pollution on something they think they’ll eventually build, not something they have today.
At least with these land based ones if they take on building the SMRs or massive solar fields, we can repurpose those after. If they do it with natural gas, that’s much less ideal, sure the power might be useful, but we need to move away from it and that is the wrong direction. Space based ones will just flop with no backup benefit.
You aren’t thinking about the scale properly.
“some fields”, “parking lots”, “maybe some windmills” is a wildly vast under estimation of what would be required to do it.
7.65GW is A LOT.
The USA doesn’t even have an active 1GWh solar farm today. It’s 900MW, and it is 4600-5000 acres and is 1.3 million panels
To run batteries off that, you need to 3-4x that would be a 30GWh generation capacity
batteryto run in the non sun periods and give some buffer for bad weather. Then you need a massive battery as well to store it all. It would need to be able to run at least for a couple days, which means it needs to store 367.2 GWh (48h storage), so my solar numbers above are actually probably wrong since you won’t fill that up in 1 day, but maybe that’s okay if the size is just an emergency amount. 122GWh battery would probably the be minimum for a perfect summer condition 100% of the time day/night cycle.The largest upcoming battery installation in the USA is 1.15 GWh, which is not built yet.
I’m not saying don’t do it, but you’ve really glossed over the sheer scale of it, and I doubt they’d be able to easily put the solar field near the datacenter, so there’s extra costs around transmission lines as well.
Edit: For 30GW of solar you’re looking at around 153,333 acres, or about the size of Dallas Texas, before the battery storage
Edit: Updated battery figures
I concede that my grasp of the scale involved was unrealistic. However, I don’t see a viable solution. The issue, I have realized, isn’t how to generate the power, it’s the scale of power needed to run these data center monstrosities.
As you say, 7.65 gigawatts is a lot of power. According to ERCOT, Texas just set a new record for power demand in July of 91gw. So this plant, built to supply this one data center complex, will produce power equal to 8.5% as much power as the entire state of Texas required at peak demand this year (so far, anyway). 8.5% of peak demand for the entire state of Texas which, by the way, is the largest energy consuming state. Texas uses about 15% of the total power generated in the entire USA according to EAI.gov.
So if solar and wind (with batteries, of course) would be very challenging, it is at least vaguely possible. Texas is a great state for wind, solar, and geothermal power. It already generates about 30% of it’s power from renewables. Texas has a lot of sunny days and a lot of open spaces for wind farms particularly in the western parts of the state. I can think of worse ideas than covering all those oil fields with solar panels or wind farms. You wouldn’t even need to stop producing oil. You can do both.
Nuclear really isn’t viable. Starting the approval process now, it would easily be 20 years before a new plant would start producing power. No one is waiting two decades to open their data center.
The only other option to power these data centers, the one they are going with, is fossil fuels. That means multiplying the amount of toxic and greenhouse gasses being pumped into our air which is the absolute worst option for everyone on the planet.
(Lemmy failed to send my original reply and I lost it I think, but if another similar one eventually show up… that’s why)
I wouldn’t rule nuclear out in the not so far future if SMRs properly come online. I think those could scale up demand at the pace they could build these monstrosities in the not so distant future, like give it another 15 years and it might start being okay?
And ya, the renewable way IS possible, and Texas would be a good spot for something of this scale, but it would be a huge undertaking. I imagine we would see amazing improvements in solar/battery tech building something like that out with economies of scale and additional research from the huge amount of funding it would create. Also there’s no reason they can’t make them do some % with renewables.
This is aslo why Space based stuff maybe isn’t as far flung as the idea seems. It obviously requires new tech that doesn’t function yet.
e.g Starship needs to get down to its targeted $50-$200/kg) which is not a given.
New chips that can run hotter so the radiator size is reasonable. Cheaper solar panels as space based panels are very expensive right now.
But with that (and all the other things) you get 7.65GW of power being 7.65GW of power and minimal batteries required. Space solar is more effective as well since there’s no atmosphere, so the panels will be 30% more efficient or something like that than on land. It really does solve some problems, but it obviously creates others.
(edit: I worked it out once, and at the launch cadence of F9, they could put ~1GW of solar up a year, which would be pretty fast compared to land)
All that being said, while I don’t doubt they can launch these satellites, being profitable is an entirely other matter, but being profitable on land is also a huge unknown for something of these scales. Like, do they think there’s enough demand for these large quantities of huge datacenters? I’m not convinced this is a, if you build it, they will come situation for LLMs. True AI absolutely, but LLMs not a chance. So ALL of this IMO is betting these insane amounts of money and pollution on something they think they’ll eventually build, not something they have today.
At least with these land based ones if they take on building the SMRs or massive solar fields, we can repurpose those after. If they do it with natural gas, that’s much less ideal, sure the power might be useful, but we need to move away from it and that is the wrong direction. Space based ones will just flop with no backup benefit.