cross-posted from: https://lemmy.world/post/53694832
The boom is set to produce enough greenhouse gases to cancel out 15 years of declines in power-sector emissions, according to a Sierra Club analysis.
We should be destroying them
What makes me really sad about this is they don’t have to be if they are engineered right, but the current government is all nah f that let’s burn it all down.
Maybe a new government can retroactively penalize carbon emitting data centers and force them to shutter until they get to carbon neutral. We really need to get aggressive on this.
Assuming they can actually get built
Assuming there is actually enough money to run them, or clients to buy compute.
Once all the AI startups have burned off their VC money, no one will be using AI data centers. The Koolaid is so strong that the plan to actually make money is to plan to use AI to tell them how to make money, because the currently have no clue.
We’re so fucked.
Why do you think Trump was trying to colonize greenland?
I wonder if there would be more public tolerance if new DCs were targeting land that would otherwise be considered horrible. I’m thinking places with historical industrial pollution unsuitable for other residential or commercial space.
How much complaints would happen if a DC was being built on top of a garbage landfill, bonus points if the DC taps the landfill for its methane to power the DC (many landfills simply burn off the methane in a flare wasting the energy)?
Would the public be against a DC being built on an old industrial complex polluted with mercury and PCBs? (the power question would still need to be answered)
I believe this is fixed now, but how about in historical Flint Michigan where the water being delivered was too polluted with lead to be consumed. A evaporative cooler in a DC wouldn’t care about that type of pollution.
How about near productive natural gas fields where a certain amount of excess ethylene is burned, and if there is still too much ethylene, the well production has to be halted. If DCs could take that excess ethylene to burn it, it would make the natural gas field productive again.
In short, instead of targeting GOOD land, GOOD resources, have them target UNDESIRED land, water, and power sources.
Brownfield development would be perfect for data centers, but it’s slower and costs more.
There isn’t undesired water or electricity sources available at anything like the scale the AI companies want though; these things use truly huge amounts of energy
There isn’t undesired water or electricity sources available at anything like the scale the AI companies want though
This is my second followup reply to the same circumstances described:
Lake Boehmer is a roughly 60-acre body of water in west Texas that sprouted from a former oil well that’s been allowed to leak since 1951. It produces 642 gallons of toxic salt and sulfured water a minute. It is undrinkable and simply spilling out on the ground. A 100MW datacenter consumes 367 gallons per minute, so not quite half of what is coming out of the ground every minute. Additionally the existing 60 acre lake can also be drawn from to increase the water consumption. There are also almost no one living out there meaning it would not impact the quality of life of residents.
West Texas is also the home of some of the largest wind farms in the USA producing 11.431GW of electricty. Wind, unlike solar, also has 24/7 generation, which more fits the model of a data center for power consumption. Solar could also be installed to supplement during daylight ours of course.
There is also lots of unused land out there. There are a number of datacenters already going up in west Texas so the know-how to build them already exists locally too.
Water that salty is corrosive. You can’t just use it for datacenter cooling (or running a power plant) without removing the salt at considerable cost. It’s cheaper to bribe politicians to transfer fresh water into their use than it is to desalinate that and use it.
Water that salty is corrosive. You can’t just use it for datacenter cooling (or running a power plant) without removing the salt at considerable cost.
Using traditional systems designed for fresh water sure, but there are methods to using high saline water already. Here’s the science . It also likely won’t surprise you to learn that the nations in the Middle East already has high saline evaporation cooling deployed in industrial applications for years now.
There isn’t undesired water or electricity sources available at anything like the scale the AI companies want though;
A solution doesn’t have to be 100% perfect to be an improvement.
Lets take an example of what I would consider a fairly advanced Landfill energy management site. Columbus Ohio SWACO (Solid Waste Authority of Central Ohio) operates a commercial landfill gas (methane) recovery operation. Today that methane powers the municipal buses, garbage trucks, and snow plows (in winter). Excess beyond the city needs is fed back into the natural gas pipes and sold at commercial rates to residential, commercial, and industrial customers. They’ve been operating this successfully for 11 years already.
The operation currently produces approximately 3,800 MBTU of methane daily. A high efficiency methane power plant (at 50% efficiency) would generate 557 MWh of electricity daily from that. More than half a terrawatt of electricity is certainly in the small to medium sized DC.
Further, the landfill produces 26 million gallons of leachate waste water. This is only a fraction of the water an Open Loop cooling system uses, but its a current waste produce, and its already onsite in the exact same location as the landfill gas which potentially could be used to offset other water usage.
Again, its not a perfect replacement for the land, water, and power, but its a very significant offset. A number of Ohio landfills are doing much the same as Columbus, and some are even supplementing with PV solar overtop of the landfill. If a DC was built on the Columbus landfill, the roof and remaining land could be covered with solar PV, and the recovered landfill gas stored during high solar production hours allowing even more gas available for electricity generation in the non-solar hours.
A single new datacenter in the pipeline from Amazon is going to consume above 7000MW. Landfill can and should be burned in useful ways, but it’s not going to supply anything like what the tech industry is planning to use.
Right, I addressed that. It doesn’t be a complete net zero replacement. DCs are going to be built. Any offset is an improvement on the status quo. Building on the landfill isn’t just about the electricity, its about reducing water consumption, land use, and impact on residents.
A single new datacenter in the pipeline from Amazon is going to consume above 7000MW.
That size of DC is the exception not the rule. Its the largest DC ever prospose in the USA. The currently largest built DC in the world is only 1000MW
Partial offset doesn’t stabilize temperature. It’s going to need to be 100%
They’re not even trying though
Partial offset doesn’t stabilize temperature. It’s going to need to be 100%
If you’re seeking a single silver bullet to the problem, you’re never going to find it. It will be a combination of solutions. My proposal is just one of them.
They’re not even trying though
There’s some trying, but not enough for my liking. Google has DCs are run by hydropower (with more being built). Microsoft has bought up an existing decommissioned nuclear power plant to put it back into service. Its not enough, but its also not nothing.
I’m not looking for a silver bullet. But part of success is demand avoidance. And creating demand at the scale the datacenters are doing is going to make it far far harder to get to zero.
