So you think they are 100% thermally efficient?
Even modern 2x1 (2 gas turbines, one steam turbine to capture that waste heat) combined cycle plants generate tons of waste heat.
There is an 800mw unit in my area that lost is make up water feed from the municipality, they only had 8hrs of pondage before they were coming off line, granted it was high 90s but still.
Also isn’t the proposed meta facility in LA about the size of manhattan, 2,250 acres?
Show me any area that could use that waste heat that has that much free space.
Maybe some industrial manufacturing that could be built next to it?
A lot is what you are willing to do. Their business is providing "data".
Every time you extract work from heat, the harder it is to extract more energy out of whatever is left, because it is lower temperature. It is pretty easy to extract work out of the heat from a gas turbine engine, because it is still extremely hot. But extracting the heat out a steam turbine is harder, because the exhaust temperature is much lower, but you can still extract plenty of work out of it.
You can run that heat through a low temperature steam turbine to capture more energy, or use it to run Stirling generators to capture a greater percentage of the heat. The heat out of THOSE can also still be used in something like district heating, or absorption chilling.
Depending on what you want to do to try to capture that heat energy, you could, in theory, come fairly close to getting to 100% efficiency. You'll never get there, but you could get close.
PROBABLY you can't get economically close to that. Are you going to locate a city near it and run steam pipes everywhere? I went to PSU, they had an on-site coal power plant, that also used the waste heat for steam heating campus in the wintertime. I'd imagine they were probably ballpark 70% or so efficient out of the input energy there in the wintertime (the way the steam lines and such were run, there was a lot of heat loss in the system, prior to actually heating buildings).
What kind of industrial processes are going to use that heat?
But if up-front cost were no object, I'd imagine you could install a combined cycle natural gas plant and then use the steam cycle waste heat for absorption chilling and probably get the total efficiency into the 70% range for utilizing input energy. Maybe even high 70% range. Long term, I'd bet that is probably also a cost savings, but Google, Tesla, MS, etc. mostly care more about upfront costs. How fast can they scale "AI" to beat their competitors. The long-term cost of operation is a footnote, especially if it delays construction and turning it on by weeks, months, or years.
Again, back to Grok and xAI's use of natural gas turbines that were NOT permitted, lied on permit applications, etc. They could care less about breaking the law if it gets them up and running sooner. Waiting months or a couple of years for grid ties, permits, etc. is not something they care about. Just like they could care less if it costs them tens or hundreds of millions of dollars more a year to operate. That is a worry for someone else in a couple of years, until then, just burn more money and hope we "win".
Heck, if they cared the most about long-term costs, they'd be pushing solar facilities with battery nearby. That is still typically cheaper than building new power plants and fueling them, even without any subsidies now. But that'd also cost TIME, possibly a few years to permit and eminent domain for transmission lines, buy/lease that much land for a GW scale solar facility.
Which is also why these facilities generally aren't doing any solar on site at all. They'd likely recover any input costs in a few years on the electrical generation, but it would delay getting them up and running by at least months. In slight fairness, I get that. If you told me it'll take me 18 months to construct a facility that was going to be generating a billion dollars a year in revenue and cost $175 million in electricity to run it, but if I added solar and batteries to the project it would save me a million dollars a year and delay the project by a month, I'd probably laugh at you and ask if maybe once it was up and running we could consider putting it in if it didn't disrupt anything. That month delay would be 80 years of revenue! Not including maybe $10 million in capital costs to install it too. Even if the capital costs would pay themselves back in a few years.
Same with things like cogeneration. Heck, that probably would save a ton more, maybe $20, $30M a month or more! But it probably would also delay construction by a few months, costing hundreds of millions of dollars in revenue. Likely also add tens of millions to the construction costs. And worse, delay "winning" by months. I might have lost because of that delay! Egads!
PS Despite how negative I am being about it from a business side, that is why Gov't exists and should bring down the hammer and say "not today". This is all socializing costs on everyone else, while privatizing the profits. This is drastically driving up costs on everyone else by slurping down power. If Gov't was stepping in and mandating best practices to reduce facility energy use, along with slowing it up to mandate that utilities increase generation this wouldn't be happening. But it would "slow" "AI" by months and months.
Oh no. In other news.
