16
votes
Wind farms are building datacenters to offload electricity in non-peak hours
Link information
This data is scraped automatically and may be incorrect.
- Title
- Wind Farms Have a DARK Secret - Here's How we FIX It
- Authors
- Two Bit da Vinci
- Duration
- 12:11
- Published
- Nov 4 2023
Wind farms face limited profitability at certain times of the day due to variable electricity demand. Some developers are incorporating power-hungry datacenters into their wind farms to run only at times of low demand. At other times, the farm puts energy into the grid instead of the datacenter. It's a neat solution to the financial problem of renewable energy profitability.
I see the utility in this solution for datacenters that do reasonably useful things, like protein modeling for medical research. However, I'm less excited about use-cases like crypto blockchain computations or certain applications of LLMs, whose social benefits seem opaque. Separate from renewable energy in particular, I do worry that we are increasingly spending money to produce monstrous amounts of electricity for things that "turn a profit" in a financial sense but are not strictly beneficial to society.
The video also makes a few remarks about the high water usage of datacenters, which isn't something I've thought about much.
And it's not just wind - Microsoft is in the speculation and planning phases of using nuclear power to become the backbone of their data center and AI pushes - putting bets both on Small Modular Reactors and Fusion Reactors - https://www.theverge.com/2023/9/26/23889956/microsoft-next-generation-nuclear-energy-smr-job-hiring
But honestly, I'm just disheartened with the fact that Federal Research is at all time lows in proportion to the federal budget - https://www.aaas.org/sites/default/files/2022-09/Budget_Federal.png
For profit must chase things they expect to eventually be able to milk for much more than their investment costs.
But even in day to day life, I see so many returns on things I never expect to be repaid on. I invest my time in likely dead ends, or things I think might be one offs but want to do it right rather than doing it sloppily and possibly struggling in the future if it ever becomes relevant again.
Our future should not be decided by what's viable, as you never know when research into bug microbiomes could result in new ways to make bio reactors more efficient and safe like the research here on trying to optimize biome balance for the intended metabolic reactions - https://www.sciencedirect.com/science/article/abs/pii/S1385894721031685
Or finding out bumblebees like to play! Why could this be important? What possible profit could be gleaned from knowing we could set up bee attractions? But that's not the point! The point being that we're still vastly in the dark about the world around us - if bees can play, bees can think and learn, and we also learned that playing is a fundamental part of cognition, I believe being part of a reward mechanism built into psychology to encourage growth of pattern recognition and familiarization.
https://www.sciencedirect.com/science/article/pii/S0003347222002366
So, now we have a new set of possible test subjects for cognition research that are fast at reproducing and have built in controls for limited biodiversity within a sample population!
Termites we're still learning architectural design techniques from - https://www.bbcearth.com/news/what-termites-can-teach-architects&ocid=twert
You NEVER know what research has the potential to improve our lives, which is why we should always strive to ask questions we've never thought to ask, reassess assumptions, and never be afraid of hitting dead ends. But I feel like all of these are the exact reasons why capital based research is a horrible policy for the world at large.
You are absolutely preaching all the things I know and love about this.
So many people are screaming "AI WILL SAVE US ALL FROM WORK!" ... not with the private sector at the helm it won't. And with such uninformed, blathering idiots in charge that's unlikely to change.
I’m not going to watch this video, so I’m just going by other people’s summaries. But I’ll point out that the limiting factor is capital costs. Data centers and computers cost money whether you use them or not. If they only run part time, that’s less revenue and the cost is the same, so it might not work out to be profitable.
This all depends on the numbers. A natural gas “peaker plant” might only run when electricity demand is highest, but it’s still worthwhile because they’re relatively cheap. Similarly, solar panels are so cheap nowadays that it can make sense to over-provision so that peak power is higher than the grid connection can accept.
Over-provisioning is a form of “waste” that can make rational sense, but there’s incentive to try to do better, figuring out ways to improve efficiency. The amount of waste is proportional to capital costs and how much time the equipment is idle.
In real estate, this kind of waste happens in places that are very seasonal like beach towns and ski resorts. Figuring out some alternative use for housing and other infrastructure during the off-season improves efficiency.
For data centers that only run part time, it might make sense to move the computers somewhere else during the off season. I think that might have happened with Bitcoin mining and hydroelectric plants in China?
Bitcoin miners have been paying renewable energy providers like wind farms for their over-generated energy for a while now, curtailing what is often called the "duck chart".
Theres some papers on this:
I assume same will be found with large AI companies as well.
Unfortunately these papers are paywalled. What I can see don't explain a whole lot about what Bitcoin miners have been doing so far.
The first paper, from 2022, seems to be about making an estimate of what Bitcoin mining could do for Texas.
The second one is a highly jargon-filled modeling paper based on data from Texas in summer 2022:
I think what they're saying is that in the summer of 2022 in Texas, cryptocurrency farms didn't use electricity as much when prices were high. That's about what I'd expect. It's still not all that clear how good this deal is for Texas, though. How much did these businesses earn for not using electricity? I'm not sure if that's in the paper.
The third paper seems to be about what Bitcoin might mean for investors. They claim that adding Bitcoin to a portfolio would increase profits on average ("enhance the risk–return relationship of the portfolio") and reduce the portfolio's carbon emissions, which I'm not sure is a real-world effect I care about. It's unclear what data they used from the non-paywalled portion.
We recently discussed an alarming news report about 2023. It would be good if someone analyzed that case properly.
It was Microsoft that successfully tested submerged data centres, there must be some available synergy if you can sink a few tanks and erect some turbines in close proximity, I'd have thought.