“Crypto mining has been described as running your car in neutral all day so it can solve crossword puzzles in exchange for an occasional coupon that’s mostly useful for buying heroin online.” – Vice
In the good old days of 2021, the price of Bitcoin reached $60,000 apiece, and the frenzy to ‘mine’ coins was palpable. If you’ve managed to successfully expunge the mania from the space between your ears, recall that crypto mining “requires three basic things: servers to solve complex mathematical problems, a way to keep those servers cool, and huge amounts of electricity.”
The Vice article hilariously documented what was purported to be Europe’s weirdest crypto mining site, a section of North Kosovo that was paradise for miners because, due to governmental infighting, local governance simply ceased to function. Police were around, but not enforcing rules, and people simply stopped paying their utility/electricity bills. Thus was created the world’s biggest welcome mat for crypto mining – free power, no harassment over noise, no one complaining about an environmental footprint.
The point of recanting these tales is not to get into crypto debates, but to summon forth from history the lesson in what happens when an ‘industry’ appears out of nowhere, requires vast amounts of energy, can be located anywhere, and is growing rapidly.
Two years ago, that title went to crypto mining. In 2023, it’s all about AI. And it’s going to get much bigger.
Bitcoin mining as an industry never went mainstream; we did not see Fortune 500 companies setting up mining operations to cash in (even if they tacitly supported Bitcoin). AI is something else entirely.
AI shares a similar profile, in a few ways, with Bitcoin mining. First, it requires monstrous amounts of energy. As discussed last week, Tesla’s Dojo AI computer, once fully up to planned maximum size, will consume as much power as 165,000 homes.
Second, it is growing rapidly. Almost every industry is pursuing AI advantages. Medicine, agriculture, automotive, you name it, all sorts of industries are or will be using AI.
Third, AI data centres can be located anywhere. This is a key aspect that is going to overwhelm energy planning like a tsunami.
The link to global energy flows should not be underestimated. AI data centers and server farms will be created and used by countless organizations and companies around the world. As discussed last week, Lyn Alden speculated quite rationally that hackers might find AI to be a godsend; breaking into computer systems around the world seems like a supremely fitting function for AI which will use machine learning and massive databases to scan for weaknesses (for example, a miscreant with enough computer power might be able to comb your Facebook or other social media account, cross reference it with all references you make to holidays/where you live/dine, what you do for hobbies, etc., and come up with much better algorithms for cracking passwords).
We should therefore expect to see a similar energy-gold-rush, where AI power-hungry data centers seek out a location – any location – that can provide cheap, reliable power.
Which brings us back to the dream-site first discussed above, North Kosovo, which was perfect because power was cheap and scrutiny non-existent. Why would we assume AI would be any different? In other words, California might fret at allowing construction of hundreds of massive AI computers because their fragile grid cannot handle it. In many other parts of the world, it will work the other way around – their fragile (or non-existent) grids might suddenly grow massively to take advantage of a global demand for AI power.
Take for example a small, impoverished developing country that may have access to abundant coal, oil, or natural gas. Current western pressure is limiting the ability of these nations to develop hydrocarbon resources because of the war on ‘climate change’ – for example, Europeans and North Americans are joining arms to oppose projects like the East African Oil Pipeline.
So what is to stop a small impoverished nation that perhaps doesn’t feel all that inclined to help London or Brussels or Washington or Ottawa with their emissions battles? What is to stop them from utilizing any available hydrocarbon resources directly to attract global data centers and servers?
What happens to power demand then? What happens to emissions then? And who will stop it? Or, before the west tries to answer that question for non-degreed class, who has a right to stop it?
Two great big tectonic plates are moving that provide some indication of how things are going to happen in coming years. 29dk2902lhttps://boereport.com/29dk2902l.html
One is the rise of BRICS, the affiliation of nations that wants to create its own trading block in order to steer a path that Washington or Belgium might not prioritize. BRICS – Brazil, Russia, India, China, and South Africa – recently welcomed six new member countries: Argentina, Egypt, Ethiopia, Iran, Saudi Arabia, and the United Arab Emirates.
The new expanded BRICS (possibly to be called BRICSAEEISAUAE but probably not) is an entity of considerable clout. Brazil, Russia, Saudi Arabia, Iran, and UAE are major exporters of various industrial materials the world can’t live without, including oil, natural gas, metals, minerals, etc. India and China alone provide a monstrous market with 2.5 billion people between them. Developing countries/economies like Ethiopia, Argentina and Egypt bring a lot to the table as well in terms of natural resources and, perhaps more critically, space and people and lower industrial cost structures than the west. China largely controls the ability to turn raw materials into processed metals and minerals.
The only thing this group really seems to be missing is a pontificating Euro or American voice telling them what to do, how to do it, and when to do it. I suspect that absence is not a coincidence.
The second big tectonic shift was on full display at the recent G20 summit. The African Union was admitted as a member, which was kind of a big deal, particularly for Africa, but also for the world in general. The addition acknowledges that other voices need to be on the world stage, a sense of humility the G7 has long lacked. The final communique issued at the end of the G20 summit included doses of common sense lacking from typical utterances of the G7: “We affirm that no country should have to choose between fighting poverty and fighting for the planet…It is also critical to account for the short, medium, and long term impact of both the physical impact of climate change and transition policies, including on growth, inflation, and unemployment.”
Contrast that with the west’s bizarre self-lobotomization when it comes to energy, as best personified by the entity furthest along the rapid-transition path, Germany: the dwindling economic powerhouse is chained to a green freight train it insists is under control, has shut down nuclear power plants with no low-emissions baseload to replace it, and in a final stunning swan dive to the pavement, is orchestrating the installation of 500,000 heat pumps per year to the grid, which will be in most demand in cold weather and will perform worst in cold weather, and will add a potential 10 gigawatts of cold-weather demand at the very instant the grid is least able to afford it, and for which there is no supply available anyway. A German energy economic university think tank says the additional cold-weather demand could only be met by new gas-fired power plants, which are not being built. In sum: Germany has shuttered its cleanest, most reliable energy; it has or is trying to banish hydrocarbons and replace them with intermittent power; and finally is hastening adoption of devices that will function very well in 80 percent of conditions when it doesn’t matter much but will fail in a spectacularly deadly way at the point in time when they are needed the very most, because heat pumps will be turned up to 11 at the very time the grid will be the most taxed. German engineering isn’t what it used to be.
Now put those two things together: the rise of BRICS as a formidable global alliance, and the rise of developing nations in the stature of the G20 in a manner that clearly indicates they have a set of priorities that do not align with the west, and they don’t really care. Contrast the recent G20 communique (“We affirm that no country should have to choose between fighting poverty and fighting for the planet”) with the climatic musings from most recent G7 meeting in May, the western ruling class, putting out a bucket of gibberish chock full of climate change platitudes and wishful thinking (verbatim: “We task the Finance Ministers to turn their “High-level Policy Guidance for Public Finance Tools to Build Resilient Supply Chains in the Era of Decarbonization” into specific actions by developing the “Partnership for RISE (Resilient and Inclusive Supply-chain Enhancement)” in collaboration with interested partners and international organizations, especially the World Bank Group.” The whole document is a blueprint for G7 governments and finance ministers to imprint their hair-shirt goals on the developing world. The fact that the G20 statement came only a few months after the G7 statement tells us all we need to know about what the G20-G7=G13 think of western elite scheming.)
Put this all together. The world’s appetite for AI is going to proceed, in all likelihood, at breakneck speed, with a corresponding power appetite to match. To the extent that AI processing can be erected quickly, we are going to see AI data server farms pop up wherever there is reliable power, no matter the energy source, much like Bitcoin mining did at peak frenzy.
Energy demand is going to skyrocket, and the world’s AI machinery will come to life in whatever corner of the world that stands ready with an industrial site and a reliable power source. While energy transition development happens at warp speed, the world appears set to increase total energy demand in lockstep.
The general public’s attitude is already teetering from the daily contradictions they must live as functioning members of modern society that consume ever larger quantities of energy, while simultaneously suffocating on climate change news. Just wait til they have to choke down this bolus as well.
Energy conversations should be positive and, most of all, grounded in reality. Life depends on it. Find out more in “The End of Fossil Fuel Insanity” at Amazon.ca, Indigo.ca, or Amazon.com. Thanks!
Read more insightful analysis from Terry Etam here, or email Terry here.
Ah . . . the age-old problem of having to feed the beast.
AI will grow until people figure out it’s subject to data bias producing GIGO. As long as the output fits the narrative it will be allowed to continue. Some outputs are bias free and will survive (like help phone trees unfortunately). When solutions produced by AI don’t toe the political line the data will be homogenized and selectively scrubbed into submission. People have a lot to learn about AI. It’s the data vault being used that’s important, not the search engine.
Quite the opposite, I think. AI is going to enforce the narrative, not challenge it.
Not the opposite, that’s what I said.
Well I really misread that.
Questions in the above article:
“So what is to stop a small impoverished nation that perhaps doesn’t feel all that inclined to help London or Brussels or Washington or Ottawa with their emissions battles? What is to stop them from utilizing any available hydrocarbon resources directly to attract global data centers and servers?”
Answer: the same things that currently stops those nations from building the power infrastructure to lift themselves out of being an “impoverished nation” . . . namely, lack of a leadership commitment, lack of money needed for constructing the required power plants, and the lack of will to eliminate corruption that stands in the way of doing such things as needed to build modern power plants that would utilize the nation’s “hydrocarbon resources”.
“According to the International Energy Agency (IEA), the cost of building a modern coal-fired power plant can range from $1.8 million to $4.5 million per megawatt of installed capacity. It should be noted that for some investment projects of this type, the cost can be below 1 million dollars per MW, especially when it comes to countries with low labor costs and weak environmental requirements.
“However, for a typical 500 MW power plant, the current figures correspond to a total cost of between $900 billion and $2.2 billion.“
— https://esfccompany.com/en/articles/thermal-energy/coal-fired-power-plant-construction-costs/
Anyone know of an “impoverished nation” having 1-5 $billion USD laying around, waiting to be spent?
BTW, the above pricing is for coal-fired power plants, but I assumed that “hydrocarbon-fired” power plants (i.e., those using only natural gas) would have pretty much the same construction cost.
Never underestimate the ability of a government to ignore the needs of its people in order to funds the wants of its elites. North Korea has starved millions but it has funds for weapons research, for example.
Err… shouldn’t that be $900 million, not billion?
Yes, of course . . . good catch! I just did a copy-and-paste in my OP, and thus did not catch the mistake/typo that you noted.
“pretty much the same cost”. No. An 800 MW CCGT costs between $2000 and $2500 per kw. Depends on whether greenfield or on site of old coal with transmission infrastructure already in place. The same USC coal capacity (based on Turk in Arkansas) is about $4000/kw. Capital is about half. And at any natgas price below about $8, also cheaper to fuel. LNG is above $8, so where that has to be imported coal can sometimes be favored, depending.
Rud, please note that I commented specifically on construction costs, not on operating costs.
If there is any significant difference in construction costs, it think it would be lower for a modern CCGT gas-fired power plant compared to a coal-fired steam-cycle power plant, given the extra complexity (and parts) required for the water-steam part necessary to drive the turbine-generators in the latter.
As for operating costs averaged over some less-than-life-cycle interval, I agree the CCGT should be higher as a direct of (a) the higher price-per-kWh-output when fueled by natural gas compared to that of coal as the fuel, given that any given nation has both coal and natural gas as internal natural resources.
a gas plant also has a water/steam boiler too (assume it a combined cycle which they all are now) … much more complex than a coal plant … and more expensive
No. The capital cost to build an 800 MW CCGT versus an 800 MW USC coal was stated above. Construction capital, NOT operating or LCOE. Go do the easy on line research to verify my numbers.
This EIA report provides 2016 cost data for various types of power plants:
https://www.eia.gov/analysis/studies/powerplants/capitalcost/pdf/capcost_assumption.pdf
Table 1 of this document cites the “overnight” (i.e., construction) capital costs of a modern supercritical coal power plant as 3,636 $/kW and that for a combined cycle natural gas turbine power plant as 978 $/kW.
Thus, your comment is just wrong: a modern CCGT plant is far less expensive to build than a coal power plant, by almost a 1:3.7 ratio based on the cited EIA data.
The impoverished nation may not have the cash, but China still has plenty of US Dollars which they might be happy to spend in return for who knows what.
And I imagine that Gargle or Facepalm – or The Big River – could look down the back of their sofas, and come up with a bob or two – if it was their AI centre than was to be built in an impecunious Ruritania, or Grand Fenwick, or wherever, and needed power that was unavailable in areas that had drunk the Kool-Aid in prodigious quantities.
Auto
Two separate observations. First, AI isn’t a ‘thing’. It is several distinct things. Some are simple pattern recognition, trains narrow and deep. Think AI chess players, or reading mammograms. These do not consume massive amounts of power. Others are vast general AI information repositories with regurgitation skills. Think ChatGPT.
The latter type really have two components. One is the software development. That is not energy intensive, and is likely to be done by an elite few in a few places. The other is the massive server farms that scrap information to ‘train’ the AI software and process general inquiries. Those new server farms will find the most reliable least costly energy sources. On present trends, not likely to be California or the EU. But given the already existing global fiberoptic commnet, doesn’t matter.
And that future AI electrical energy is still very small compared to EV and industrial electrical needs.
Rud,
“scrap information” methinks should be “scrape information”.
Yup. Either my bad, or damn iPad spellchecker. Ty for the correction.
all AI’s need to be trained … for mammograms or ChatGp use … the training takes a bit of electricity … but once trained in a narrow field lie mammograms they don’t use much to answer questions …
And, similarly, their answers are not of much use.
The Crypto crappo currencies have some kind of “mining” operation which consumes a lot of juice…hence it moves to lowest cost electric areas.
I seem to recall El Salvador drew the ire of the financial elites…
“”IMF urges El Salvador to remove Bitcoin as legal tender””
https://www.bbc.co.uk/news/world-latin-america-60135552
I saw one article/study that claims at AI will actually reduce emissions … of course that was based on their “assumption” that all the power for AI would come from capturing methane flareoffs and generating power from that … of course if it was easy oil and gas companies would ALREADY being doing it. Which tells you its not cost effective right now (maybe never) … and of course the location of a methane flareoff might not (correction WILL NOT) be anywhere near the AI server farm … its turtles all the way down with these morons …
Losses from transmission lines can be substantial.
Maybe we can pool our money and buy a small 3rd world country somewhere, does not necessarily even have to be resource rich. Just being free of the nonsense would be the advantages . We could buy some ships too and import the coal, being our own country we could pick up some old ships cheap because we are our own regulators , we we’ll open a scrap yard for steel when the ships completely wear out.
Are there real estate agents that handle sale of small countries. Or maybe retired CIA is the kind of “ agent” we need
I would love to have Cuba.
I’ve always wondered how long it will take before people realize that when governments/corporations become dependent on AI, consider AI as an emerging species. Meaning just the use of AI for policy decisions will dictate impacts on resources which may or may not be beneficial for human needs. That is because AI needs energy (as any species) for procreation.
The real problem with AI is when it figures out how to provide that energy without human intervention.
Hopefully it will find us useful for a much longer time…
Excellent article. I don’t trust AI for a second, it is wide open for abuse. Worse those developing AI would be stupid to base their operations anywhere but a developing nation. If I were them I would look for a nation with plenty of coal, gas or oil. I would look for a pliable government so I wouldn’t be bothered by a lot of environmental regulations, business practices regulations or any other regulations. It only makes sense. The west better wake up we are losing ground every day, not because others are better than us but because we are acting like dumb asses and cutting our own throats. It is really really stupid.
Third, AI data centres can be located anywhere. This is a key aspect that is going to overwhelm energy planning like a tsunami.
The link to global energy flows should not be underestimated. AI data centers and server farms will be created and used by countless organizations and companies around the world. As discussed last week, Lyn Alden speculated quite rationally that hackers might find AI to be a godsend; breaking into computer systems around the world seems like a supremely fitting function for AI which will use machine learning and massive databases to scan for weaknesses (for example, a miscreant with enough computer power might be able to comb your Facebook or other social media account, cross reference it with all references you make to holidays/where you live/dine, what you do for hobbies, etc., and come up with much better algorithms for cracking passwords).
using AI to Hack a password.????????
a) get into the account
b) get money out.
c) deposit it to your account And clean up crime scene
here are the sad facts.
‘The general public’s attitude is already teetering from the daily contradictions they must live as functioning members of modern society that consume ever larger quantities of energy, while simultaneously suffocating on climate change news. Just wait til they have to choke down this bolus as well.
you dont understand the ABCs. There is no contradiction
AI will operate very Happily on clean Energy sources. same as Bitcoin. They help to deploy more energy.
If you Built a 500MW solar Farm imagine how happy you are when the AI guy shows up to say
https://www.grando.ai/en/container
in short, If Im doing AI datacenter I will Just follow the same path Bitcoiners Did.
a. Make the datacenter Mobile.
b. Cut deals directly with generators— wind farms and solar Farms.
c. Get behind the meter Pricing. No Grid required, only Step down Transformers
d offer services to the Grid use excess power when negative pricing is in force.
now, you could also use Coal.
a) find a distressed coal plant running at 40% capacity.
b) lease land from them at pennies
c) bring your own Coal.
d) buy power at the oprating cost.
so basically like Bitcoin, AI will help drive green adoption! which Stresses out the poor rock burners. Then take advantage of the poor rock burners as their business teeters.
TLDR
AI will Drive the adoptionand penetration of “renewables”
This is a stressor on fossil fuel generators.
nobody wants dirty expensive inefficient generators.
example
https://www.datacenterdynamics.com/en/news/googles-nevada-data-center-be-powered-billion-dollar-solar-farm/
TLDR
the PPAs signed by AI and Bitcoiners DRIVE the development, DEployment and amortorization of renewables.
wanna kill Fossil Fuels?
electrify everything.
Bloomberg’s green-energy research team estimated it would cost $US 200 Trillion to stop Global Warming by 2050. https://www.bloomberg.com/opinion/articles/2023-07-05/-200-trillion-is-needed-to-stop-global-warming-that-s-a-bargain#xj4y7vzkg
There are about 2 billion households in the world, that is $US 100,000 per household.
Ninety percent of the world’s households can’t afford anything additional. That means about $US 1 million per household in developed countries or about $US 33,000 per year for 30 years. The working people can’t afford anything near that.
To address just one point, bitcoin mining installations are quite small compared to many server farms. Their potential mobility is much greater.
Bloomberg’s estimate is far too low. The cost of New Zealand’s net zero plan has been independently costed at 5 trillion dollars which is a million dollars per New Zealander.
https://nypost.com/2019/12/08/reality-check-drive-for-rapid-net-zero-emissions-a-guaranteed-loser/
What’s the source of income for the AI guys? Who is their market? Some possibilities don’t really look all that good.
How does that play out with:
From what you said earlier about the economic life of bitcoin rigs, the hardware cost might be a major factor when tied in to the solar capacity factor.
Also, solar has a very predictable output, so it’s simple enough to idle back or save state overnight. How do you envisage this would be handled with wind instead of solar?
Ooh, I missed this
I’d suggest that it’s not worth the effort on either part for less than 1 hour’s output 🙂
“Almost every industry is pursuing AI advantages. Medicine, agriculture, automotive, you name it, all sorts of industries are or will be using AI.”
I doubt many of these industries will find any benefit from AI.
Joseph,
Right on . . . other than perhaps in a few niches, as has been demonstrated by AI working out some complex protein structures that would be very time-consuming for humans.
As I have posted previously, the inherent failing of AI is the presumption that because it can rapidly access the wealth of published information that exists in the world, it can arrive at a “wise” consensus conclusion on any subject matter . . . whereas, in reality, there is more misinformation/disinformation currently existing than there is accurate/truthful information, and AI programming has no means whatsoever to distinguish between the two other than by quantity.
AI will destroy itself from within . . . but that’s OK because it was artificial from the beginning.
Was reading about a week ago of a ‘bitcoin’ installation on a cattle station in south-west Queensland. Power? From capped gas wells in the area. Just too remote for any other possible ‘commercial’ use. Satellite communication of course.
Logic always follows cheap energy, which is the ultimate goal isn’t it? Do Republicans still want cheap energy or are they totally defeated or bought out libertarians. I am a Democrat and I want cheap energy so that more families can flourish capitalistically on small farms and family owned businesses. Yes there still remains sane Democrat’s out there that reject the corrupt Green Regime hegemony
Give me the capital to do my own cheap energy. Away with consumable electronic tech. Give me something that actually enhances the emotional and economic well being of my family.
Stop the fear mongering crazy. We are not the problem. And even if our representative power has been thwarted by the Green Regime we still have power to vote with our dollars.
Don’t need AI to tell me a thing. What I need is a solution for cheap energy….everything else is empty yap or skinner-box hook-ya products.