Ethereum Energy Consumption
ethereum.org
ethereum.org
How many years has it been, at this point? I distinctly recall "imminent" plans to switch over to PoS as early as 2019.
The latest test merges have gone very well, but when you're dealing with something this valuable, you do a lot of tests, just to be sure.
What I recall is a number of public announcements over the years claiming that the PoS transition was imminent, only to have it delayed at the last moment. That kind of repeated whiplash is not confidence inspiring, either from a design or a community standpoint. Is there a reason to expect this time to be different?
Well...
- five+ different consensus clients, all with a full PoS implementation and a working PoW -> PoS transition mechanism
- four different execution clients, all with a working PoW -> PoS transition mechanism
- Multiple (dozen+) of successful merges on testnets and shadow forks. Ropsten, an ancient test network, was successfully merged to PoS from PoW a couple weeks ago.
There's two more existing testnets that will be merged soon. Sepolia on the 5th of July, and Goerli on some date TBD. If those are successful and there are no major issues found, it's time for mainnet.
[0] Not even 24 hours ago: https://news.ycombinator.com/item?id=31954276
If you're talking about when it started being claimed that it was happening soon, the beacon chain launched at the end of 2020, so probably 2019/2020 sometime.
https://news.ycombinator.com/item?id=27194586
They keep making the same announcement every few months.
How is "[using] dramatically less carbon" more secure? What are the security implications?
> Since its inception, Ethereum has aimed to implement a proof-of-stake consensus mechanism, but doing this without sacrificing security and decentralization has taken years of focused research and development.
A rather notable criticism of PoS, and its variations, is that it /does/ sacrifice decentralization to achieve its aims. I know that much smarter people than me have been working on this, and I don't follow research into consensus algos super closely, but I haven't heard or read anything validating PoS as the silver bullet it's made out to be.
PoW burns a lot of energy, but the algorithm has an elegant simplicity to it. Crucially, the _work_ is impossible to fake.
PoS (very roughly) means that the richest control the network... which seems reasonable. But suppose you decide to rewrite the blockchain to say you're the richest, do you then control the network? It's a circular loop, and defending against those kinds of attacks is dramatically more complex. I've been lightly following development and I get the sense they've been playing whack-a-mole with vulnerabilities and bugs since ~2016. It's hard to have confidence in the result.
Coming back down to earth, this is why ETH requires checkpoints[1], but PoW chains do not[2].
[1]: https://ethereum.org/en/developers/docs/consensus-mechanisms...
[2]: https://bitcoin.stackexchange.com/questions/75733/why-does-b...
According to this, the PoS schedule first slipped in 2017. (search for "snooze button")
The production staking network was launched in December 2020, was upgraded in mid-2021, and since then there's been extensive public testing of the merge with the rest of the network.
I've heard the argument that if you take into account all the human-associated energy costs Bitcoin's energy use compared to a company like Visa would be comparable. Now I'm not so sure.
Notably this 0.26 TWh/yr number doesn't include things like the energy used by each human to drive to and from work or the amount of energy needed to make the food that fed them that year but now we're getting into some anti-human territory.
Perhaps the goal shouldn't be to minimize energy use but to maximize energy output.
1. https://app.impaakt.com/analyses/paypal-consumed-264100-mwh-...
Even the most charitable argument, that crypto-relieved folks are trying to get governments involved.
Likewise financial systems have little to do with protecting fiat, banks are backed by governments businesses like paypal hold money in banks, etc.
So to compare bitcoins energy consumption to an entity that spend energy in an unrelated way isn't entirely fair - there is some merit there, but not much.
Bitcoin's energy use contributes no more towards preventing armed antagonists destroying or seizing assets, harming their economic value by destroying the underlying economy or compromising the network by seizing ASIC farms than Visa's. Bitcoiners outsource that bit to governments too.
Bitcoin can perform without a government - the involvement desired by some people is not at all, nor is it claimed to be, or understood by anyone but you, to be necessary to the healthy functioning of Bitcoin or other sufficiently decentralized open protocols.
You do not get stable fiat without a trusted central authority and you do not get a trusted central authority without military power.
By the way, bitcoin is also "fiat", in the sense that it isn't backed by anything (this is what "fiat" means).
That's not what fiat means in any sense. It has latin origins for "let it be done" which is to say that it is a decree by an authoritative actor.
But both bitcoin and fiat are backed by something - bitcoin is backed by a lot of energy consumption and fiat is backed by government authority.
> Do you mean that private property is enforced by violence? How else do you expect to enforce it?
I don't expect it to be enforced any other way, I didn't mean to give the impression that bitcoin was somehow free from the necessity of enforcing private property. However, bitcoin is different in that you only need to protect people and the network, with fiat you have much more to protect, such as political positions of power and physical buildings like fort knox.
The OLG models with money attribute to fiat money two basic characteristics: (a) it is intrinsically useless – that is, it cannot be directly used in production or consumption 3 – and (b) it is inconvertible 4 – that is, the issuer does not give a commitment to redeem it into commodities.
Handa, J. (2008). Monetary economics. Routledge. (p. 719)
Bitcoin is a fiat currency, just like the US dollar. And you're also wrong that with bitcoin "you only have to protect people and the network". What? Bitcoin itself can't be consumed (because it's fiat, remember), therefore it's worthless unless it can be exchanged for goods and services that can be consumed, which means these goods and services also need to be protected.
Both bitcoin and USD (as an example of traditional fiat) are absolutely backed by something - they are backed by quite a large economy, for one.
I'm not sure if you intentionally left this out, but following your quote literally just one sentence further you get:
> Further, it is assumed that fiat money is costless to produce, to store and to transfer, and does not pay interest to the holder
Which is quite obviously not a definition that fits bitcoin.
> I'm not sure if you intentionally left this out, but following your quote literally just one sentence further you get
It says that these economic models assume fiat money costs nothing to produce, which is not accurate, because it does cost money. However, the production cost is not relevant for the purposes of these models, and therefore they can make this simplification and assume that it costs nothing.
It definitely doesn't, at least not the way you're suggesting - you can find article after article talking about oil backed USD. Further, you can find plenty of definitions for the gold standard that don't use the word "backed" because it's not a precise term.
> It says that these economic models assume fiat money costs nothing to produce
You can either use this models parameters as a definition for fiat or as simply parameters for the model, you don't get to pick and choose only the parts that support your argument while ignoring the parts that explicitly don't.
Regarding your other point, I'm not sure what you're trying to say. Are you saying that the dollar is not fiat money because it costs something to produce?
To be honest, I don't care. I have nothing else to add to this conversation, therefore I'm out.
The protection of private property without violence may be considered the innovation behind Bitcoin. Big deal.
Gold and bitcoin aren't debt, they're assets. You could issue bank notes backed by gold or bitcoin.
The amount of misinformation around crypto energy usage has been out of control in the past few years. It's still weird to dip into an HN comment section and see people spreading the false claim that mining is powered by renewable energy sources that would otherwise go to waste (an absurd claim on every level) or that Bitcoin is somehow a "store of energy". The extreme energy inefficiency of cryptocurrencies is well-documented and doesn't even come close to any other financial system.
> Perhaps the goal shouldn't be to minimize energy use but to maximize energy output.
The real goal is to minimize harmful emissions. If we had some hypothetical infinite clean energy source then mining wouldn't be a big deal. However, even when mining is colocated with renewable energy sources, that renewable energy is prevented from going into the grid and offsetting fossil-fuel power plants.
I’m not sure that’s how energy markets work… do you have any further reading? I’ve heard generally miners buy the cheapest electricity possible—typically oversupply that would otherwise be curtailed.
I don’t understand why Bitcoin mining is incompatible with emissions reduction. It sounds like an additional source of demand may make economical plants that otherwise could not be built.
Some years ago, there was a book published where the author systematically looked at the energy consumption (not just the of the UK and compared it to the total possible renewable energy sources therein. He found that even if you covered every hectare of the UK with solar panels, sent every raindrop through hydro power plants to extract energy as it runs to the sea, etc., it still would not provide enough renewable energy to satisfy UK consumption.
Switching existing energy sources to renewable isn't sufficient to avoid climate change; we need to reduce aggregate energy consumption. And that means that Bitcoin mining, adding a massive energy sink where one did not previously exist, is doing the exact opposite of what needs to be done.
In the book he tries to determine if renewables could support UK's energy needs. He estimated the UK consumes about 195 kWh per day per person (food, transporation, heating, miltary defense, ...)
If the UK massively invested in all the renewable options it stacked up to 180 kWh per day per person (obviously he provided lots of caveats).
His assumption on Solar PV was covering 5% of the UK land with 10% efficient PVs (about 200m2 per person). He estimated that could contribute 50kWh/d/p.
In the 14 years since the book was published it seems like Solar PV efficiency is now commonly 17-19%. So might only need to use 2.5% of the land.
Believing that taking into account human-associated energy costs would put them on rough parity would require assuming that computers take up 1/1000th the total energy costs. And that's before you actually look at how many human jobs would actually disappear were cryptocurrency to magically replace the financial industry.
That is a red herring, as what secures the U.S. dollar is much more than individuals eye testing physical currency.
Paypal and any other company is interested in efficiency. Servers, air conditioning and power all cost quite serious amounts of money. UPSes and backup generators have to be sized for the load. Bringing large amounts of power into a building is also non-trivial. All of this means that most companies want to be reasonably efficient.
If Paypal found that their needs are covered by half the servers, leading to shutting down half their datacenters, they'd be quite happy about it. Plus less wages to pay!
Proof of Work on the other hand intrinsically wastes power. If somebody made a 2X more efficient ASIC/GPU, that'd mean that mining is now cheaper, which means you can do twice of it, which means that quickly everyone scales up and power usage remains the same.
The output we want from a blockchain is capacity to do work, that is transactions.
PoW means you're burning more and more energy, while the amount of work the network remains constant, because none of that energy actually results in more useful work being done. Thus the efficiency of PoW chains goes down and down and down.
Correspondingly, if I buy a washing machine, I can do one load of laundry per hour. If I buy a second, I can do two loads per hour at the expense of more water and electricity. If washing machines were powered by proof of work, adding more washers and using more water and power wouldn't increase the number of loads I can wash per hour. The algorithm would see that too many loads are being done per hour and decrease the efficiency of each washer to compensate.
If they work from home? How would you even measure this energy? They have to power their home and feed themselves so that they can do their job to keep the system running. Is that taken into account?
Surely the crypto numbers don't include the same figures for folks working on all of the Ethereum infrastructure. PayPal employs 31000 people. Surely there are more people globally working on Ethereum (in mining facilities, writing code, etc).
Proof-of-work isn't used just to add cost, but to make sure that the longest chain has longest history in physical time, and make it objectively verifiable.
Proof-of-stake doesn't have this property. Whole history can be recreated in a microsecond, and there's no way to tell by looking at the blockchain. Instead, network participants have to ask others for the correct chain, or refer to a trusted source when there are disputes.
- For starters, if say I'm accepting a BTC payment, I need to ask a trusted source about the market value of BTC.
- I need to ask which software I can trust to implement the Bitcoin protocol.
- If I don't want to verify every transaction from the genesis, I need to ask for a checkpoint.
- In case I'm being eclipsed, I need to ask whether the chain I'm seeing is the canonical one, or what approximate hash rate the canonical chain should have.
PoS doesn't have this single truth, and thus it's incapable of converging on it. "call a friend"-consensus fails when there are disputes.
If you want to use something, this is much more credible from Cambridge University (showing radically different numbers): https://ccaf.io/cbeci/index
For an example on what digiconmist gets wrong (there are many!): https://bitcoinmagazine.com/business/not-science-digiconomis...
The Cambridge University estimate for the Bitcoin network is 92.93 TWh with a lower bound of 51.01 TWh and an upper bound of 153.61 TWh. (I don't see an Ethereum comparison on the Cambridge site, so I can't check those numbers)
Given that the Digiconomist estimate is within the bounds of the Cambridge estimate, I don't see how this is really a debunking. I'm not really inclined to put too much faith into an article from "Bitcoin Magazine" on the topic when a quick look at the numbers shows that the two primary sources don't really disagree and have similar estimates.
Just incredible.
Not trying to be snarky, just generally intrigued.
The first is that there’s still no date for The Merge beyond “expected to happen in the second half of 2022.”
Secondly, the Ethereum organization can only ask people nicely to stop mining the POW version of the coin. It’s possible that the current version of the Ethereum blockchain will continue to progress, like Ether Classic has, albeit at a greatly reduced value.
https://github.com/ethereum/pm/blob/master/Merge/mainnet-rea...
Notably, there is only 2 testnet merges left to do. Ethereum has 3 testnets to try things out. One testnet has already merge and is running on PoS (Ropsten testnet). The merge was a success although it allowed to uncover some minor issues. The next one, called Sepolia, is happening this week. After that, there is only one other testnet merge to occur, likely in August. If all that goes well a date will be set.
Ethereum does have a way to force the death of PoW. The Ethereum protocol has always built-in the so-called "difficulty bomb". The difficulty bomb is simply a target date in the future when the mining difficulty starts rising exponentially. Making finding new blocks exponentially harder until the chain freezes to a halt. Its death is always coded-in, so to speak. The difficulty bomb needs to be delayed periodically so Ethereum can keep working. Last month the difficulty bomb was pushed a few months to occur in September. If all testing goes well, the merge will be triggered in September aligned with the difficulty bomb which will cause the death of PoW Ethereum.
You could argue that you can always fork Ethereum, remove the difficulty bomb, and run that instead. Which is true. Except there won't be economic value in that chain, so there won't be fee revenues to capture which means it won't be profitable to point your hash-power to this forked chain.
I'm much more bullish on Chia and proof-of-space/time — same nakamato consensus as proof-of-work without the energy consumption: https://chiapower.org/ Also helps that Chia can be farmed from a raspberry pi or similarly low-performance device, making it so there are over 100k farmers around the world currently.
I don't think it's true that Proof is Stake is less resilient personally, of the 30-100 proof of stake chains that have been running for over a year now, I don't believe there have been any consensus attacks, and many of those would be far easier to attack (as the cost of doing so is relative to the market cap of the native coin)
On the other hand, we did see consolidation of >50% of the hashrate on Bitcoin, due to mining pools. That was only short-lived because the mining pool which accumulated that much hashrate agreed to cap it at 48% or something
The opposite is true, proof of work is hopelessly insecure and centralized. It's hopelessly insecure because resources needed to mine are external - so any attack by the only realistic adversary (ie. a state) is always just a question of resources, and the most pessimistic cost is in low billions of dollars. Realistically - America could already 51% bitcoin by forcing regulations on existing miners for nearly zero cost.
It's also centralized because mining has infinite economies of scale. Bitcoin mining started with hobbyists mining on home pcs, now there are companies buying power plants just to mine. This only makes it easier for governments to control.
The only reason no big PoW coins were successfully attacked is because there's no clear motivation for anyone - nothing in the real world relies on any cryptocurrency to continue existing. The only realistic attack would be to try to exchange tokens on the attacked PoW network for tokens on another - which is easily done for low 9 figs, but becomes increasingly hard afterwards - making any attack uneconomical. On top of that there are legal issues - it's very possible a 51% double spending attack would be legally treated as theft.
This all changes completely the moment major countries and corporations go down if some blockchain goes down. Imagine Iran defaulting if bitcoin or ethereum gets taken over in a 51% PoW attack, with all transactions from Iran addresses censored. It would be orders of magnitudes cheaper than any serious military action. Fortunately for Iran and crypto holders that's not true. A significant adoption of a PoW network would end up in an inevitable disaster and potentially destroy any trust in cryptocurrencies forever.
Of all invented consensus methods only proof of stake can survive in a truly hostile environment. No government can print eth at will to take over ethereum, and stakers can easily stay anonymous, unlike massive corporate miners.
Bitcoin holders are very good at propaganda and pretending PoW is safe, which works as marketing, but wouldn't do anything against a real military incentive to take control over bitcoin.
>I'm much more bullish on Chia and proof-of-space/time
No, Chia has the same issue. Anything that relies on external resources as consensus votes is fundamentally insecure.
Already custodied stake is centralized. In Ethereum today Infura could selectively withhold and distribute transaction data in a manner suitable to attack - with 64x the chain space reliance on data providers which are already highly centralized increases.
In both scenarios the security comes from the motivation to not sully trust in the network you have invested so much to attack, so you are correct in pointing out how externalizing the work function provides a dangerous way in. The issue with PoS, of course, is that while the work function is insular, purchasing stake is not.
The only solution is to design a consensus that pays for the whole network (so Infura type entities are not necessary) and which always has a negative return on attack. The seed to achieving this is attaching work to every transaction, rather than every block. Saito goes further.
https://saito.tech/eliminating-51-attacks-in-proof-of-work-b...
>The issue with PoS, of course, is that while the work function is insular, purchasing stake is not.
That's not the issue, that's the 'S' in PoS. Consensus partially breaks if majority of state is hostile (no imminent double spend risk unlike PoW, but may require a manually coordinated fork with minority of validators). It breaks for 2/3 of hostile validators in the sense that new nodes would follow a different chain while honest minority is building on its own fork.
Both failure modes require some manual coordination, but rather than successfully censor or steal, they create temporary chaos - at the cost of millions of eth. It's impossible to even buy enough now, and any attack is self extinguishing because the available supply shrinks.
Under PoW the failure mode is trivial and deadly. If the attacker owns majority of hashrate he can censor anyone forever, including other miners. There's no real defense because there's no in-protocol way to delete the attacker's asics. Switching to a gpu pow achieves nothing - because the only entity prepared for that would be the attacker itself - but it would be pointless in the more general sense, as nobody would ever trust PoW anyway, because the same attack can always be repeated. The issue is use of external resources not controlled by the protocol.
Attack from the state is one thing, infinite economies of scale is another. An entity controlling majority can multiply its income by censoring everyone else. Eventually, everyone else mining at loss disappears. Then they can only mine at a fraction of hashrate and instead of paying for energy, use a fraction of excess profits to buy more asics. In the case of any competitor they would turn all of them on again until competitor gives up. This one is eventually inevitable, hard to say how soon.
These issues are completely unsolvable in PoW.
>so Infura type entities are not necessary
It's just a rpc - it's always necessary to answer queries and send transactions. At best it could work inside js in metamask as a light node. The only theoretical attack infura could do is lie about balances to someone, but the discrepancy would be detected and that would be the end of people trusting infura (to report correct state).
Infura is not 'just an rpc.' It takes real economic output to do what Infura does and its necessary for Ethereum to operate as well as it does today. If you force users to run their own lite nodes in their browsers they will simply leave, the alternative is that chain data interfaces centralize exactly as they have.
> Infura is completely irrelevant as far as security goes, they have zero power to do anything related to consensus.
Not directly, but the majority of data which ends up in block chains runs through Infura which accounts for transaction fees and affects the profits of miners. There are certainly levers to be pulled there. At the end of the day networking operations are important to security and therefore should be directly tied into consensus and the incentive system. That's why I bring up Saito, its work function is the quality of open networking provided.
As far as staking goes. Stake isn't protected from custodial pools, its already condensing in pools since people want to stake and can't afford 32 Eth. Not only does that defeat the purpose of having an arbitrary requirement, it incentivizes users to give up control of their keys.
Your failure condition for PoW is that someone eventually accumulates 51%, and you simply assume that it will happen. Convenient you get to ignore one of the best arguments for PoW, which is that sustainably mining, monopolizing mining, and centralizing the resources for profitable mining are all made extremely difficult by the distributed and remote nature of cheap energy. Scaling up with thin margins is not as simple as just buying another ASIC.
But just like in Ethereum's proposed PoS system, controlling a PoW network to the point of disaster does nothing but damage the attacker's own investment. The incentives are basically the same: extract as much value without going noticed. Its not great in either case.
Appreciate the enlightened discussion. It's in the spirit of crypto these systems are highly debated and it helps everyone get ahead and learn to participate in them.
Suggest reading what Vitalik has to say if you're going to offer this kind of objection:
https://github.com/ethereum/research/blob/master/papers/disc...
Infura collects upwards of 80 percent of the fees that flow into Ethereum and is in a position to control exactly who participates profitably. If your solution is "minority will fork" the obvious question is surely "with what scalable infrastructure?"
This is obviously false. Can you provide a source?
What does the discouragement attack paper have to do with this?
I really hope it takes off though, because if crypto is here to stay it really should stop wasting so much energy. We'll never solve the climate crisis this way.
I thought it was just an arbitrary decision saying "We'll merge on date X".
Why is that? PoS security model is fundamentally different from PoW - for instance - you cannot determine which of the two chains is "correct" anymore, you have to consult "a friend" [0]. That's not the case in Bitcoin, where the correct chain is simply the longest.
All PoS algos are very complex and have a ton of seemingly bandaid solutions for various problems that arise when you can create alternate histories freely (even if you use VDFs I think), and there is a disturbing lack of critique on that. So, why is it seen as "more secure"?
[0] https://yanmaani.github.io/proof-of-stake-is-a-scam-and-the-...
If the correct chain is actually the most difficult chain that follows the accepted rules, then you're still checking with a friend to find out what software implements the correct rules.
Most bitcoin people tell me the latter is the actual case, which means there's a level of social consensus for PoW as well as PoS.
Consensus on which rules to use happens on a different level. I'd say that it's economic consensus enforced by network participants. There's economic incentive to use rules which will not split the network.
That means you can't tell what the correct chain is just by counting hashes. You have to ask someone what the currently correct software is, and then you can find the correct chain for that software.
This puts you in the same position as someone on a proof-of-stake chain.
This is completely different issue than chain consensus disputes, which means that inside the chosen rules there's disputes on which history is correct. I.e. it's the double-spend problem, which PoW solves. PoS is subjective, and it's impossible to determine the correct history without asking someone. [0]
[0] https://hugonguyen.medium.com/proof-of-stake-the-wrong-engin...
In this sense there could only be Bitcoin. PoS cannot come close to that at all.
P.S. To clarify a bit more - I don't think the most-difficult chain can be with broken nonsensical rules, not in the long term.
Compared to Bitcoin as it exists in the real world, PoS has advantages.
You assume that the community somehow can agree which chain is the correct one in proof-of-stake. This is not true because there's no longest chain rule. Without an authority, users would never be able to converge into the same chain, after a split has happened.
There are ARM nodes that require less than 10W. Then there are more conventional servers that require ...150W? Average here is pure guesswork, but error bounds aren't that large - let's say 100W/node.
How many nodes are going to exist - I think 10k is a reasonable maximum number.
The result is 24000kWh per day. At $0.15/kWh, that means $3600 in daily energy costs.
There's also the hardware cost itself and internet connection cost (which can be zero if there's an already paid for connection) - but all in all, total, real expenses to run ethereum become a rounding error of current ones. Guesstimating hw depreciation and connection cost, daily total at $5k translates into $1.825M/year. The value of block rewards during the last 24h is ~$15M.
In other words resources required for running ethereum for ~3 hours now are going to be enough to run it for a year after the merge.
With higher throughput, more strategies that once again saturate the block space, that could be more lucrative than it seems?
Don't forget about MEV (maximum extractable value). Block producers can front run your trades and do other interesting stuff to about 6% extra a year using something like flashbots.
That math is going to change as the Ethereum miners reconfigure for altcoins and the difficulties change, but the bottom line is that mining will continue as long as some coin is still profitable to mine.
We'll see if it plays out, but cryptocurrency prices falling have certainly reduced the energy consumption as well (it becomes less profitable to mine at a given difficulty when your mining reward is worth less).
If Ethereum does eventually manage to overtake Bitcoin, that'll bode well for crypto's energy footprint in the long term (fingers crossed for this happening, or at least for there to not be another Bitcoin/PoW bubble)
Given this, a LOT of mining capacity is going to go offline when this occurs.
Proof-of-work actually does work - it is no longer a debate, and these people (many of the same people who were saying “b-b-but tulips” a couple years ago) can’t stand it. I wouldn’t trust them to engineer their way out of a paper bag - definitely not after some muppet on the TV reminds them to get back in line.
Capitalism is still winning again, folks. Deal with it.
Climate change is painting different picture.
Bit of a Catch-22 for you there, sorry. It’s just reality.
The logic of whom not to trust in this chain is pretty simple. There is no catch-22 here without some magical stretch of logic.
I mean my reason for not trusting said capitalists is that I don’t want people’s success to be vested in the destruction of our planet. I honestly don’t understand yours.
What a great argument against PoS.
Even if this merge is done, a greener pyramid scheme doesn’t change anything about the speculators and scams that Ethereum harbours as Ethereum becomes even more centralised and plutocratic.
Gas fee surges will always still be a problem after the merge which only favours the rich that can afford them making Ethereum useless at scale for on chain usecases.