Update 2021-04-26: Thank you everyone who provided feedback (here and here, mostly). I have updated the post with a new, clearer example of how an ETH presale insider could have manipulated the presale to their advantage. To make it more clear what new kinds of risk are introduced by the possibility of manipulation by presale insiders, I added a hypothetical example of how manipulation of the ETH presale could potentially create problems for Ethereum users down the line. I also added more context about why bitcoin extensibility matters in the Ultra Sound Money debate, namely the role of extensibility in supporting bitcoin’s long-term security budget. And I added a reply to a section from David’s post comparing the sustainability of bitcoin security post-coinbase subsidy to Eth2’s dynamic issuance policy.

Update 2021-05-03: Added a subsection about credible neutrality to the “General comments” section.

Last weekend I read David Hoffman’s recent post comparing ETH to BTC, “Ultra Sound Money“.

I read this post while doing research for a longer report I’m working on about the investment case for BTC (publish date TBD). Coincidentally, the other day I also saw a tweet from Ryan Sean Adams suggesting that no Bitcoiners are willing to debate the ETH = Ultra Sound Money thesis. So I figured I would publish my response to David’s post to get the debate going.

I’ll start off by saying I’m a fan of the Bankless community’s work. I have followed Ryan Sean Adams (one of the Bankless founders) on Twitter since 2018 and though much of our interaction is just me pushing back on his overly-bearish claims about BTC or overly-bullish claims about ETH I still enjoy much of the content he and his Bankless community put out. Bankless very clearly has a heavy Ethereum bias but I think Bitcoiners and the Bankless community have a lot to learn and gain from each other’s perspectives. So I’m happy to have a chance to dialogue with the Bankless community by responding to the Ultra Sound Money post.

My summary of David’s post

Just so we know we’re on the same page:

• ETH is used in Eth2 to protect the blockchain against 51% attacks via bonded PoS.
• The more ETH-denominated transaction fees are paid in Eth2, the more ETH will be locked in staking to earn transaction fees, leading to a reduced circulating supply (i.e. supply shock).
• BTC miners have to sell most of the BTC they earn due to opex costs; Eth2 stakers will not have to sell as much of the ETH they earn because they have significantly lower opex costs (all else being equal) leading to less overall sell pressure on ETH relative to BTC (all else being equal).
• Annual ETH inflation will drop by 75% under PoS, from today’s ~4.5% to ~0.5-1%.
• The ETH inflation rate will adjust based on the amount of ETH that is participating in Eth2 staking; lower if there is less ETH staking, higher if there is more ETH staking.
• The more economic activity there is on Ethereum, the more ETH will be burned via EIP1559, potentially resulting in a net reduction (i.e. deflation) in the ETH supply.
• The above factors make ETH even more sound (“ultrasound”) than BTC, because a) ETH is more likely to have a sustainable monetary policy, and therefore b) the Ethereum blockchain is more likely to be secure in the long run.

General comments in response to David’s post

Even if all of the points made about ETH and the beneficial properties given to it by PoS and EIP1559 are true (and I believe the basic technical facts are true, see specific replies to the contrary below) there are several other important factors that must be considered when comparing ETH monetary properties to BTC that were not mentioned in David’s post.

The ETH premine and presale dynamics

There’s no way to know how decentralized the distribution of the ~70 million premined/presold ETH (>60% of the current supply) really was. What we do know for sure is that one entity received all of the money from the presale, which opened up opportunities for manipulation of the initial supply (e.g. the seller could borrow BTC and put it into the sale to artificially inflate the initial supply of ETH, then repay the borrowed BTC after the presale is over; some related dynamics are discussed here). Ethereum users are trusting that this manipulation did not happen, or that if it did happen, that the effects won’t be that bad.

Decentralization of supply distribution is vitally important for any asset aspiring to be sound money (let alone “ultra sound money”) to avoid the enshrining of a privileged class who can use their privilege to disproportionately/unfairly influence governance going forward. This is a kind of risk that won’t be realized until the stressors are in place that cause this problem to rear its ugly head, and even then it is unlikely to be obvious what is really going on.

As a hypothetical example of how the ETH presale dynamics could be abused to manipulate Ethereum governance, imagine that a presale insider (“the insider”) did use their privileged position to unfairly grant themselves an outsized portion of the initial ETH supply. Then one day a controversial EIP is proposed that the insider doesn’t like. This EIP ends up roughly splitting the community in half, and the EIP supporters feel so strongly about it they go ahead and implement the proposal, causing a chain split. Now there are two versions of Ethereum: A, which the insider likes, and B, which the insider does not like.

In this hypothetical example the insider now has ETH on both chains. What they can do is they can use their outsized share of the ETH supply to sell ETH-B in exchange for ETH-A. This suppresses the price of ETH-B and supports the price of ETH-A. They can keep doing this until they run out of ETH-B. By then, however, Ethereum-B might be dead, or close to it, due to the unfair price suppression. This is just one example of how the manipulation enabled by the Ethereum presale could cause governance problems down the line.

BTC on the other hand has been mined on equal terms by all participants since the very beginning. No individual or organization has ever been singled out and given a built-in privileged position in the bitcoin protocol. The 50 BTC coinbase reward minted by Satoshi Nakamoto in the genesis block can’t even be spent! Bitcoin is as decentralized and trust-minimized as electronic money can get, which greatly reduces or eliminates entire classes of risk (key person risks, legal risks, governance capture risks).

(Aside, not a reply to David’s post but responding to something related to this section that came up in a comment thread about this post: some people seem to think that early BTC miners have the same privilege as ETH presale insiders. This is false equivocation.)

Ethereum is not credibly neutral

Neutrality is another vitally important feature for any asset that aspires to be sound money. If users of a money cannot expect to be treated equally and fairly, then why should they consider that money to be a sound store of value? If an alternative monetary asset cannot credibly claim to be neutral, users might as well stick with the money they already have (for most today, that would be fiat currency).

There have been several instances throughout Ethereum’s history where participants have broken the principle of neutrality, making changes (or not) in favor of some users but not others.

One example is The DAO hard fork in 2016. This hard fork broke The DAO contract, allowing some users to withdraw funds that they were not previously entitled to withdraw, and allowing other users to not withdraw funds that they were previously entitled to withdraw.

In contrast to The DAO fork, in 2017 all Parity multisig wallets were frozen due to a misconfigured contract deployment. Although the proposed fix was simple and would enable all Parity users to move their funds to a more secure wallet, Ethereum developers never implemented the fix.

The fact that The DAO mainly affected Ethereum insiders (early adopters, prominent developers, etc) and the Parity issue mainly affected developers of a competing platform (Polkadot) points at potential bias in terms of what contract “fixes” are implemented and which are not.

Another example where credible neutrality was broken is with the Istanbul hard fork, which intentionally broke thousands of contracts including my own Aragon multisig contract. The specific effect of the Istanbul hard fork on my Aragon contract was that it would no longer be safe to use ETH with any other smart contracts (for example, I couldn’t use the Aragon contract to deposit ETH into a Maker vault). This violated the principle of forward compatibility and could have led to significant losses for any users who weren’t closely following this consensus rule change.

Bitcoin, on the other hand, has arguably only once ever violated the principle of neutrality. The incident occurred in 2010, before BTC had any significant monetary value, when Satoshi pushed a soft fork that fixed a value overflow bug. This soft fork deleted billions of bitcoins that had been minted in a single transaction. Strict neutrality would have let the user keep their billions of bitcoins, however this would have violated a previously established expectation of a 21 million BTC supply limit with a fixed emission schedule. The previously established limit and emission schedule won out.

The fundamental differences between PoS and PoW, and the implications for protocol governance and security

It’s one thing to use PoS as a layer on top of PoW, either directly on the mainchain (as in hybrid Casper FFG) or as a mechanism for securing a sidechain (as in Nomic). It’s something else entirely to view PoS as a replacement for PoW and to fully rely on PoS to secure the base layer mainchain. This is relevant to a discussion about the soundness of ETH as money because if, after thorough analysis, it turns out that PoS introduces new tail risks that PoW does not have then that might negate any perceived benefits introduced by PoS and challenge any notion of monetary soundness (all other issues with ETH/Ethereum notwithstanding). Here’s my take on this and for balance, here’s Ethereum founder Vitalik Buterin’s take.

ETH may be either inflationary or deflationary, while BTC is strictly deflationary

It should be noted that while under Eth2+EIP1559 ETH may be either inflationary or deflationary depending on the staking rate & BASEFEE burns, BTC is, in the long run, a strictly deflationary asset. Due to BTC’s protocol-enforced supply cap, the maximum circulating supply of BTC can only ever go down. In addition to permanently lost coins (estimated to be in the several millions, but impossible to know with 100% certainty) BTC has been provably burned by application-layer protocols such as Counterparty and is also burned by the Stacks protocol. Proposed (but not yet implemented) protocols such as Spacechains and StakePeg could also result in BTC being burned at rates proportional to the success of the protocol, resulting in even more deflation.

Bitcoin has a clear path to survival under a fee-only regime

One main concern that is brought up in David’s post is the instability of bitcoin’s security budget under a fee-only regime. While people have been speculating about whether or not bitcoin will survive in a fee-only regime (contra today’s coinbase + fee regime) since bitcoin was first introduced, the pessimists were given intellectual ammo with a 2016 paper published by Carlston et al, On the instability of Bitcoin without the block reward. This paper describes several theoretical issues that could arise in a fee-only regime, namely a new “undercutting” attack and several downstream effects caused by this attack. It should be noted that the results of this paper were later contested in a 2020 paper published by Gong et al, Towards Overcoming the Undercutting Problem.

Even disregarding Gong et al’s rebuttal of the earlier Carlston et al paper, however, there are reasons to be optimistic that bitcoin fees will eventually make up for the lost coinbase subsidy. One reason is that, long term, fee revenues on bitcoin have been trending upward. Provided that bitcoin adoption continues to grow, fee revenue is likely to continue increasing as well.

Source: https://charts.coinmetrics.io/network-data/

Another reason for optimism about bitcoin’s long-term fee revenue is that all of the smart contract capabilities of Ethereum (and any other altcoin) can be replicated using embedded consensus, Layer 2, or sidechain protocols built on bitcoin. See RSK for the most obvious example of this (detailed Ethereum <> RSK comparison here) as well as what projects like Nomic and Stacks are working on. These additional capabilities unlock new use cases for bitcoin, providing additional sources of fees for bitcoin miners and contributing to bitcoin’s security budget.

(Note, however, that the security model of BTC on sidechains is not exactly the same as mainchain BTC. You can read about the different security models of the two-way peg mechanisms that enable BTC sidechains here.)

Replying to specific sections from David’s post

‘Buying’ money charges it with power. ‘Selling’ money discharges its power.

I think this is a flawed metaphor. As a simple example to show why, consider someone who purchases 1 DOGE for $100 USD, while the rest of the market prices DOGE around $0.20 USD. Is the 1 DOGE that was purchased “charged” with $100 USD of purchasing power, to be later discharged on demand? No. The buyer will be unable to recapture that $100 USD of value with that 1 DOGE until they can find a buyer willing to buy at least 1 DOGE at that price.

Considering how much of the post relies on this flawed metaphor to make its points, one has to wonder how different the conclusions in the post might be if the author used a more accurate metaphor instead.

(I could also be misinterpreting the metaphor; if I am, let me know.)

Crypto-economic systems protect the economic engines, but they do not tinker or manipulate them.

Crypto-economic systems are not led by a collection of human brains; they are powered by numbers and algorithms. Crypto-economic systems do not make subjective investments or bets on the future. They do one thing, and one thing only: provide economic security.

This constraint on the scope of crypto-economic systems is theorized to produce better social-scalability. No longer do the economic actors on these platforms need to place their trust in the competence of human leadership; instead they have assurances about the long-term operation of these systems, as these systems have automated away the need for human coordination, and replaced centralized leadership with code and math.

I can think of two counterexamples:

1. “Crypto-economic systems” like Decred and Tezos that have human-directed treasuries funded directly via seignorage.
2. Ethereum, which has had multiple instances of human intervention in the monetary policy, thereby “tinkering” with the economic engine by manipulating the inflation rate of the money used by actors in the economy. Also, the Ethereum community intervened with a hard fork after The DAO incident, and the Istanbul hard fork caused thousands of smart contracts to become unusable, both actions that I would consider tinkering with or manipulating the economic engine.

That is to say, either the definition of “crypto-economic system” given here is inaccurate, or the systems I mentioned as counterexamples are not crypto-economic systems. (In which case, what are they?)

Gold is secured by the energy of a supernova. After the energy in an exploding star is released, all gold has been minted from this event and no more. All economic costs to maintain the security of gold is paid for at genesis. There are no ongoing costs to secure the value of gold; it gets its security for free.

I may be misinterpreting this claim (I also watched the Justin Drake interview and didn’t get much more context) but gold has many ongoing security costs once mined, including assaying, branding and certifying, vaulting, etc.

Bitcoin Proof-of-Work operates through an endless competition to mine Bitcoin blocks faster than everyone else. Those that produce more hashes mine more blocks, and receive more rewards.

This PoW competition perpetually gets more and more fierce as the value of BTC goes up.

I wouldn’t say this competition “perpetually gets more and more fierce as the value of BTC goes up”. Eventually the competition will be entirely dependent on the value of transaction fees, which will be a product of the value (or price) of BTC, the fee rate, and volume of transactions. It’s possible for miners to earn less value per block even as the value of BTC is increasing, if for example the volume of transactions and fee rates paid are decreasing for some reason.

As the margins for PoW miners get slimmer and slimmer, more and more BTC must be sold to pay for the operational costs. Selling the unit discharges the energy, and Bitcoin security is fundamentally designed to instigate a race to consume as much energy as possible.

(I have already pointed out earlier why “selling discharges energy” is a flawed metaphor.)

It is true that in the giant “order book” of global trade, net selling of BTC for USD will marginally reduce the USD price of BTC. But what if the BTC that miners are “selling” is being traded directly for electricity, and the power company hodls? Or what if the power company pays its employees BTC, and they hodl? Does this still result in selling pressure on the BTC price? I don’t think it necessarily does.

Imagine an entire circular economy that runs on BTC. Does the BTC circulating through this economy put sell pressure on BTC and reduce its overall purchasing power? Again, I don’t think it necessarily does. Only if there are net sellers. But in a circular economy, trades are generally 1:1 in value so buying and selling balance out.

Bitcoin security is fundamentally designed to instigate a race to consume as much energy as possible.

I would rephrase this to say “Bitcoin security is fundamentally designed to instigate a race to consume as much energy as the block reward will purchase.”

Energy will only be consumed by bitcoin miners up to the point that it is no longer economically rational to consume any more energy. There is an upper limit on the amount of energy that will be consumed by miners, and that is the amount of energy that the value of the block reward (coinbase + fees) will purchase for them.

Bitcoin has no capacity to give long-term assurances to its security providers about the future size of the security budget… Not being able to meter the economic resources inputted into the Bitcoin security engine means Bitcoin miners have an unpredictable security budget… Proof of Stake rewards increase or decrease as a function of the supply of ETH being staked to Ethereum. If there is less ETH staked, ETH rewards are higher and vice versa… Through this mechanism, Ethereum organically discovers the optimal balance between Ethereum security and ETH issuance. 

David says “Bitcoin has no capacity to give long-term assurances to its security providers about the future size of the security budget”. However, even using the adjustable rewards mechanism described in the above quote, Eth2 isn’t able to give long-term assurances to its security providers about the security budget either. Sure, there’s an algorithm that says “If x ETH is staking, y ETH will be minted as staking rewards” however that doesn’t tell stakers anything about the value of those rewards, which are constantly in flux due to market dynamics.

So the comparison here isn’t “Bitcoin, which cannot give security providers long-term assurances about the security budget vs Ethereum, which can give security providers long-term assurances about the security budget”. The comparison is, “Bitcoin, which after 2130 will exclusively fund security using transaction fees that vary in value vs Ethereum, which after the transition to PoS will fund security using a combination of fees and staking rewards that vary in value”.

If the Bitcoin economy is running hot, Bitcoin forwards all of this revenue directly to the miners, instead of taking those economic excesses and saving them for later.

Why classify higher fees as “economic excesses”? Maybe a higher fee rate is what is “normal”, and the lower fees were subpar. Who’s to judge what is “excess” and what is not?

Not having the option to pocket these fees and save them for later is the forfeiting of an extremely powerful value-capture mechanism, and leads to less soundness in BTC over time.

Citation/explanation needed, in particular for “leads to less soundness in BTC over time”. I personally would not mind having an EIP-1559-like mechanism on bitcoin, but I don’t think that not having it makes BTC any less “sound”.

Bitcoin is designed to preserve the 21 Million hard cap supply of BTC, and everything else is a means to produce this end.

Au contraire, bitcoin’s raison d’être is irreversible transactions, and having a sound monetary policy is a means to that end.

Stakers are not required to sell their ETH rewards to pay for the economic costs of providing security.

Assuming ETH block production is the only source of income for stakers (as seems to be assumed for BTC miners earlier in the piece) I imagine stakers will have to trade at least some of the ETH they earn from block rewards for the hardware, internet, and electricity costs required for staking, since these resources are not free.

As more and more validators stake ETH to Ethereum, the ETH-denominated rewards are reduced. This process will wash out the lesser-ETH bulls who are less interested in reduced rewards, leaving those who are willing to receive the least amount of ETH possible.

This seems to me to be conjecture. Both ETH-bull-stakers and non-ETH-bull-stakers (let’s call the latter “entrepreneurs” and the former “ideologues”) want to maximize returns as measured by purchasing power; ideologues, so they can accumulate more ETH, and entrepreneurs, so they can reap more profits in their currency of choice.

If ideologues see an investment opportunity that will provide them with greater returns than staking ETH, they will redirect their capital to that investment so that they can earn more purchasing power and purchase more ETH. There’s no reason to expect that they’ll continue staking all of their ETH if an enticing enough opportunity (relative to the returns provided by ETH staking) presents itself.

Also, given the attention paid in this piece to the supposedly endemic sell pressure on BTC as a result of high mining opex, it’s worth noting that “washing out lesser-ETH bulls” would also result in sell pressure on ETH that will have to be absorbed somehow, either via new capital flowing into ETH or via depreciation in the value of ETH.

If an interested party wants to attack Ethereum, they will need to source 3,618,200 ETH in order to do so. $6.7B of buying pressure on ETH the asset is likely to significantly increase the price of ETH, so the US-Dollar value of Ethereum security is actually likely much higher. 

This assumes that the ETH required for an attack must be bought, but the ETH could be borrowed (which could also be a nice setup for a leveraged short trade, introducing the possibility to profit further from an attack if it causes the price of ETH to drop).

EIP1559 is a mechanism that associates economic output to economic unit scarcity. ETH gets more scarce as Ethereum’s economy produces economic output.

Assuming more ETH is burned than is issued, yes.

Sound money is money that grows in scarcity as a function of the growth in the size of the economy that uses it.

This is notably different than any definition I’ve ever known for the term “sound money”. Words change in meaning over time, so I don’t expect perfect consistency here, but I do think it’s worth noting that this is an abnormal definition of the term. Here are a few definitions I found for “sound money”, from a diverse variety of sources:

American Enterprise Institute:

“[Sound money is] money that has a purchasing power determined by markets, independent of governments and political parties.” (source)

Ludwig von Mises:

“The principle of soundness meant that the standard coins — i.e., those to which unlimited legal tender power was assigned by the laws — should be properly assayed and stamped bars of bullion coined in such a way as to make the detection of clipping, abrasion, and counterfeiting easy.” (source: Human Action, p. 776)

Sound Money Defence League:

“Sound money is money that is not prone to sudden appreciation or depreciation in purchasing power over the long term, aided by self-correcting mechanisms inherent in a free-market system.” (source)

Thomas J. Jordan, Chairman of the Governing Board, Swiss National Bank:

“Money is a generally accepted medium of exchange, and performs two other important functions, namely as a unit of account and as a store of value… money is sound if it fulfils its functions as well as possible.” (source)

There is no matching ETH’s level of trustlessness, as all other assets on Ethereum were issued by humans (who are folly) or code (which can be buggy).

ETH is issued both by code (e.g. Geth code) and by humans (e.g. EIPs that modify the ETH issuance rate). Perhaps rephrase?