Recently, in an apparent response to a largely-flawed critique of stablecoins from the Open Markets Institute, cryptocurrency exchange FTX clarified its position on transaction fees for withdrawals.
Its blog post was striking in that it appeared to associate proof-of-work (PoW) blockchains with high fees (which users are partly responsible for upon withdrawal) and proof-of-stake (PoS) blockchains with low fees. The conclusion: FTX wants to encourage users to use low-fee, less-energy-intensive, proof-of-stake blockchains. We can see the appeal of associating PoW with extractive, consumer-unfriendly, high fees, and PoS with efficiency and user-friendliness. But FTX is mistaken to associate consensus and fees.
In its article, FTX claimed:
“The actual amount that a blockchain requires to send a transaction differs widely based on the underlying structure of that blockchain. Platforms like Bitcoin and Ethereum are known as ‘Proof of Work’ blockchains, where the ‘work’ required to add that transaction to the blockchain uses a large amount of computing time and energy. On such platforms, average transaction fees are quite high: around $2 per transaction for Bitcoin, and around $40 per transaction on Ethereum!”
Leaving aside our surprise at seeing a major exchange take such a partisan approach, the analysis relies on a misconception regarding the relationship between consensus (or Sybil resistance) methods and blockchain fees. There simply is no inherent association between proof of work and high fees, or proof of stake and low fees. The fact that the only meaningful fees exist on two blockchains (Ethereum and Bitcoin), both of which currently happen to be PoW-based, does not mean that PoW implies or causes fees. It simply means that the two most popular blockchains both use PoW and are somewhat congested, leading to high fees (Ethereum, more so than Bitcoin).
Proof-Of-Work Basics
In PoW coins, “work” must be performed and verified before a block is appended to a blockchain. Producing work requires miners to perform several attempts before finding the number that grants them permission by the protocol to add a block to the blockchain. At first glance, it may appear that proof of work’s trial-and-error architecture naturally entails a delay in block production and that, in times of congestion, that delay pushes fees higher. However, this is a misunderstanding of what drives throughput.
The time in between blocks is not what determines throughput in crypto networks. Instead, the main determinant of throughput is block size, i.e., the number of bytes (and hence, transactions) that can fit into a block. Consider that a blockchain designed to produce one block per second with 1,000 transactions in each block has the very same throughput of a blockchain that produces one block per minute that is large enough to fit 60,000 transactions.
Critics of proof of work might be tempted to claim that an increase in the interval between blocks affects settlement time, which in turn increases congestion. That would also be misguided. A transaction included in a block is not final. All blockchains, including those that follow new architectures such as Solana, require users to wait before considering a transaction final. The reason behind this wait is that there are events that might take place within that period where the blocks in the blockchain are reorganized. Depending on the severity of these events, a transaction that was once in a block might be permanently removed from the blockchain.
Fees Are A Function Of Supply And Demand
The cause of fees is simply more demand for blockspace than there is available supply. Under conditions of scarcity, a prioritization method for transactions must be determined. One way is to create an auction in which eager transactors can pay up for priority inclusion in a block.
Having material fees is extremely healthy for a public blockchain system: it eliminates the spam problem by making it costly to insert junk data, and it constitutes “protocol revenue” that can be directed to a number of causes.
In Bitcoin’s case, this fee-based revenue will pay for security once issuance trails off. For Ethereum, fees are already being burnt to introduce a deflationary mechanic. You could also redirect fees to finance various public goods like paying Core developers. To make a rough corporate analogy, fees are “revenue” and issued supply is “equity.” Many firms do finance their operations by continually issuing stock, but shareholders generally prefer not to get endlessly diluted. The existence of fee revenue frees blockchains from dependence on dilution-based financing.
In such blockchains, fees also play a critical role in supporting their long-term security. They make it costly for information to be stored on the blockchain, thereby disincentivizing spam and DDoS attacks that have historically plagued zero-/low-fee networks, like Nano, EOS and XRP. Most crucially, fees promote a competitive environment among miners which in turn makes it prohibitively expensive for single parties to successfully attack a network. Thus far, proof of work in high-fee environments is the only battle-tested mechanism known to the industry to be resilient against attacks.
In its post, FTX claimed that “the ‘work’ required to add [a] transaction to the blockchain uses a large amount of computing time and energy.” This is erroneous. Contrary to this common characterization of PoW, there is no “energy payload” required to make a transaction. You are not using joules to push transactions through the pipes. Making, registering and validating a transaction costs very little, computationally.
The thing which is expensive (financially, and, in the case of PoW, in terms of energy, too) is winning the eligibility rights to include a block, obtainable by brute-forcing for a valid nonce. And it’s expensive because the reward for creating a block is significant — around $290,000 at the time of this writing. Logically, miners will pay up to $99 to win a bounty worth $100. But this bounty exists due to the issuance of new coins as fees are de minimis (in Bitcoin at least). The bounty is also available whether a block contains 4,000 transactions or none.
The per-transaction energy cost figure that FTX and the affiliated Solana make frequent reference to is not a useful analysis. Bitcoin could produce far more blockspace, thus driving fees to zero (as BSV did indeed do, for instance), without expending a joule more energy. Bitcoin could also process zero transactions per block, and miners would expend virtually the same amount of energy. There simply isn’t a linear correlation between transactions and energy expenditure, and there is barely any causal linkage between the two.
Why Limit Block Space?
As to why fees exist in the first place, they are the consequence of crowded block space. Congestion exists in a blockchain context because the basic security model of blockchains requires that end users can independently audit and verify the transactional history from the very first block should they choose to, and there’s a limit to the quantity of data that can be audited per unit time.
A blockchain is a replicated ledger. The orthodox security model requires that users be able to actually run a current version of that ledger, and recreate and validate all historical transactions, thereby ensuring that the rules are being followed. Bitcoin’s design philosophy aims to permit anyone with at least a weak internet connection and consumer-grade hardware to perform a full audit of the transaction log.
Ethereum takes a more liberal approach, adding computational complexity and some scalability at the cost of more challenging and expensive verification. But still, running an Ethereum node should be doable on high-end consumer hardware if users discard some historical information after validating it, a technique called “pruning.” It is not out of the reach of a somewhat technical individual with a modest budget.
The design philosophy of both Bitcoin and Ethereum (at least in its current form — founder Vitalik Buterin has more ambitious plans which deviate from this idea) stresses the importance of an individual being able to run a current copy of the ledger. Therefore, the growth of the ledger must itself be constrained to keep the cost of node operation within reasonable bounds. The major constraints are disk i/o, bandwidth and storage capacity.
It’s not enough to store the blockchain — you have to stay up to date with its latest entries, which means downloading a lot of data and performing new computation by verifying data as it arrives. Here is where we arrive at the key constraints: There’s only so much computation modern hardware can perform per unit time – only so many signatures that can be verified and state changes verified. Of course, node software can (and has been) optimized, to eke more computation (and hence transactional validation) out of the same number of bit flips. Storage and bandwidth are generally becoming cheaper with time, too. But these still represent genuine constraints grounded in the laws of physics. A computer can only do so much.
So, we arrive at the status quo. Bitcoin’s protocol makes available a theoretical maximum of 4 MB of new block space every 10 minutes — in practice, this hovers around 1.2 MB at the current weekly average. Ethereum post-EIP-1559 creates roughly 6 MB every ten minutes. If demand exceeds supply, a queue emerges, and the highest bidders get priority access to block space. Hence, fees.
As demonstrated, fees are not a “PoW thing” or an “energy thing.” They are a “security model” thing. If you want to keep the decentralization high, you want to keep the cost of node operation low, and thus you want to limit the quantity of data a validator must process per unit time. If you do all of these things, and your blockchain is popular, fees will organically emerge, as they did in Bitcoin and Ethereum.
Now, if you take a much looser view of security, and you are content to have a small number of very performant nodes doing all of the validation, then you can create more block space, and drive fees effectively to zero.
This is not a new idea; it’s the foundation of the “big block” movement in Bitcoin, which embroiled the protocol in a civil war for the better part of a decade. That movement gave birth to the perfect counterexample to the claims of FTX: BSV.
The designers of BSV created virtually-unlimited quantities of blockspace, content as they were to have a small number of industrial nodes perform validation. Fees are effectively zero in BSV. But this is a PoW network, and its miners absolutely consume energy. Conversely, at some point next year, Ethereum will move to a proof-of-stake model, at which point it will stop consuming meaningful amounts of energy. But I expect Ethereum will still having meaningful fees at the base layer — and these fees will be considered desirable in many respects, since they support the deflationary mechanism introduced with EIP-1559.
The reason that Solana, for instance, has low fees, is simply because the designers of that network were happy to adopt a different security model from Bitcoin or Ethereum. In Solana, there is virtually no difference between running a node for the purposes of verifying the integrity of the chain and running a node for mining blocks. As such, running a Solana node requires extremely specialized hardware and an experienced devops team.
We can attest to this, as Coin Metrics runs one (alongside 100 other nodes spanning 25 distinct Layer 1 blockchains). It costs Coin Metrics dozens of thousands of dollars a month to run a SOL node. That is a magnitude higher than the couple of hundreds of dollars a month we spend running BTC nodes.
At current rates, Solana produces approximately 550-times more blockspace than Bitcoin per day. Solana validators, at current rates, must process around 100 GB per day of data, or 36 TB per year. Most of that data is removed, or pruned, which impacts the ability of third parties to check all transactions from genesis.
Bitcoin node operators, by contrast, ingest around 180 MB per day, or 65 GB per year. Solana validators must therefore manage two orders of magnitude more data than Bitcoin validators. Ethereum is a bit more complex and computationally intense than Bitcoin, but still far more limited than Solana in terms of the computational work validators must do to maintain the ledger.
Solana can offer users more abundant blockspace and therefore a cheaper all-in transactional experience, but this comes at a cost. The network has recently experienced outages, as its relatively few nodes were successfully targeted with DDoS attacks. Effectively, Solana obtained (a measure of) scalability, but at the cost of more centralization, and consequent fragility.
Ultimately, the Sybil-resistance mechanism used is largely irrelevant to the question of fees. A PoS network could be completely costless from an energy perspective and constrict block space, causing fees to emerge; a PoW network could increase blockspace and drive fees to zero.
Heavy Is The Head That Wears The Crown
While FTX’s analysis is off base on the question of fees and PoW, we can nevertheless sympathize with the desire of an exchange operator to align itself with proof-of-stake networks, and to minimize the importance of PoW networks.
After all, if you can influence the world toward an outcome in which PoS-based monetary goods are dominant, and you run a large custodial exchange which stands to accumulate lots of those PoS assets, your incentives are clear. Other things being equal, you probably prefer to have more rather than less influence over the world’s future monetary protocol.
In a PoS-dominant world, exchange operators, custodians and banks that accumulate the most coins are king. Users that deposit coins generally surrender their coin-based network voting rights to the exchanges themselves. There are already examples of exchanges being used to influence PoS networks, as occurred when Justin Sun colluded with Binance, Huobi and Poloniex to commandeer the Steem network. These exchanges voted with user funds in Sun’s favor, demonstrating an obvious principal-agent problem created by the custody of PoS assets.
In a PoW world, large intermediaries are much less empowered. The failure of SegWit2x, a movement supported by most of the large exchanges and custodians at the time, demonstrates this. Imagine a similar movement today, except taking place on one of the larger PoS networks. The largest exchange operators, custodying as they do a large plurality of all the outstanding coins, would simply shape the protocol to their liking with no resistance.
And in a world where operating an exchange is a decidedly hazardous profession, as demonstrated by the travails of BitMEX, Huobi and OKEx executives, the inclination is surely to offend the powers that be as little as possible.
So, it stands to reason that FTX leadership would align itself with “ecological” PoS, eliminating what has historically been the most strident objection to public blockchains from the policy crowd. Why rock a boat which is already swaying quite precariously?
But we would argue that even though the naive analysis suggests that exchanges should, as a group, support and foster the growth of PoS while marginalizing PoW, this is unwise in the long run. If these exchanges/brokerages/banks accumulate a large fraction of all the coins, they will amass enormous political power, especially if these blockchains become monetary assets of global consequence. At that point, accumulating voting power proportional to coins held becomes a poisoned chalice. The exchange becomes a gigantic honeypot for the state — a state which will not surrender its power of sanctions easily.
As we transition from a world where the U.S. projects power through correspondent banks and international systems like SWIFT, to a world of stablecoins, MetaMasks and Layer 2 protocols, the state will have to develop new ways to control financial flows. It would be convenient in the extreme if a small handful of exchanges accumulated a large portion of supply in PoS networks, and then submitted (as they ultimately must and will) to increasingly onerous regulation.
At this point, exchanges would simply become deputized — just as banks are today — into carrying out state policy, which could well extend to controlling public blockchains at the protocol layer. PoS networks explicitly grant control and discretion to the largest stakeholders, so at this point, the jig would be up. The state would be free to pursue its merry ambitions of deep financial deplatforming.
This isn’t just fantasy. Already, the U.S. financial policy establishment is demanding that stablecoins obtain federal bank charters, which would bring issuers directly under the aegis of the Federal Deposit Insurance Corporation (FDIC), the Office of the Comptroller of the Currency (OCC) and the Federal Reserve. The exchanges, currently loosely regulated in the U.S. under a patchwork of state-by-state regulations, will likewise be asked to submit to federal regulation.
So, the exchange CEOs that lionize purportedly ecological PoS and dismiss the merits of PoW should be careful what they wish for. It may seem appealing on a surface level to control consensus from the seat of a large custodial exchange, but it is a power that is best spurned in the first place.
Public blockchains exist to eliminate centralized points of control and to remove the political constraints that are inherent in traditional finance. The combination of PoS and large quantities of coins held in regulated exchanges or banks is one that is very conducive to the state reasserting control over these nominally-decentralized systems. Unless you are eager to be deputized into a hall monitor for the new financial system, it is best to repudiate the influence that helming a PoS network would grant you.
This is a guest post by Nic Carter and Lucas Nuzzi. Opinions expressed are entirely their own and do not necessarily reflect those of BTC Inc or Bitcoin Magazine.