TL;DR
Introduction
In our last few posts, we went deep into the economics of the Ethereum network, breaking down the layer 1 fee structure, various layer 2 business models, and highlighted the impacts of recent network upgrades on the overall economic dynamics.
One of the most impactful network upgrades was the rollout of Proto-Danksharding through EIP-4844. At the highest level, this introduced a new transaction type specifically designed for Ethereum’s layer 2 networks, called ‘blobs’. This effectively created a new fee market, shifting the majority of rollup data away from costly call data towards lower-cost, ephemeral blobs. As we highlighted, this reduced L2 fees paid to L1 by 95-99% overnight and substantially lowered gas prices, and thus fees and revenues on Ethereum mainnet.
Source: Aggregate fees paid to ETH L1 by L2 networks via growthepie.xyz
Following the Merge and prior to this upgrade, Ethereum had been largely deflationary, with the burn from transaction fees more than offsetting new network issuance. Following EIP-4844, this has no longer been the case.
Source: ultrasound.money – the impact of EIP-4844 on Ethereum’s burn is clear
Is the fee burn enough?
Following EIP-4844 and the general migration of most execution to L2s off of mainnet, a major question arises in Ethereum’s tokenomics: can the network return to net-negative issuance through its fee burn mechanism?
To quantify that question: based on the current issuance (~0.79% per year, ~948,000 ETH) and price (~$2300), can the network facilitate enough mainnet transaction activity to burn $2.24B in base fees each year?
Source: growthepie.xyz – Layer 2 transaction share relative to Ethereum
While $2.2B in a year seems like a major departure from the ~$600M that Ethereum is currently generating, it is worth noting that the network generated $3.78B revenue in a single quarter (4Q 2021), followed by an additional $2.16B in 1Q 2022, per TokenTerminal. But as seen in the chart above, the vast majority of EVM activity was still occurring on mainnet back in 2022 – a far different picture than what we see today, with Ethereum facilitating just 8% of overall EVM transactions. (Note we are using “EVM” here as a simple shorthand for all Ethereum-network layer 2 solutions, not necessarily any EVM-compatible chains.)
On an absolute basis, actual activity on Ethereum has held remarkably consistent since 2021, despite the protracted bear market sandwiched between relatively strong bull markets, typically hovering around 30-35M transactions per month. Note the near-doubling in L2 transaction activity in March 2024 after EIP-4844 went live.
Source: Ethereum and aggregate L2 transaction count by month via growthepie.xyz
So where does that leave us? Overall transaction count on Ethereum mainnet is at levels that historically have allowed for net negative issuance via fee burns while transaction activity on L2s is exploding, with over 90% of activity happening on these scaling networks (almost 450M transactions in August 2024). Without EIP-4844, value accrual to ETH would simply be going parabolic in lockstep with activity. But with EIP-4844, the opposite seems to have occurred, and the industry finds itself divided over whether this is the beginning of the end of ETH’s dominance or simply a necessary, minor step back on the path to a hyperscale future.
What do the numbers tell us?
To answer this, let’s take a quick dive into the numbers to get a rough estimate of what is needed for Ethereum to return to its former ‘ultrasound’ glory.
Current circulating ETH supply is ~120,300,000. At the current ETH price of $2,363 (at the time of writing) this gives ETH a circulating market cap of $284B.
As we discussed here, ETH staking rewards asymptotically approaches ~3.3% based on target 1M validators securing the network. At current levels, this works out to roughly 0.78% of new issuance per year: 948,000 new ETH issued worth $2.24B. We will hold this constant as actual numbers are approaching the network’s target levels and should generally hover around this level going forward. For reference, 1.75M ETH have been burned in total since the mechanism was put in place in August 2021.
Source: Glassnode
The network burns the base fees spent by users to execute transactions on the network. Average gas used per transaction has been ~100K since 2022. Simple ETH transfers require 21,000 gas, ERC-20 transfers (e.g. simple contract interactions) require 65,000, and more complex contract interactions rise rapidly from there. Median gas fees are slightly lower (80-85K), reflecting this higher cost of many transactions.
Gas prices on Ethereum are down substantially from historical norms, especially if measured since 2021. Over the previous cycle, it was not uncommon to see daily average gas prices exceed 100 or 200 gwei. While leading to high revenues for the network (often due to exorbitant transaction costs), this is the reason why layer 2 scaling solutions have been such a priority over the past few years. Since EIP-4844, average gas prices have continuously declined, from 25+ gwei for much of 2023 and 2024, down to below 6 throughout September. (As a reminder, gwei is a unit of measurement representing one 100 millionth of ETH. That is, 1 gwei = 0.000000001 ETH.)
Source: Etherscan
Taken together, the average transaction over the past several weeks has cost ~800,000 GWEI at a reasonable average gas cost (100,000 gas times 8 gwei). At $2363 per ETH, that amounts to roughly $1.89 base fee per transaction. This is a far cry from the $30-$50+ fees seen through much of 2021 and early 2022, but still orders of magnitude higher than the sub-cent fees on L2s and alt-L1 networks like Solana.
What does all of this tell us? Using the 2024 monthly average transaction numbers on mainnet (426M annualized) and the average gas price recently (2-5 gwei, the lowest level ever), the path to returning to a net negative issuance is daunting, to say the least.
Holding gas prices constant (again using 8 gwei optimistically), the network would need to grow annual transaction volumes by 178% to offset net network issuance, or 1.2B transactions. For reference, the highest 4-quarter period in Ethereum’s history was throughout 2021, where the network saw an aggregate 461M transactions. This was during the peak of the last bull cycle when almost all activity was happening on ETH mainnet. If we use yesterday’s average gas price of 2 gwei, the network would need 4.7B transactions, or 1012% higher. With the emergence of layer 2s and truly competitive layer 1s, it is difficult to understand where this massive increase in transaction activity will come from when those viable alternatives exist.
Given how consistent transaction counts on Ethereum have been over the past 3 years, we can also look at what gas price level is required, given current transaction numbers, to return to a net positive burn. This is likely a more viable path to net negative issuance, but also definitionally results in a negative impact on user experience given this means either higher transaction costs or, alternatively, far lower ETH prices. Based on current numbers, gas prices would similarly need to grow 180% to achieve that level of burn, or ~22 gwei. Historically pre-EIP-4844, this is not out of the realm of normality, but the resultant $6 transaction fee would be prohibitive for many users and limit any consumer-facing scalable use cases.
The third possibility is for the amount of gas used per transaction to increase substantially, up to an average of 278K per transaction. Simply, this would mean that the majority of transactions on the network would need to be heavily skewed towards far more complex interactions, such as heavy DeFi use. The ultimate impact of this, all else equal, would be a similar increase in transaction fees as the gas price, again a net negative impact on most users of the network. While DeFi users may be more insensitive to these higher transaction costs, the marginal more price-sensitive user would rationally leave to an L2 or alternative L1.
Naturally, each of these are also interactive and as such, changes in any of these metrics lead to changes in all of the others. For example, a higher skew towards DeFi-heavy transactions increases the average gas used per transaction. But those same transactions would push the amount of gas used per block higher (definitionally so if holding # of transactions constant), which programmatically pushes the network’s base fee higher, meaning fewer transactions are actually needed to reach that level. At the same time, the higher fees would drive many users offchain, meaning the transaction volumes would need to be sourced from marginal net-new users.
What about the blobs?
Missing from these calculations is the impact of L2 transaction growth and blobs on ETH fee generation and revenue. We are ignoring these for the time being given how low the fees L2s are actually paying post EIP-4844 despite record levels of activity (note: based rollups such as Taiko are slightly different here, but have minimal adoption compared to optimistic or ZK rollups). Since March, blobs fee burn has amounted to just 1,020 ETH (~$3.5M), vs. almost 500,000 in new ETH issuance over that same period.
Source: ultrasound.money
Fees have barely budged above the base fee of 1 wei aside from just a few days (LayerZero airdrop and “blobscription” mania). Note ‘wei’ here is not a typo. Wei represents 1 one-hundred millionth of 1 Gwei, or 10^-18 of 1 ETH. Given a single blob requires 131,072 gas, this fee is astronomically small. Over the past 30D, L2 blobs have accounted for just 103 gwei burned – essentially $0, despite representing over 400M transactions over that period. (5000 gwei is ~$0.01). The transactions to ‘carry’ these blobs are accounted for in the normal L1 transaction activity counts. In short, rollups are receiving 100% discounts on data costs compared to if they kept committing as calldata:
Source: Dune via Galaxy Research
There are discussions ongoing about how to ‘fix’ this pickle Ethereum finds itself in. One of those proposals is EIP-7623, which aims to specifically increase the cost of calldata for rollups. The reason and rationale behind this is beyond the scope of this post, but curious readers can dig in deeper here. At the highest level, EIP-7623 attempts to increase costs for any ‘user’ that uses Ethereum primarily for data availability, thus maintaining higher fees from L2s while ensuring fees for all other transactions on L1 remain low.
While we applaud the ingenuity of the proposal, we are not convinced that price discrimination with the goal of charging 90% of your customer base higher prices is the path for long-term sustained success, especially when the rest of the market sees fees trending to 0. Even if the rollups have the margins to accommodate fee increases, the level of increase would have to be substantial to generate adequate fee burn. Ultimately, those increased data costs would ultimately be passed through to users, increasing transaction fees, given L1 costs are explicit components of L2 transaction fees:
Source: An example fee breakdown of a transaction via Basescan
Conclusion
Generally speaking, Ethereum mainnet economic activity needs to increase roughly 180% from current levels in order to return to its former net-positive burn glory. Increasingly, this is a daunting hill to climb given the massive proliferation of L2 and L3 networks and the rise of viable alternative L1s where fees are orders of magnitude lower than Ethereum.
Viewed through this lens, we believe the most viable path requires a marked change in the makeup of L1 transaction types, skewing far more heavily towards high-gas contract interactions. This naturally requires a major shift in user base towards more price-insensitive users, specifically institutions that are willing to pay for the security of transacting directly on mainnet and at high enough transaction values to justify the fee levels. All other transaction types (simple sends, gaming, NFTs, consumer/retail interaction) are likely to move to L2s and alt L1s, and as we showed above, L2 settlement fees likely will not result in adequate fee burn for the network for the foreseeable future.
While there are early signs of this transition occurring, such as Blackrock’s BUIDL fund and the signs of adoption of RWAs, this is still very much a long-term transition; 5-10 years or more, and not a short-term transition by any means. With that said, the ecosystem of Ethereum as a whole (mainnet + L2s and L3s) continues its growth and dominance within the industry, by essentially any measure. While we do not expect the EVM’s dominance to suddenly drop off a cliff, we do expect the makeup of Ethereum’s economic activity to be markedly different going forward than what we have seen throughout its history to date.
He brings a wealth of knowledge from BlackRock, BGI, Goldman Sachs and more.