One of the key stories in crypto over the last two years has been the rise of Layer 2s. In a short span of time, they have attracted billions in TVL, and transaction volume on Arbitrum alone now rivals that on mainnet Ethereum.

But a key question remains. Are L2 tokens just worthless governance tokens? Or can real value accrue to these L2 tokens?

In this blog post, I'll dive into the key mechanisms through which L2s will accrue value.

L2s will primarily accrue value through two mechanisms:

1. MEV

2. Transaction fees

Let's explore each of these in greater detail.

Currently L2s don't generate MEV revenue, as they are run by centralized sequencers that don't want to be extractive to users. However, eventually the sequencers will be decentralized, which will open these rollups to MEV extraction.

To understand MEV better, we can think about it from two different lenses:

1. Harmful vs Good MEV

   a. Harmful MEV negatively impacts users. A sandwich attack on a DEX is a good example of this, as users end up executing their trades at worse prices.

   b. Good MEV is beneficial to users. For example, price arbitrage across DEXes helps prices across different DEXes stay aligned.

2. Intra-domain vs cross-domain MEV

   a. Intra-domain MEV occurs within a rollup, such as if you arbitraged prices across two DEXes on the same rollup.

   b. Cross-domain MEV occurs across different rollups, such as arbitrage across two DEXes on different rollups.

By combining these classifications, we arrive at four subtypes of MEV.

As an end state, most harmful MEV will likely be mitigated or rebated back to users. Why? Because this type of MEV directly harms users, there’s a large incentive for participants at all levels of the stack (wallets, apps, rollups) to create ways to more effectively deal with it.

Some great teams are already working on solutions to harmful MEV:

• CowSwap is a DEX that leverages batch auctions to mitigate harmful MEV

• MEV blocker is RPC endpoint that shields your txs from harmful MEV

• MEV Share from Flashbots is a protocol that designed to rebate MEV back to users

The methods to reduce harmful MEV will continue to improve, and it will get easier and easier for users to avoid harmful MEV.

Rollups themselves will also want to help users mitigate or rebate this MEV, in order to offer a superior user experience and get more adoption. Very few businesses accrue value by exploiting their customers!

So ultimately, harmful MEV is unlikely to be a source of value accrual for rollups.

On the other hand, good MEV benefits users and can serve as a significant source of value accrual for rollups. Since I expect most rollups to use shared sequencers in the future (such as @EspressoSys and @AstriaOrg), the question is how will this good MEV be allocated between the sequencer and the rollups.

Here it can be useful to separately consider both intra-domain MEV and cross-domain MEV.

Intra-domain MEV

It’s been argued that no rollup would join a shared sequencing network if it didn’t at least receive back intra-domain MEV, because a rollup could earn this amount of MEV on its own without joining the shared sequencer.

I disagree with this assessment, because whether a rollups receives back it’s intra-domain MEV ultimately depends on the network effects of the shared sequencer, which remains to be seen.

Shared sequencers enable cross-rollup interoperability. If the network effects from this are extremely powerful, and it ends up being a winner takes all market, rollups would choose to join shared sequencer network despite having to give up a portion of intra-domain MEV. Why? Because if they don’t offer this interoperability, while other rollups do, they could have difficulty getting users (and valuable tx fees).

That said, my view is that the network effects of shared sequencers will be moderate, due to lack of atomic cross-chain composability. With shared sequencers, you get a guarantee that two transactions will be included atomically across two chains, but not that they will be executed (since one of the transactions could be invalid and revert on one of the chains). James Prestwich lays it out simply in his blog post.

With moderate network effects, I agree with Sreeram that shared sequencers will likely need to rebate intra-domain MEV back to rollups.

Cross-Domain MEV

Cross-domain MEV will likely be split between rollups and shared sequencers. It's hard for rollups to capture cross-domain MEV themselves, so shared sequencers will want to take a cut of this MEV.

The exact division will depend on the pricing power / network effects of each specific shared sequencer.

Summarizing the above, here’s how I expect MEV value accrual to play out for rollups.

Note that there is one very important question arising from the diagram above. What percent of MEV is harmful vs good MEV? If a majority of MEV generated is harmful, then that limits the amount of value that rollups can generate from MEV. The answer to this question is still unclear.

Aside from MEV, transaction fees will also serve as a critical source of value accrual for rollups. Here are the various fees that rollups will collect from their users.

At a minimum, rollups will charge users (directly or indirectly) for the two costs they incur.

1. L1 costs for data publishing and execution

Rollups must pay for posting data and proof verification on the L1. They will recoup these costs from users.

2. Fees for operating the L2

Rollups networks incur various costs for operating the chain, including costs for sequencing, proving, execution, etc. They will recoup these costs from users.

In addition to recouping their costs, they could also optionally earn revenue from the following

3. Implementing minimum fees to protect against spam

In scenarios where their direct costs are very low, rollups will likely need to set a minimum fee that is artificially higher to deter spam and prevent bloating of the L2 state. So even if EIP-4844 brings L1 costs down to almost nothing, all of these savings won’t get passed to the user since it may otherwise result in lots of spam.

4. Implementing minimum fees to maximize profit

Additionally, rollups may opt to charge more than their directs costs plus spam protection. Why? Like any other business - to maximize profits. Just because it costs Starbucks $0.10 to produce a cup of coffee, doesn't mean they sell it at that price. They sell at the price that will maximize profit.

Example:

Imagine a rollup that incurs a 0.01 gwei costs for L1 and operating costs. They put a small 10% markup on that, and charge their customers 0.011 gwei, netting 0.001 gwei profit / tx. Now imagine that instead this rollup decided to set a minimum L2 gas price of 0.1 gwei. Even if the number of transactions reduced by 75%, they would still be making more profit. Like any product, there will be a profit maximizing price that the rollup can set.

In fact, L2s may already be doing this today. Arbitrum currently implements a 0.1 gwei minimum L2 gas fee for transactions. This is 100x what Optimism's min L2 gas fee is (0.001 gwei)

It’s important to note that only rollups that generate real network effects, and aren't easily copied (ie. have valuable state), will be able to generate pricing power and charge fees above and beyond the costs they incur, thus leading to additional profits. Otherwise competition will drive pricing down to operating costs.

But those that can generate these network effects and valuable state may become extremely valuable.

Lastly, L2s will generate congestion fees.

5. Congestion fees

Congestion fees occur when blockspace is saturated. Here you can see the spikes in fees over the last year when Optimism blockspace became congested.

As you can see, there were limited instances of congestion over the last year on Optimism (the same is true for Arbitrum). However, as demand for blockspace grows, fees from congestion could increase.

Summarizing the above, here’s how we can think about value accrual for L2s from transaction fees.

To put it all together, I believe L2 value accrual will stem from

1. Intra-domain good MEV

2. A share of cross-domain good MEV

3. L2 gas price above costs for spam protection

4. L2 gas price above costs to maximize profits

5. L2 congestion costs

As mentioned above, certain L2s will likely be quite valuable. Which L2s will be the most valuable? Those that can generate network effects and valuable state, which then allows them to command pricing power for their blockspace.

They will not run a simple middleman model where they charge a fixed % above the what their operating costs are. Instead, they will be in a position to set floor prices for their blockspace (both to reduce spam and/or maximize profit), as well as earn from congestion fees, and value from good MEV.

Finally, it should be noted that app-specific rollups have a lot of room to implement entirely different business models! For example, dYdX and ImmutableX both charge users a percent fee on trades. These different business models will allow these rollups to accrue even more value than had they just charged fees for gas / congestion.

Special thanks to @apolynya, @jskybowen, and @avichal for providing valuable feedback on this post

One of the key stories in crypto over the last two years has been the rise of Layer 2s. In a short span of time, they have attracted billions in TVL, and transaction volume on Arbitrum alone now rivals that on mainnet Ethereum.

But a key question remains. Are L2 tokens just worthless governance tokens? Or can real value accrue to these L2 tokens?

In this blog post, I'll dive into the key mechanisms through which L2s will accrue value.

L2s will primarily accrue value through two mechanisms:

1. MEV

2. Transaction fees

Let's explore each of these in greater detail.

Currently L2s don't generate MEV revenue, as they are run by centralized sequencers that don't want to be extractive to users. However, eventually the sequencers will be decentralized, which will open these rollups to MEV extraction.

To understand MEV better, we can think about it from two different lenses:

1. Harmful vs Good MEV

   a. Harmful MEV negatively impacts users. A sandwich attack on a DEX is a good example of this, as users end up executing their trades at worse prices.

   b. Good MEV is beneficial to users. For example, price arbitrage across DEXes helps prices across different DEXes stay aligned.

2. Intra-domain vs cross-domain MEV

   a. Intra-domain MEV occurs within a rollup, such as if you arbitraged prices across two DEXes on the same rollup.

   b. Cross-domain MEV occurs across different rollups, such as arbitrage across two DEXes on different rollups.

By combining these classifications, we arrive at four subtypes of MEV.

As an end state, most harmful MEV will likely be mitigated or rebated back to users. Why? Because this type of MEV directly harms users, there’s a large incentive for participants at all levels of the stack (wallets, apps, rollups) to create ways to more effectively deal with it.

Some great teams are already working on solutions to harmful MEV:

• CowSwap is a DEX that leverages batch auctions to mitigate harmful MEV

• MEV blocker is RPC endpoint that shields your txs from harmful MEV

• MEV Share from Flashbots is a protocol that designed to rebate MEV back to users

The methods to reduce harmful MEV will continue to improve, and it will get easier and easier for users to avoid harmful MEV.

Rollups themselves will also want to help users mitigate or rebate this MEV, in order to offer a superior user experience and get more adoption. Very few businesses accrue value by exploiting their customers!

So ultimately, harmful MEV is unlikely to be a source of value accrual for rollups.

On the other hand, good MEV benefits users and can serve as a significant source of value accrual for rollups. Since I expect most rollups to use shared sequencers in the future (such as @EspressoSys and @AstriaOrg), the question is how will this good MEV be allocated between the sequencer and the rollups.

Here it can be useful to separately consider both intra-domain MEV and cross-domain MEV.

Intra-domain MEV

It’s been argued that no rollup would join a shared sequencing network if it didn’t at least receive back intra-domain MEV, because a rollup could earn this amount of MEV on its own without joining the shared sequencer.

I disagree with this assessment, because whether a rollups receives back it’s intra-domain MEV ultimately depends on the network effects of the shared sequencer, which remains to be seen.

Shared sequencers enable cross-rollup interoperability. If the network effects from this are extremely powerful, and it ends up being a winner takes all market, rollups would choose to join shared sequencer network despite having to give up a portion of intra-domain MEV. Why? Because if they don’t offer this interoperability, while other rollups do, they could have difficulty getting users (and valuable tx fees).

That said, my view is that the network effects of shared sequencers will be moderate, due to lack of atomic cross-chain composability. With shared sequencers, you get a guarantee that two transactions will be included atomically across two chains, but not that they will be executed (since one of the transactions could be invalid and revert on one of the chains). James Prestwich lays it out simply in his blog post.

With moderate network effects, I agree with Sreeram that shared sequencers will likely need to rebate intra-domain MEV back to rollups.

Cross-Domain MEV

Cross-domain MEV will likely be split between rollups and shared sequencers. It's hard for rollups to capture cross-domain MEV themselves, so shared sequencers will want to take a cut of this MEV.

The exact division will depend on the pricing power / network effects of each specific shared sequencer.

Summarizing the above, here’s how I expect MEV value accrual to play out for rollups.

Note that there is one very important question arising from the diagram above. What percent of MEV is harmful vs good MEV? If a majority of MEV generated is harmful, then that limits the amount of value that rollups can generate from MEV. The answer to this question is still unclear.

Aside from MEV, transaction fees will also serve as a critical source of value accrual for rollups. Here are the various fees that rollups will collect from their users.

At a minimum, rollups will charge users (directly or indirectly) for the two costs they incur.

1. L1 costs for data publishing and execution

Rollups must pay for posting data and proof verification on the L1. They will recoup these costs from users.

2. Fees for operating the L2

Rollups networks incur various costs for operating the chain, including costs for sequencing, proving, execution, etc. They will recoup these costs from users.

In addition to recouping their costs, they could also optionally earn revenue from the following

3. Implementing minimum fees to protect against spam

In scenarios where their direct costs are very low, rollups will likely need to set a minimum fee that is artificially higher to deter spam and prevent bloating of the L2 state. So even if EIP-4844 brings L1 costs down to almost nothing, all of these savings won’t get passed to the user since it may otherwise result in lots of spam.

4. Implementing minimum fees to maximize profit

Additionally, rollups may opt to charge more than their directs costs plus spam protection. Why? Like any other business - to maximize profits. Just because it costs Starbucks $0.10 to produce a cup of coffee, doesn't mean they sell it at that price. They sell at the price that will maximize profit.

Example:

Imagine a rollup that incurs a 0.01 gwei costs for L1 and operating costs. They put a small 10% markup on that, and charge their customers 0.011 gwei, netting 0.001 gwei profit / tx. Now imagine that instead this rollup decided to set a minimum L2 gas price of 0.1 gwei. Even if the number of transactions reduced by 75%, they would still be making more profit. Like any product, there will be a profit maximizing price that the rollup can set.

In fact, L2s may already be doing this today. Arbitrum currently implements a 0.1 gwei minimum L2 gas fee for transactions. This is 100x what Optimism's min L2 gas fee is (0.001 gwei)

It’s important to note that only rollups that generate real network effects, and aren't easily copied (ie. have valuable state), will be able to generate pricing power and charge fees above and beyond the costs they incur, thus leading to additional profits. Otherwise competition will drive pricing down to operating costs.

But those that can generate these network effects and valuable state may become extremely valuable.

Lastly, L2s will generate congestion fees.

5. Congestion fees

Congestion fees occur when blockspace is saturated. Here you can see the spikes in fees over the last year when Optimism blockspace became congested.

As you can see, there were limited instances of congestion over the last year on Optimism (the same is true for Arbitrum). However, as demand for blockspace grows, fees from congestion could increase.

Summarizing the above, here’s how we can think about value accrual for L2s from transaction fees.

To put it all together, I believe L2 value accrual will stem from

1. Intra-domain good MEV

2. A share of cross-domain good MEV

3. L2 gas price above costs for spam protection

4. L2 gas price above costs to maximize profits

5. L2 congestion costs

As mentioned above, certain L2s will likely be quite valuable. Which L2s will be the most valuable? Those that can generate network effects and valuable state, which then allows them to command pricing power for their blockspace.

They will not run a simple middleman model where they charge a fixed % above the what their operating costs are. Instead, they will be in a position to set floor prices for their blockspace (both to reduce spam and/or maximize profit), as well as earn from congestion fees, and value from good MEV.

Finally, it should be noted that app-specific rollups have a lot of room to implement entirely different business models! For example, dYdX and ImmutableX both charge users a percent fee on trades. These different business models will allow these rollups to accrue even more value than had they just charged fees for gas / congestion.

Special thanks to @apolynya, @jskybowen, and @avichal for providing valuable feedback on this post