Rollups, which execute transactions outside of Ethereum (off-chain) and post integrity (execution) proofs and transaction data to Layer 1 for verification, have scaled Ethereum in a secure and performant way, enabling significant reduction in transaction fees and improvements in user experience. However, while improving scalability, rollups also introduced siloed execution that led to liquidity fragmentation and broke composability in Ethereum. In this article, we talk about the rollup fragmentation problem, its implications for users and developers, and we explore how emerging technologies can restore composability. Ethereum ultimately needs to provide a unified and composable ecosystem where users can interact with different applications seamlessly, across rollups. To get there, the focus must be on shared sequencing and real-time proof generation.

I would like to thank Dankrad Feist, Justin Drake, Noah Prevecek, Murat Akdeniz, Guy Wuollet, Mike Manning and Orest Tarasiuk for their valuable feedback!

The Evolution of Ethereum and Scaling Solutions

Ethereum has been at the forefront of blockchain technology, serving as the foundational infrastructure layer for decentralized finance (DeFi), non-fungible tokens (NFTs), and countless decentralized applications (dApps). However, as the Ethereum network has grown, so too have its scalability challenges, paving the way for alternative Layer 1s and rollups.

During the DeFi summer of 2020, we experienced the explosive growth of decentralized finance (DeFi) protocols like Uniswap, Compound, and Aave as DeFi grew from a $600mn to a $15bn ecosystem within a year [1]. Often referred to as “Money Legos”, these applications that run in the same execution environment all work in a composable manner, interacting and integrating with each other like building blocks. This composability property allows users to combine various financial services such as lending, borrowing and trading, enabling unprecedented innovation in dApps. However, this tremendous growth of DeFi also led to a dramatic surge in Ethereum gas costs. The network’s capacity was stretched as the average gas price skyrocketed from around 10-20 Gwei earlier in the year to over 100 Gwei, leading to transaction costs going well over $100 [2]. As a result, many smaller users were priced out of Ethereum, highlighting the scalability challenges Ethereum faced at the time. This created an opportunity for EVM-compatible alternative L1s like Binance Smart Chain and Polygon to emerge as a new home for developers to deploy low-cost applications and attract otherwise priced-out smaller users.

Ethereum introduced the rollup-centric scaling roadmap in October 2020 to increase transaction throughput without compromising on decentralization or security [3]. The rollup-centric approach leverages Ethereum’s inherent security while enabling high throughput. By allowing rollups to execute transactions off-chain and using the base layer for secure data availability and settlement, Ethereum promised to keep up with with alternative L1s, which often face challenges maintaining decentralization and security as they scale. Rollups also provide a new design space for innovation; developers can experiment with new VMs, mechanism designs or other primitives to provide superior experience and capture market share.

The Rollup Fragmentation Problem

While rollups are able to scale Ethereum, they also introduce fragmentation. There are two main reasons behind this fragmentation: The first is siloed execution across different rollup ecosystems and the second is long proving or settlement times on L1 [4].

Siloed Execution

Rollups and Layer 2s are execution silos. They each have their own state, sequencing rules and settlement methods. As a result, applications are only composable within a given rollup stack and completely disconnected from the rest of the Ethereum ecosystem. This leads to liquidity and a broader composability fragmentation, which are detrimental to the experience of Ethereum user and developers:

• Liquidity pools on one rollup are inaccessible to users of another rollup, which reduces the efficiency of DeFi markets.

• More broadly, applications on one rollup are not aware of the state of applications in other rollups and cannot make calls to (interact with) them.

Slow Settlement

Settlement (e.g. withdrawals) on Ethereum requires rollup execution to be verified on Ethereum. This introduces a latency which we call finality or settlement latency. Currently, optimistic rollups have a 7-day settlement latency due to fraud proofs and zero-knowledge rollups have a settlement latency on the order of hours [5].

In essence, rollup fragmentation is akin to the problem of walled gardens in Web2. Users and developers are vendor-locked-in into isolated ecosystems, leading to a suboptimal user and developer experience, as well as limited network effects.

How Is Rollup Fragmentation Impacting Us?

Let’s look at how fragmentation is impacting users as well as application and rollup developers.

Users

As a part of the rollup-centric scaling roadmap, Ethereum expects most of the usage to go to rollups. However, this is not the case today. Only 13% of active Ethereum users (EOAs or Externally Owned Accounts which are essentially user wallets on Ethereum) have also been actively using one of the 6 major rollup ecosystems in the last 30 days; on average 5.68mn EOAs were active on Ethereum per day, of which only 723K were active on at least one rollup [6]. While EOA-based and Ethereum-as-entrypoint-conditioned user analysis has its limitations, it’s also the best metric we have. Based on this; most Ethereum users are not using rollups. Rollups, while playing an important role in scaling Ethereum, are not serving the majority of Ethereum users yet.

Note that during the same period a total of 27mn EOAs were active on the mentioned rollups above.

Bridging problems

One thing we know for sure is that the bridging experience does not help the adoption of rollups. Assets on rollups can either be bridged by the canonical rollup bridge, by using external cross-chain messaging protocols like LayerZero or Chainlink CCIP or minted natively on the rollup like USDC or rollup native tokens.

Most assets on rollups are bridged using the canonical bridges. For example, of $13.20bn TVL on Arbitrum, approximately $10.55bn TVL is driven by Ethereum native assets (excluding ARB, Arbitrum native tokens or tokens from alternative L1s like BSC). Of the Ethereum native assets, 80% are bridged via the canonical bridge, 14% of assets native to Ethereum are minted on Arbitrum and 6% is bridged via external bridges. [7]

However, as mentioned earlier, canonical bridges introduce latency in Ethereum settlement. In order for fast withdrawals back to Ethereum to happen, users need to rely on third party liquidity-pool bridges or RFQ bridges. Liquidity-pool bridges are deposit contracts on chains that are controlled by a handful of entities, mostly using multisig wallets. If Alice wants to withdraw WBTC from Arbitrum to Ethereum, Alice deposits WBTC to the bridge contract on Arbitrum and she can withdraw WBTC on Ethereum given there’s enough WBTC on the liquidity-pool of the bridge on Ethereum. Once assets are bridged via the canonical contract they can be deposited on liquidity-pool bridges.

The cross-chain transfers these bridges can facilitate are capped by the deposits they have on the destination chain. If we look at the top 3 liquidity-pool bridges on Ethereum by TVL; Wormhole (Portal) [8], Axelar [9] and Stargate [10] collectively have approximately 96.5K ETH deposits. The amount of ETH deposited on the Arbitrum canonical bridge contract alone is approximately 1.28mn ETH or 13x the ETH TVL of top 3 liquidity-pool bridges on Ethereum [7]. The amount of ETH deposited on the 6 rollups analyzed above via canonical bridges is 2.6mn. Less than 4% of this amount can be bridged within a couple of minutes via liquidity-pool bridges given low deposits. The main reason users don’t deposit funds on liquidity-pool bridges is because it’s very risky. The liquidity-pool bridges have been exploited and lost billions of dollars. While the deposit amounts may be low, these bridges are efficient under normal circumstances, as they aggregate liquidity net transfers against each other. Major liquidity bridges have 3-4x monthly volume compared to their TVL [20]. However in black swan events, these bridges impose limitations on individual transfer amounts. Larger withdrawals require longer processing times and higher fees. I was using Wormhole during the Terra collapse to get UST out to Ethereum and convert it to USDC. The transfer took hours, during which UST lost a significant amount of value.

A safer alternative to liquidity pool bridges are intent based RFQ (request for quote) bridges. RFQ bridges replace the reliance on liquidity pools with market makers or solvers. These market makers provide Alice funds on the destination chain and receive Alice’s funds on the source chain. Across is emerging as the winner in this space. This model relies on market makers charging fees to cover for their cost of capital, which get significantly high with larger operations.

An unspoken fact in bridging funds between different ecosystems is the use of centralized exchanges. Market makers have a very high cost of capital compared to retail users. As a result, fast bridging is extremely important to them, not because of superior UX or cross-chain application composability but because bridging time actually impacts their return on capital. Interviews with market makers consistently imply that most market makers heavily rely on centralized exchange bridges to move funds between Ethereum and rollups.

UX problems

Users who don’t want to pay L1 fees not only decide what application to use but also need to decide which rollups to interact with. The process of choosing a rollup from Metamask is a significant UX friction. When we call an Uber, we care about getting to our destination, not about whether Uber’s backend runs on AWS or on GCP. As we are trying to reach a billion users, we have to provide better UX.

Developers

Fragmentation is an equally big issue for cross-chain applications. Most DeFi protocols deploy on different chains to meet the user demand, without a cohesive cross-chain product strategy. Applications on different chains feel different to the users. AMMs with deeper liquidity provide better execution to its users. AMM liquidity varies between different rollups: Uniswap has the deepest liquidity pool for the WETH/USDC pair on Ethereum. However on Base, Aerodrome offers the most liquid WETH/USDC pool. This is not good for users as they need to navigate which AMM gives them the best execution in each rollup.

In an ideal world, the experience of a given application should not depend on the infrastructure behind the instantiation of such an application. Most cross-chain teams spend significant amounts of time maintaining different deployments in different chains rather than focusing on shipping innovative features that would allow them to provide more value to users.

Recently, several Ethereum applications such as Frax, Aevo (formerly Ribbon Finance), Synthetix and Uniswap have launched or announced plans to launch their own rollup. These teams face the trade-off between providing low-cost transactions and enabling composability with the rest of the Ethereum ecosystem. Given these applications are Ethereum-native and their users are mostly on Ethereum, losing composability with the Ethereum application ecosystem is a significant friction. When we look at Frax Finance, only $12mn of the $470mn TVL is on Fraxtal, the rollup built by FRAX, while 95% of the TVL is on Ethereum [11]. In an ideal world, interacting with a low transaction cost version of your favorite app should feel like interacting with just another smart contract on Ethereum. And you shouldn’t even know about it.

Finally, application developers who are building new products on rollups are required to make a bet on the success of that specific ecosystem. While it’s possible to deploy to several rollups, in an ideal world, application developers should be betting on the success of Ethereum rather than that of a set of individual rollups.

Rollup liquidity

While rollups provide an environment for permissionless innovation and experimentation with new primitives, it’s really difficult for them to bootstrap liquidity. Lengthy withdrawal time creates a significant friction for users and TVL becomes a moat. It’s important for a rollup to attract liquidity because DeFi applications require liquidity to function effectively. Even with the strong brand of Frax, Fraxtal has struggled to attract non-Frax (FXS, FRAX etc.) assets to the rollup [11]. The friction in bootstrapping liquidity makes it very difficult for rollups to innovate and provide a great UX. Most rollups therefore resort to airdropping tokens (directly or through ecosystem applications) to attract liquidity which in most cases is mercenary, unsustainable capital.

Overall, fragmentation creates significant issues for all stakeholders involved in the ecosystem. As long as Ethereum liquidity and users are fragmented, Ethereum’s network effects are diminished.

The Endgame…

The ideal end state for Ethereum and the rollup ecosystems is one where everything feels as a single unified ecosystem with low-cost transactions. To achieve this, we need to achieve Universal Synchronous Composability, a term coined by Justin Drake. Universal Synchronous Composability means that applications across different rollups should be able to work together seamlessly and in real-time. To be more precise, a smart contract on rollup A can call a function on smart contract on rollup B and learn the outcome of that call, all in the same block. Let’s unpack this a little bit [4].

Blockchains, which are state machine replication systems, function when participating nodes agree on i) the ordering of transactions and ii) the state after the agreed-upon transactions are executed. When we apply this to the definition of Universal Synchronous Composability, we need both;

• A smart contract on rollup A can call a smart contract on rollup B, in the same block: This requires blocks to be sequenced by the same entity or entities that somehow form consensus on the ordering in the block for both rollup A and rollup B.

• A smart contract on rollup A can learn the outcome of the contract call initiated on rollup B, in the same block: This requires rollup A to either execute transactions for rollup B, or to verify a proof of execution integrity of rollup B [12].

In order to achieve Universal Synchronous Composability, we need to have shared sequencing and real-time proving. Real-time proving is the ability to prove state transitions in a rollup within one base layer block, which is 12 seconds for Ethereum. Real-time proving allows rollup deposits to be withdrawn immediately (real-time settlement).

Rollups today have only centralized, siloed sequencers, and are highly asynchronous, with a 7-day and hour(s)-long latency for optimistic and zero-knowledge rollups, respectively.

While Universal Synchronous Composability really gives superpowers to blockchains, it’s important to remember that the internet is asynchronous. The Internet is not synchronous, but is extremely fast. As a result, we don’t observe a problem with its asynchrony. There are prominent voices like Vitalik Buterin, who believe synchronous composability is overrated [13]. If our aim is to provide Web2.0 level UX, then low latency is sufficient for us to restore a unified UX for Ethereum.

And how to get there!

Earlier, we established that we need both shared sequencing and real-time proofs to achieve Universal Synchronous Composability. Let’s take a quick deep-dive into these areas:

Shared sequencing

Shared sequencing networks allow rollups that opt-in for shared sequencing to have their blocks be sequenced (built) by the same entity during a specific time period. Such a builder therefore has monopoly power over ordering transactions across multiple rollup ecosystems and can thus provide guarantees on including cross-domain transactions (e.g. contract interactions).

Shared sequencing networks introduce an auction model, which requires builders to bid for sequencing rights, usually above a reserve price set by the rollup operator. In this model, shared sequencers are going to be able to bid above the reservation price when there are cross-chain MEV opportunities that individual rollup sequencers cannot tap into. Users can pay shared sequencers a premium for cross-chain composable transactions. As more users demand cross-chain bundles rather than single rollup transactions, the revenue potential of centralized sequencers, that can only serve users in a single rollup, will go down, making it easier for shared sequencers to bid competitively. Shared sequencing will also be profitable in blocks where there is significant cross-chain MEV. The MEV opportunity has to either come from CEX-DEX arbitrage during times of high volatility or from a builder-favorable ordering (yes, including sandwich attacks!), which is a practice that current centralized non-shared sequencers do not engage in.

Based sequencing with pre-confirmations

In order to extend synchronous composability from only between rollups to also include Ethereum, more work is needed. This goal can be achieved by having Ethereum proposers (or builders, in reality) also serve as builders in a shared sequencer model for the rollup ecosystem.

In this scenario, Ethereum proposers, who are selected for a slot of 12 seconds, also have sequencing (ordering) rights for all participating rollups. However, such a window of 12 seconds is too slow for rollups as they need to provide much faster “soft confirmations” to their users. As a result, we’d need to introduce “pre-confirmations” (aka “preconfs”), i.e. commitments from the L1 proposer to sequence and include certain rollup transactions in sub-Ethereum block time. These out-of-protocol pre-confirmations require the proposer to provide a stake (a slashable collateral). Initial discussions for the stake amount of these builders range around 1000 ETH [14]. EigenLayer AVS, the dominant Ethereum restaking provider currently has 76 operators that has more than 1000 restaked ETH [15]. In addition to high collateral requirements, these cross-domain builders will need to have high bandwidth, high uptime and low latency. These requirements will be a centralizing factor.

Real-time proof generation

Traditional real-time proving requires the ability to create a SNARK (Succinct Non-interactive ARgument of Knowledge) or STARK (Scalable Transparent ARgument of Knowledge) within an Ethereum slot’s duration (i.e. 12 seconds). Such constructs are cryptographic “proofs” that ensure the validity of an execution trace, e.g. the state transition in the case of zero-knowledge rollups. Currently, STARKs and SNARKs are costly and take a long time to produce off-chain, and are expensive to verify on-chain. Scroll, the largest ZK rollup in terms of TVL, currently posts such a proof only once an hour [16] and proof verification accounts for ~75% of their on-chain costs [17]. While there’s progress with hardware acceleration, ASICs etc., we don’t expect such traditional real-time proof generation to be large-scale viable in the next few years.

One idea that’s been proposed by Justin Drake is to use liquidity providers and solvers who would take the risk of execution faults to provide loans to users [18]. However, as discussed earlier, such market makers or solvers would need to charge users for their risk and cost of capital, rendering such an approach a band-aid solution only.

Another alternative to STARK/SNARK real-time proof generation is via Trusted Execution Environments (TEE). Such TEEs are designed to be secure areas within processor hardware able to run code. TEEs guarantee that the code and data running inside the TEE cannot be tampered with. TEEs provide cryptographic proofs (“remote attestation”) to verify that the code they’re running  is genuine and untampered. Any changes or unauthorized modifications to the TEE break the remote attestation and would be detected. However, while TEEs have extremely valuable properties, they have been exploited in the past [19]. Whether for privacy or for verifiable compute, we must therefore reinforce them with defense-in-depth strategies.

How 𝚝𝟷 contributes to the endgame?

𝚝𝟷 is a rollup designed to address fragmentation and composability challenges in the Ethereum ecosystem. By leveraging AVS-secured Trusted Execution Environments (TEEs), 𝚝𝟷 introduces real-time proofs (RTP) that prove the integrity of 𝚝𝟷 execution to Ethereum in less than 12 seconds. RTP allow instant settlement on Ethereum and composability across different rollups. While this composability is NOT fully synchronous, 𝚝𝟷 provides a single-block asynchrony window, which we believe is acceptable while we as Ethereum community work on longer-term solutions.

Conclusion

In this post, we’ve explored the challenges of rollup fragmentation within Ethereum and its effects on users, developers and the broader ecosystem. Rollups, while essential in scaling Ethereum, have led to liquidity fragmentation and have broken composability across applications, limiting the network’s full potential. The ultimate goal for Ethereum is to achieve a unified, composable ecosystem where users can interact with different applications seamlessly across rollups. To get there, the focus must be on introducing some type of shared sequencing as well as making real-time proof generation feasible.

At 𝚝𝟷 𝚙𝚛𝚘𝚝𝚘𝚌𝚘𝚕, we are addressing the latter. Our mission is to unify Ethereum and the broader EVM rollup ecosystem, ensuring the scalability and composability needed for Ethereum to thrive. We believe TEEs are a practical solution to address a critical problem that exists today.

For more insights on Ethereum’s rollup fragmentation and how 𝚝𝟷 is addressing it, follow us on Twitter. We are actively seeking to collaborate with application developers and rollup builders for a more composable Ethereum ecosystem. If you share our excitement about this being a problem worth solving, get in touch with us today!

References

[1] "DeFiLlama," [Online]. Available: https://defillama.com/. [Accessed: Oct. 7, 2024].

[2] "Etherscan Gas Price Chart," [Online]. Available: https://etherscan.io/chart/gasprice. [Accessed: Sep. 30, 2024].

[3] V. Buterin, "A Rollup-Centric Ethereum Roadmap," Ethereum Magicians, Dec. 2020. [Online]. Available: https://ethereum-magicians.org/t/a-rollup-centric-ethereum-roadmap/4698. [Accessed: Sep. 20, 2024].

[4] J. Drake, “Comment on ‘Ethereum 2.0: The Road to Proof-of-Stake’,” Reddit, Mar. 21, 2023. [Online]. Available: https://www.reddit.com/r/ethereum/comments/191kke6/comment/kh78s3m/. [Sep 25, 2024]

[5] “L2 Risks,” L2Beat. [Online]. Available: https://l2beat.com/scaling/risk. [Accessed: Oct 2, 2024].

[6] M. Glinka, “Dune Analytics Query 3752807,” Dune. [Online]. Available: https://dune.com/queries/3752807?days_in_scope_t6c1ea=10&window_length_t6c1ea=30. [Accessed: Oct 5, 2024]M. Glinka, “Dune Analytics Query 3984338,” Dune. [Online]. Available: https://dune.com/queries/3984338?days_in_scope_t6c1ea=10&window_length_t6c1ea=30. [Accessed: Oct 5, 2024]

[7] “L2BEAT - Arbitrum TVL Breakdown,” L2BEAT. [Online]. Available: https://l2beat.com/scaling/projects/arbitrum/tvl-breakdown. [Accessed: Oct 13, 2024].

[8] “Wormhole: Portal Token Bridge - Multichain Portfolio,” Etherscan. [Online]. Available: https://etherscan.io/address/0x3ee18B2214AFF97000D974cf647E7C347E8fa585#multichain-portfolio. [Oct 13, 2024].

[9] “Axelar Gateway - Multichain Portfolio,” Etherscan. [Online]. Available: https://etherscan.io/address/0x4F4495243837681061C4743b74B3eEdf548D56A5#multichain-portfolio. [Oct 13, 2024].

[10] “Stargate Finance - Pool,” Stargate Finance. [Online]. Available: https://stargate.finance/pool. [Oct 13, 2024]

[11] “Frax Finance - Protocol Information,” DeFiLlama. [Online]. Available: https://defillama.com/protocol/frax-finance#information. [Accessed: Oct, 11, 2024].

[12] “We’re All Building the Same Thing,” DBA.xyz. [Online]. Available: https://dba.xyz/were-all-building-the-same-thing/. [Accessed: Oct 6, 2024].

[13] V. Buterin, “A lot of people forget this, but the reason why DAOs are interesting is because they are efficient, not because they are a legal entity. They enable large-scale cooperation that is not feasible with existing systems,” X (formerly Twitter), Feb. 27, 2022. [Online]. Available: https://x.com/VitalikButerin/status/1803459217933623418. [Accessed: Oct 6, 2024].

[14] D. Hoffman, R. Cooper, and J. Drake, "Fragmentation: The Future of L2s and Rollups," Bankless, Jun. 27, 2022. [Online]. Available: https://www.bankless.com/fragmentation. [Accessed: Sep. 25, 2024].

[15] “EigenLayer Operator Dashboard,” EigenLayer, [Online]. Available: https://app.eigenlayer.xyz/operator. [Accessed: Oct. 15, 2024].

[16] “Liveness,” L2Beat, [Online]. Available: https://l2beat.com/scaling/liveness. [Accessed: Oct. 13, 2024].

[17] “Scroll Onchain Costs,” L2Beat, [Online]. Available: https://l2beat.com/scaling/projects/scroll#onchain-costs. [Accessed: Oct. 15,2024].

[18] J. Drake, “Ethereum Sequencing and Preconfirmations Call #6,” YouTube, Apr. 12, 2024. [Online]. Available: https://www.youtube.com/watch?v=w91otDFmGpo&t=343s. [Accessed: Sep. 21, 2024].

[19] “SGX.Fail” [Online]. Available: https://sgx.fail/. [Accessed: Sep. 30, 2024]

[20] “Bridge Volume on Ethereum,” DefiLlama.com [Online]. Available: https://defillama.com/bridges/Ethereum. [Accessed: Oct. 16, 2024].”