This report analyzes the current discourse surrounding ZK rollups in the L2 scaling space. It examines how ZK technology is positioned as a contender in the L2 scaling competition. Furthermore, the report compares and contrasts the offerings of various ZK rollup projects, which is then used to create a framework to determine which ecosystems have the greatest likelihood of success.
With Ethereum abandoning its sharding based-scaling roadmap, for a rollup centric-roadmap the crypto market has largely agreed to a consensus that future Ethereum scaling will come from a mix of general purpose and app-specific rollup projects. While Ethereum, a ‘monolithic’ blockchain is the core-focus of the rollup thesis other upcoming chains built on a ‘modular’ architecture (Celestia), also plan to develop a robust ecosystem of rollups with them
In this piece we aim to look at ZK rollup landscape and identify projects that are best positioned to win this budding market. Since, the monolithic base layer + rollup execution layer based blockchains are currently under research and early development, we will largely focus on ZK rollups building on Ethereum
To date some the major ZK projects have received billions of dollars in combined funding and significant attention from the market
Some numbers for context:
Matter Labs (ZK sync): $458 mn
Starkware: $282 mn
Hermez (acquired by Polygon): $250 mn
Aztec Network: $119 mn
Net sector = $1.4 bn~
If we look at the state of Ethereum L2 market, the combined TVL across all Optimistic & ZK rollups, Plasma, Validium solutions is equivalent to $3 bn~ (estimated based on data from L2beat, Defi Lama and Arbiscan). This is roughly equal to 5% of all TVL across different blockchains as per data from DeFi Lama. At present roughly 80%~ of the rollup TVL belongs to Arbitrum and Optimism, which are both Optimistic rollups. However, this does not imply that ZK based rollups are inferior to ORs, infact much of the success of the current ORs can be attributed to market timing. Optimistic rollups launched at a time when most new users were basically priced out of using the main Ethereum chain with gas costs as high as 30$ per send txn
This implies two main points:
ZK based rollups have yet to fully launch to market and onboard key projects that will eventually drive usage. Currently ZK rollups account for only 0.5% of all DeFi TVL most of which is accounted for by dydx
The demand for cheaper EVM-equivalent blockchains is growing massively, this can be seen from the 15x growth in Optimism and Arbitrum daily transactions from Nov 2021 to Jan 2023. Currently, Arbitrum and Optimism together settle about 600k-800k txns per day compared to 1 mn of ETH
In this section we will be comparing the key differences in the value proposition between ZKRs and ORs. Rollup centric discussion is often centered around computational complexity, trust assumptions and the system wide game theory. But it is probably better to take a step back and ask the more basic questions around which approach is the ‘better’, these questions are:
How robust is the system? Rich ecosystem? Deep Liquidity?
Is the system hack-proof? Can i withdraw and deposit easily from mainnet?
When looking from the above lens we see the following benefits for each side :
Optimistic rollups
slightly lower gas costs ($0.2~ compared to $0.4~ of ZK implementations)
Longer withdrawal periods/finality (upto 7 days)
Easy EVM compatibility
Lower compute cost and can work on less specialized hardware (easily decentralized)
ZK rollups
Almost instant finality
lower costs at scale
Harder to implement a general purpose EVM
Can offer privacy features, unique to ZK
Based on the comparison above, the larger consensus amongst the smart money investors and developers is that ZK is the superior tech. The discussion can be best summarized in an statement by Vitalik himself:
Conclusion: Zero-Knowledge rollups offer more benefits compared to OR’s and with additional features such as privacy, they will be able to serve use-cases that cannot be solved for by OR’s. Given the early state of the L2 market and significant backing and anticipation of ZK tech, it is reasonable to expect that the ZK market will at the very least 10x in usage and demand if their market share comes even close to the current market share of OR’s
However, it is likely that the key driver here will not purely be the tech but the quality of applications that are built on ZKR ecosystems. As an early proof we can already see that the Starkware based dydx is the highest revenue generating dApp with a 30D revenue of $8.3 mn compared to $3.8 mn by GMX, its close competitor on Arbitrum (data based on token terminal). The difference is less due to the base technology vs the UX and app use case.
Now that the stage is set, we can get into the current state of different ZKR’s in development. At present many teams are developing multiple flavors of ZKR’s that utilize vastly different tech and execution environments underneath. Although this creates a fragmented developer base that has to pick up one stack over the other, the different approaches breed variation which is overall good for the space. There are already projects such as Sovereign working on offering SDKs that abstract some of these complexities allowing builders to mix-n-match parts of the rollup stack
Before getting further into this discussion it is important to understand some crucial properties of a rollup. Rollups are implemented as a L2 solution on top of existing blockchains, this means that rollups sit on top of a txn’s life-cycle and are used for execution and settlement. The actual finality and consensus/DA layer activities are not in the scope of a rollup. A top down view of possible rollup configurations:
Regular Ethereum blockchain executing txns in its own EVM. Eg. Ethereum
Rollup executing txns on its own VM and posting states in form of block headers on the base chain, state proofs can be Optimistic or ZK based. Eg. Arbitrum and Optimism
Validium, similar to a rollup but the block data is posted off-chain with only a light version of the header posted on chain. This approach increases trust assumptions and is less secure
Modular rollup configuration that does not post state data on the main chain, only uses it for the data availability and consensus purposes. Such rollups will not offer bridges to the main chain, and will act as a closed L1 ecosystem
This configuration works similar to an Ethereum-alike config where a specialized settlement layer called Cevmos (in case of Celestia) can be used to bridge between different rollups. Similar to how one can move from Optimism to Arbitrum by first withdrawing to Ethereum
The last approach is similar to the Validium approach of off-chain DA, except that the modular layer acts as the off-chain data store, offering a more trusted off-chain data layer
And to top it all off, the actual execution layer (VM) itself can have many variations such as FuelVM, StarkEX (Cairo), Move, Rust, Sway, Yul and zkEVM. While some VM implementations are superior and more efficient for gas and hardware costs, EVM-equivalence (zkEVM) has become a key goal for many zk rollups as it offers the same developer experience as that of Ethereum. Both Arbitrum and Optimism benefitted massively from this approach allowing for easier network effects and developer migration. Thus, many projects choose to focus on zkEVM for their execution layer
EVM-”Equivalent” VM means that the developer experience on the VM will be 100% similar to that of Ethereum, including all the same dev tooling and methods. EVM-equivalence is a massive deal to developers since it means far less overhead when migrating from layer 1 to layer 2. Scroll in particular is hard focused on equivalence over compatibility
EVM-”Compatible” is a slightly looser version of EVM-equivalence, it basically means that while code can be written in the same language as an EVM contract (solidity), it is eventually compiled to another language that is native to that VM like in case of zkSync where it transpiles the code to another language called Yul behind the scenes. Thus, EVM-compatible rollups may not be 1:1 compatible to every Ethereum tool out there. In this regard Polygon’s zkEVM is farther from EVM-Equivalence than Scroll
For a more detailed understanding of various types of zkEVM, this article by Sino Global is a great starting point.
In this section we will discuss the major zk ecosystems under development that occupy the mindshare of most industry participants. As of Jan 23, these projects seem to be the only relevant players that are sufficiently capitalized and have enough social interest to succeed in this category. However, this does not mean that newer projects will not emerge on the scene in 2023
At the end of this section there are summarized tables that go over the key features, narratives and current state of these ecosystem, readers can reach it directly by clicking here.
One of the oldest projects in the ZK space, Starkware has two main products Starknet and StarkEX. The StarkEX product is a permissioned SaaS style rollup service that is currently implemented as siloed-smart contract based app specific rollups. Some of the apps using StarkEX infra are dydx, Sorare, Immutable, DeversiFi and Canvas.
Starknet is the general purpose version of the StarkEX implementation similar to an Arbitrum or Optimism but with ZK proofs. Interestingly fees on Starknet will be paid in the $STRK token. This has been live on mainnet since Nov 22. Overall Starkware based projects currently have about 75% of the entire ZK market, their app specific roll-out has been a great GTM strategy that earned them a lot of public trust before the general purpose rollup.
However, Starknet usage and TVL is 99% concentrated to one application (dydx) which has announced plans to leave the StarkEX service in the future, this will majorly shift market dynamics in favor of Starkware competitors
The first version of the zkSync protocol that launched in June 2020, was a ZK validium that stored data off-chain. It was also limited in capability as it could only be used for payment txns. zkSync 2.0 is the second iteration of the protocol which is a general purpose ZK rollup that is also EVM compatible. zkSync also plans to allow the V2 rollup to be used with off-chain data store. This service will be known as ‘zxPorter’ and enable even cheaper txn costs while sacrificing a little on network security
The key feature of zkSync is that it positions itself as Ethereum adjacent and its VM natively supports Solidity and Vyper. Eventually they also plan to expand to other languages such as Rust. This coupled with the fact that zkSync1.0 was well received despite its restricted use-case has created meaningful tailwinds for zkSync
Currently, the Polygon ecosystem has multiple L2 scaling projects being developed parallelly. These are Polygon Miden, EDGE, Zero, Polygon Supernets and zkEVM. The most anticipated being the Polygon zkEVM which is the the baby of Polygon and Hermez network. Polygon acquired Hermez network for $250 mn to further its zk development efforts.
The core advantages of Polygon zkEVM are that it is backed by the $10 bn FDV Matic network running as a PoS sidechain. With the way zkEVM is positioned right now, it is clear that Polygon team wants to migrate its userbase and developer base from the PoS chains and eventually sunsetting it, since many do not consider sidechains to be a proper L2. This means that zkEVM will see migration from TVL and already mature dApps currently live on Polygon PoS. Additionally, Polygon’s business team which is often considered to be one of the best in the market will also migrate to their zk ecosystem.
Scroll is also a zk rollup with its core focus on creating the most EVM equivalent execution layer. It is essentially competing for the same position as Polygon zk and zkSync but aims to offer the most ETH alike dev environments unlike like zkSync and Polygon zkEVM
Aztec team is one of the oldest team in the space and originally started the project in 2017. Aztec is a zk-SNARK based rollup with its core focus on privacy and encryption related use cases. Unlike other Ethereum rollups Aztec functions by integrating its SDK to existing dApps on Eth L1 layer. This allows Aztec users to relay txns via the Aztec Connect protocol which batches multiple user txns to one single txn but executes it on the L1. This approach offers privacy and lower gas fee due to batching but ultimately does not have dApps running on the rollup itself. Eg. If multiple users want to deposit $ETH to Lido, they can do so privately via the Aztec layer that will batch multiple requests into a single state change on the Ethereum main chain. Eventually the team plans to also allow native dApps that run on the Aztec rollup written in their custom Noir language
This approach is very different from a general purpose rollup in that it currently functions as a way to execute Ethereum txns in a private and anonymous way. A big concern with this implementation is that it could possibly result in legal scrutiny along the same lines as Tornado cash
The core thesis behind Fuel is that rollups will not only be scaling of Ethereum, but will be utilized by other blockchain ecosystems and will be essential to the upcoming modular blockchain ecosystems (Celestia). Fuel uses its own VM that is not fully EVM-compatible but is more optimized and has more features. FuelV1 focused solely on Ethereum scalability and later moved to a more general purpose execution layer called FuelV2. FuelV1 was the first optimistic rollup deployed on Ethereum in 2020. Currently, Fuel Labs is positioning itself to capture L2 market on ecosystems including and outside Ethereum. Fuel can also be implemented as both ZKR or OR, and even its own L1
However, based on the current state of the market Fuel appears to be betting on a narrative that will take much more to play out than Ethereum scaling which is an ongoing and widely fought turf war. Additionally having their own VM that isn't natively EVM compatible puts it at a disadvantage for dev onboarding
Loopring is a zk rollup that was designed for the specific use case of trading and payments. The ecosystem consists of a orderbook DEX, a wallet app and a native token. Later Loopring also added AMM pools in its ecosystem. Unlike others, Loopring is not a permission-less rollup and only applications developed by the core team are deployed on the rollup. Despite its unique positioning and trading focus, it is way overshadowed by other L2 DEXs like dydx and GMX who boast over 80-100x Loopring’s trading volume which makes it hard to make a case for Loopring in the current state
Recently, Loopring has also added NFT and gaming applications on its roadmap through partnerships, however this appears to be a move that came from trying to emulate what worked for other ecosystems
Honorable mention: Taiko is a EVM-Equivalent rollup, a new kid on the block that is currently missing from most zk ecosystem discussions. The team also hasn’t announced any fund raise for the project, however it seems to be making a buzz in the ZK forums and worth following in 2023
Aside from these protocols some other protocols claim to have zk-cryptography in their execution environment such as Dusk network, Celo, Aleph zero and Mina. However these are not rollups but their own L1s. Due to Ethereum network effects zk rollups seem to be better positioned as L2s rather than standalone L1s without any social capital. Thus, it is unlikely that these projects have a chance of competing with selection above
A summarized look at the state of different zk Rollups in development:
Key metrics for various zk Rollups at their current state (data as of 1st Feb 23)
*** Due to the experimental state of these projects and underdeveloped analytics tools, above network related data is simply indicative and could be different from the actual state*
*‘-’ implies that the stat is currently not relevant to this network. ‘?’ implies that the stat could not be found for this network ***
The above comparison matrix and network usage stats, can be distilled into a framework to rank these projects based on their the odds of success. The above discourse can be summarized as 6 key parameters that contribute to the odds of success for these projects
These parameters are:
Capital/Funding: Heavily capitalized ecosystems are obviously better positioned as capital enables more incentives, larger team sizes which is attractive for both users and builders
Winners: zkSync, Starkware and Polygon zkEVM
Ease of onboarding: As discussed in detail above ecosystems that offer similar dev environments have an easier time onboarding existing devs, applications and with them the userbase. In this case the EVM adjacent rollups will likely have an edge over their peers
Winners: Scroll, zkSync and Polygon zk
GTM: A good GTM strategy and market timing can often be the difference maker as seen in the case ORs Arbitrum and Optimism. In this regard it seems that Starkware’s rollup as a service strategy stands above its peers in establishing its brand, as seen from the success of dydx, Immutable and Sorare
Winner: Starkware
Network Effects: Network effects are a major driver in the crypto markets as networks that align themselves with existing popular communities have an easier time driving TVL and users to their platform. A good example would be Polygon’s adjacency to Ethereum that gives it substantial tailwinds compared to peers like fantom, NEAR and Tezos
Winner: Polygon zkEVM and Starkware
TAM: Most of the discussion around rollups is centered around Ethereum based rollups. While that is the most relevant use case at present, it seems that rollup tech is also key to non-Ethereum ecosystems like Celestia. Fuel appears to be the widest solution in-general offering many configurations of rollups and both ZKRs and ORs
Winner: Fuel
Specialization: General purpose rollups typically will have a bigger addressable market, however a rollup that specializes in a particular niche can also see favorable growth by cornering the market around that niche use case
Winner: Aztec Network
The evaluation framework reveals that, among all key parameters, Polygon, Starkware, and zkSync are currently the frontrunners in the ZK rollup space. Aztec, Scroll, and Fuel also show potential in individual parameters, however, they still fall behind the top three. Lastly, Loopring is relatively behind the group, except for its user-friendly frontend apps that allow for seamless onboarding.
The likelihood of success for these projects can be predicted based on the above evaluation framework, with the following ranking from highest to lowest, along with accompanying reasoning:
1st - Polygon zkEVM, zkEVM appears to be a milestone bet that would eventually replace Polygon PoS. Thus, we can only assume that the ecosystem will be extremely well backed
2nd - zkSync, due to its EVM compatibility, well funded ecosystem and a fairly successful version 1.0
3rd Starkware, due to the loss of dydx and the lukewarm reception of Starknet mainnet, which is bundled with many complexities
…
4th Scroll, outside the top 3 bigger ecosystem, Scroll is the only other ecosystem that stands to improve its standing given its scope and positioning. Additionally it also has the most EVM-adjacent VM making it an easier frontier for the Ethereum community
5th Fuel, due to its highly ambitious long term view on the market
6th Aztec, as it stands to benefit from a niche use case currently not serviced by any of the other competitors
…
7th Loopring, due to it restrictive scope and failure to truly dominate its niche despite a headstart