The Execution Layer Landscape

L2s, L3s and Everything in Between
April 18, 2023
15
MIN READ

Introduction

In the past year, the execution landscape has become increasingly crowded. While the movement towards rollups as a scaling solution has been clear for some time now, adoption is just now taking off. In the past month, the gas spent by rollup contracts accounted for well over 10% of the total Ethereum gas spend. 

from @funnyking's Dune Dashboard

The traction gained by Arbitrum and Optimism, along with the anticipated launches of zkSync, Starknet, and others, has accelerated the rate at which new execution layers are being built. Well over $2b in venture funding has gone into this category since 2018. With dozens of projects now building in the space, it has gotten challenging to follow all of them. This is exacerbated with the recent trend towards rollup SDKs/rollup as a service providers. The goal with this market map is to give a thorough overview of the execution landscape as it sits today.

Key Takeaways

Our research led to some interesting takeaways which help frame this piece:

  • The proving mechanism will be an implementation detail in the long run:
    ○  The advantage that validity-proven (zk) execution layers have over fraud-proven (optimistic)      execution layers becomes less obvious in a world with abundant blockspace. Optimistic designs      may hold up surprisingly well.
    ○  If/when zk becomes meaningfully better, optimistic systems can swap out fraud proofs for      validity  proofs.
  • As state growth becomes less of a bottleneck, execution becomes more of the bottleneck. This presents an opportunity for improved execution designs. This could be a major differentiator in the long run.
  • The deployment architecture will come down to economic preferences. More sovereign -> more value capture -> less composability.
  • Rollup frameworks and service providers will lead to an abundance of execution layers.
  • We expect that the design space for execution layers is likely to consolidate as the tech becomes more mature.
  • Remaining question marks - sequencer decentralization, cross rollup atomicity, zk proving efficiency, sustainable token models.
  • From an investor perspective - the long term winners in the execution market have likely already been funded.
  • Arbitrum is the clear category leader today in terms of both tech and adoption. Optimism has done an impressive job at attracting teams to their OP stack - most notably with Coinbase creating their own rollup - Base.
  • The key to long term success will be the ability to build a strong community of developers and users - more so than any technical detail.

We will dig further into each of these points below - starting with how to segment the market as it stands today.

Segmenting the Market 

Typically, execution layers are differentiated by the proving mechanism - either as “Optimistic” through fraud proofs, or “Zero Knowledge” through validity proofs. We will explore this distinction and explain why we believe there will be more important distinctions in the long run - namely the Virtual Machine (VM) design and the deployment architecture.

Proving Mechanism

When discussing execution layers, the primary distinction that captures attention is between optimistic rollups (ORs) and zk-rollups (zkRs). Optimistic rollups employ a system of fraud proofs to ensure the correctness of off-chain computations, whereas zk-rollups utilize validity proofs to guarantee the accuracy of the data processed off-chain.

Fraud proofs

  1. An action or transaction occurs within a decentralized system.
  2. A participant (or a set of participants) in the network claims that this action or transaction is invalid.
  3. In case of a dispute, the participant(s) can provide a "proof" that supports their claim. This proof is usually a small piece of data that can be easily checked by other network participants.
  4. Other network participants can then independently verify the correctness of the proof without having to store or process the entire blockchain.
  5. If the proof is found to be valid, the action or transaction is considered fraudulent, and the network takes appropriate action (e.g., penalizing the fraudulent actor or reversing the transaction).

Validity proofs:

  1. An action or transaction occurs within a decentralized system.
  2. The participant(s) responsible for the action or transaction creates a validity proof, which is a cryptographic proof that demonstrates the correctness of the action or transaction without revealing any unnecessary information.
  3. The validity proof is submitted to the network along with the action or transaction.
  4. Other network participants can then independently verify the correctness of the validity proof. This process usually involves checking the mathematical properties of the proof without having to process the entire blockchain.
  5. If the validity proof is found to be correct, the action or transaction is considered valid and can be included in the blockchain. If the proof is invalid or not provided, the action or transaction is rejected.

The key difference is that fraud provable systems rely on an assumption that there will always be at least one honest validator/sequencer to catch fraudulent activity whereas validity-proven systems reject fraudulent transactions from the start. Validity proven systems are generally considered more secure since they don’t rely on any trust assumption. However, validity proofs take significant time for more complex computation and the infrastructure required is very costly. With zkSync recently open sourcing their codebase, the specs for running a prover were made public. The lowest tier option recommended is a GPU system with over 40GB of vram. This would cost around $6k per month on Google Cloud, making it unsuitable for large scale decentralization. We explore the pros and cons of each in further detail below.

The main challenge in scaling blockchains today is the rate at which the state can grow without resulting in greater centralization. Any increase in throughput results in faster state growth and thus increases the cost of running the network which can lead to centralization as a byproduct. Zero-knowledge rollups (zkRs) have an advantage in this environment, as they only require the publication of proof data to the base layer, allowing more transactions to be included in a given amount of blockspace. However, as this blockspace bottleneck is resolved (EIP-4844, new DA services), optimistic rollups (ORs) may catch up to or even surpass zkRs in terms of scalability. This is because zkRs will still require costly proving infrastructure. As blockspace becomes more abundant, ORs can worry less about the cost of data publishing and maximize throughput. zkRs will also take advantage of blockspace abundance, but to a lesser extent. Plus there will always be overhead associated with running validity provers.

While the validity mechanism is an important technical discussion today, it will likely be an implementation detail in the long run.  Execution layers will converge on whichever system offers the fewest tradeoffs for their use case. Additionally, projects running on fraud proofs today will be able to validity prove their virtual machines once proving times/size comes down for zk. For example, Optimism could become a zk Rollup by validity proving mips. However, with the current state of proving systems, this would likely be too costly in the near term because of the size of the proofs required. Proof size is determined by the low level opcodes supported by the VM.  Systems like Fuel may be better suited for validity proving because the instruction set design is in some ways simpler than mips. In either case, the point is that if/when zk becomes clearly superior, fraud systems will be able to switch over. The main disadvantage these teams would face is a lack of expertise in zk, but as the space becomes more standardized and infrastructure is optimized, this will be less of an issue.

Our view is that in the long run, the proving system won’t be the crucial differentiator it is often portrayed to be today. But for this post we will go ahead and consider it. Just note that in the future, execution layers may jump categories.

VM Design

A key area of difference is in the virtual machine itself. To date, the dominant software environment for smart contract execution has been the EVM. The developer community that has been built around Solidity is robust and accounts for the vast majority of smart contracts deployed. As one would expect, this has led to the majority of new rollups and execution layers building around the EVM. Optimism and Arbitrum both implement iterations and there is a race to deploy the first zkEVM by teams at Scroll, zkSync, Polygon, and others.

The obvious benefits with this approach are compatibility with existing smart contract applications as well as being accessible to the largest community of smart contract developers and users. There has already been a clear example of this observed in alternative L1s such as Binance Smart Chain, Avalanche, and numerous others. In optimizing for compatibility with existing smart contracts, these Ethereum alternatives were able to capitalize on Ethereum’s lack of scalability and quickly attract applications and usage on their chains. Indeed, this has been playing out on early L2s like Arbitrum and Optimism as well. Nearly every major Ethereum application has instances deployed to both Arbitrum and Optimism.

There are, however, significant limitations that come in working with the EVM. The EVM, by design, cannot execute transactions in parallel.  For an L2 trying to scale the number of transactions that can be run in a given period of time, this introduces an upper bound on throughput. This has not been a major issue to date, given that the primary bottleneck is still state growth. But as blockspace becomes more abundant, the execution architecture will become more of a bottleneck. 

Starkware, Fuel, and others chose to build new systems which can provide much higher throughput than the EVM. The intricacies of these systems and languages could make for a separate research piece, but they generally share a common north star - parallelization. This alone presents a meaningful improvement over the EVM. 

The network effects which have been built up around the EVM over the years make it unlikely that one of these new systems will overtake it. As is shown in the 2022 Electric Capital Developer Report, Ethereum has far and away the most active developer community. Once you add up all the alt L1s using the EVM, this becomes even clearer. Therefore execution layers using the EVM are poised to play a major role going forward.

However, as the state growth bottleneck is alleviated, the need for more performant execution layers will become apparent. The recent announcement from Arbitrum of Stylus (supports WASM smart contracts), highlights the demand for these alternative execution designs. In the long run, there will likely be a few altVM category leaders which reach at least similar scale to Solana and other popular altEVM L1s from the last cycle.

Deployment Architecture

A final distinction worth understanding is the configuration in how the execution layer gets deployed. As briefly mentioned earlier, the most common architecture to date has been L2 smart contract rollups on Ethereum. Generally speaking, these execution layers fall across a spectrum of sovereignty. On one end you have enshrined rollups and on the other hand you have sovereign chains/rollups The definitions for these designs are quite contested, but for purposes of this post, we will use Polynya’s classification system.

Enshrined rollups: These rollups are integrated at the protocol level, with no reliance on smart contracts or governance. They provide the technical benefits of rollups at the L1 level. The downside is that upgrades can be slow, as they are tied to the consensus process of the underlying blockchain. As of now, Tezos is the only protocol moving in this direction. 

Regular rollups: These rollups rely on smart contracts deployed on a secure and decentralized chain (e.g., Ethereum) for their security. They inherit the security of the underlying blockchain but are upgradeable through the rollup's governance. Examples include Arbitrum and Optimism.

Validium: It is not a rollup but an execution layer that settles proofs on one external chain while using a different external chain for data. If validity is proven, this construction is called Validium. Examples today are ImmutableX and Sorare.

Optimistic chains: These chains operate similar to Validiums since they use an external chain for data. They can be thought of as optimistic Validiums. Additionally there is a fallback mechanism where if even one (or two) validators dissent, the chain falls back into a rollup. This implies a 1/n trust assumption as opposed to the typical ⅔ of n assumption for standalone L1s or sidechains. This is the design opted for by Arbitrum Anytrust chains like Nova. 

Sovereign rollups: These rollups settle themselves and only use the data availability layer for data and ordering. All other logic is under the sovereignty of the rollup. This results in the most L1 like experience with the major difference being how DA is published. There are a number of projects planning to build Sovereign rollups - especially within the Celestia ecosystem.

These execution layer designs offer various trade-offs in terms of security, upgradeability, and integration with the underlying blockchain. Generally speaking, more sovereignty implies fewer limits on performance or autonomy but introduces new risks around upgradability and security. There may also be added complexity around secure composability. Less sovereignty typically implies the reverse. Perhaps the most important tradeoff is an economic consideration. Our view is that the more sovereign the execution layer, the more clearly value can flow to that token. It is less clear how a regular rollup’s token accrues any value or is really needed at all. There are a number of considerations here around MEV, resource pricing, etc. which we plan to explain in detail in a future post.

Depending on the specific requirements of a project or application, developers will have to choose the most suitable rollup type to achieve their desired balance between security, decentralization, and scalability as well as economic preference. It's also important to note that much of this is still theoretical. To date, we’ve pretty much only seen regular rollups in production. The picture should become far more clear in the coming months/years.

Execution Frameworks/Services

As anticipated, the trend towards modularity in the blockchain space has given rise to numerous rollup SDK/frameworks and rollup-as-a-service providers. Companies building in this space raised more than $50m in 2022 alone. These solutions share a common objective of significantly simplifying the process of creating new execution layers. They aim to streamline the development of execution layers, akin to how the Cosmos SDK and Substrate facilitated the creation of Layer 1 blockchains. The vision is quite similar to the appchain thesis but rather than relying on building out L1 infrastructure, these appchains can inherit security from existing L1s. As Mustafa from Celestia highlighted, it’s important to understand the distinction between rollup frameworks and rollup-as-a-service.

Frameworks refer to software which can be re-used to build your own rollup. Services assist with the deployment and actual running of the rollup. A good example for understanding this is through the OP Stack and Conduit. The OP stack is a rollup framework developed by OP labs and implemented by Optimism, Base and others to build their own rollups. Conduit is a service provider which enables customers to deploy an OP stack rollup through infrastructure maintained by Conduit. 

With different service providers come different architectures and configurations. Some are configured for sovereign rollups and others are configured to be regular rollups, some are designed as L2s and some as L3s. These come with all of the tradeoffs mentioned above. Starkex was arguably the first example of a rollup as a service provider, with significant traction across dY/dX, ImmutableX, Sorare, and others. These execution layers can be configured as rollups which use Ethereum for data availability or as validiums which use an external DA source.

Generally speaking, the idea behind these frameworks and services is that there will be demand for rollapps/app specific rollups in the future - thus rollups will need to be trivial to deploy. Given the similarity to the Cosmos appchain thesis, it isn't surprising that a number of these projects are building in the Cosmos/Celestia ecosystem - Sovereign Labs and Rollkit are frameworks designed for Celestia rollups, while Caldera, Dymension, Eclipse, and others are services oriented around Cosmos/Celestia. Within the Ethereum ecosystem, the OP Stack, Arbitrum Orbit, and Stackr offer optimistic frameworks with Slush offering a zk framework. Ethereum centric services include Starkex, Conduit, and Opside.

Wrap Up

Today, the execution layer landscape looks quite differentiated. There’s zk vs. optimistic rollups, EVM vs. WASM, regular L2 rollups vs. sovereign chains/rollups, etc. As time goes on, we expect the landscape to converge down to just a few designs. Based on what we know today, it seems clear that regular EVM style rollups will be one of these designs. Given that state growth is expected to be less of a bottleneck in the future, it also seems likely that highly performant EVM alternatives will become necessary - whether that’s WASM, the Fuel VM, or another is still to be decided. The deployment architecture is a bit harder to predict, although we believe this will mostly come down to an economic consideration.

There are still open problems which will need to be addressed for execution layers to reach their final form:

  • Sequencer decentralization - essentially all of the execution layers today rely on a central point of failure. This negates much of the principles behind rollups. This is an important next step for rollup teams to take. There are also services popping up like Astria, Radius and others which offer this as a service.
  • Cross rollup atomicity - rollapps won't make sense for most financial applications unless this is figured out.
  • ZK proving efficiency - zk proofs are still too costly, although this is improving faster than expected.
  • Token design - its still unclear whether execution layers even need their own token. Arbitrum and Optimism tokens are essentially governance tokens which don’t capture value in the same way that L1 tokens do. These models are arguably unsustainable, but are there token models that make sense? As mentioned before, we plan to dig into this in our next piece.

Regardless of how exactly this shapes up, it seems clear that blockchains are nearing the point where scaling is no longer an issue. Disaggregating execution from the other layers enables never ending scale in theory. Rollup frameworks and services will make deploying new rollups trivial. The key question from here will be whether or not there will ever be enough demand to require this level of scale. We plan to discuss this question in depth in a future post, but Polynya gives a nice overview of that question here.

Given the diversity of technologies currently being used in the execution layer landscape and our expectations for future convergence, it's helpful to categorize the market to better understand the options available. With that in mind, let's segment the execution layer market into four broad categories: Optimistic EVM, zkEVM, Optimistic altVM, and zk altVM. For simplicity, we will leave the deployment architecture aside for now, given that most execution layers will enable multiple unique instances that are compatible with any deployment configuration. Below we will focus on general purpose execution layers since we already touched a bit on app specific infrastructure.

Optimistic EVM

This category is arguably quite crowded - most of the early rollup teams opted to go this route - optimistic and zk alike. Arbitrum and Optimism launched in 2021 and have steadily grown to become market leaders amidst L2s and L1s alike. In terms of TVL, DeFi Llama ranks Arbitrum 4th and Optimism 7th amongst all EVM chains with $1.86b and $875m respectively (as of 2/20/23).

Likely as a result of Arbitrum’s and Optimism’s dominance, with the exception of a strong team working on a new design called Specular,  there have not been many new teams working on optimistic EVM rollups. There are a few projects who essentially forked Optimism - Metis and Boba, but overall this category seems to have been largely decided.

Arbitrum

Work on Arbitrum began in 2018 by the team at Offchain Labs. Founded by Princeton University’s Ed Felten, Steven Goldfeder, and Harry Kalonder,  Offchain Labs raised a $3.7m seed round in 2019 led by Pantera, and subsequently raised a $100m at a $1.2b valuation in a 2021 series B led by Lightspeed.

Arbitrum deployed to mainnet in August 2021 and has since become the leading execution layer on Ethereum across nearly all metrics (outside of L1 itself obviously). Despite lacking a token, Arbitrum has been able to attract developers and users alike, with more than 200 active monthly developers according to Electric Capital’s latest report, and over 2.9m unique wallet addresses as of 2/20/23. Much of Arbitrum’s success has been a result of the ability to attract net new native projects, most notably GMX. GMX is a perpetual exchange with more than $650m in TVL and $67m in annualized revenue according to Token Terminal. This represents nearly one third of the total TVL on Arbitrum.

From a technical standpoint, Arbitrum has continued to evolve, releasing their v2 called Nitro last September which introduced efficiency and reliability improvements. Additionally, Arbitrum has begun rolling out Anytrust chains starting with Arbitrum Nova. Anytrust chains are execution layers built on the Arbitrum stack which rely on an external Data Availability Committee - an optimistic version of validiums. This results in greatly increased performance at the cost of a new trust assumption. Their plan with Anytrust is to support applications which may require less security and can take advantage of increased performance - essentially their version of appchains or rollapps. Another interesting development is their recently introduced programming environment, Stylus. Stylus will enable developers to deploy programs written in Rust, C, C++, etc., to run alongside EVM programs on Arbitrum. Stylus augments the Arbitrum EVM with a WASM virtual machine that runs alongside it. Considering this is a brand new announcement, there are still many details to be ironed out, but this will move Arbitrum into the alternative EVM category. 

The major problem which Arbitrum will need to address in the coming months/years is the decentralization of their sequencer set. To date, there has been a single sequencer managed by the Arbitrum team. This introduces significant counterparty risk. While it would be improbable that the Arbitrum team would sabotage their own project, it is still a potential attack vector for bad actors and overall goes against the concepts of decentralization and trust minimization inherent to blockchains.

Optimism

Founded in 2019 by early Plasma L2 researchers, Optimism raised over $175m across 3 rounds led by Paradigm and a16z, most recently raising $150m at a $1.65b valuation in March of 2022.

Since launching mainnet in the summer of 2021, has steadily grown to have more than $875m in TVL and over 100 dApps live. Unlike Arbitrum, Optimism launched their OP token in May of 2022 as an airdrop to early users of the network. The token currently has a fully diluted market cap of more than $11b according to Coingecko (as of 2/20/22). Similarly to Arbitrum, much of the TVL on Optimism comes from the native Velodrome DEX which launched on Optimism during the summer of 2022.

Optimism has also continued to iterate on their stack, most notably with Bedrock which is the upgrade of the network planned for this quarter. It will bring even further equivalence with L1 Ethereum along with improvements to performance. Optimism has promoted the idea of others building rollups on top of the OP stack. While the previously mentioned Boba and Metis are early examples, Optimism hopes to enable an ecosystem of rollapps deployed as L3s on top of the Optimism L2 with their upcoming releases.  

Optimism also has the same key  issue of a centralized sequencer. They plan to address this problem after the upcoming Bedrock release.

Takeaways

The main questions for both of these execution layers will be how they decentralize their sequencer sets and how they manage state growth as demand for their blockspace grows to L1 Ethereum levels.

In total over $800m in funding went towards these EVM execution layers since 2018. The category leaders in these areas seem pretty firmly entrenched. Although Arbitrum and Optimism still have some technical milestones to achieve, they have delivered on the vision for optimistic EVMs. New iterations may come about but it will be challenging to compete. 

From a venture perspective, most of the returns in this category will likely come from investments that have already been made. 

zkEVM

As mentioned above, zkSync, Polygon, Scroll, Taiko, and others have been in a race to launch the first zkEVM rollup. zkSync appears to be slightly ahead after opening their mainnet to builders and open sourcing their codebase on Feb. 16. Polygon is right on their heels with a beta of their mainnet planned to launch on March 27. The debate over what actually constitutes a true zkEVM has been quite contentious at times. There is significant differentiation in the aforementioned projects which was examined best by Vitalik himself in this piece from August. As shown below Vitalik classifies zkEVMs into different types based on tradeoffs. Lower types are generally more compatible but less performant, while higher types tend to be less compatible but more performant. Projects are not fixed in any 1 bucket, e.g. they may exist as type 3 today but move into type 2 as time goes on or vice versa.

With this categorization in mind we can examine some of the key players building in the space:

zkSync

Started in 2019 by the team at Matter Labs founded by early Ethereum researcher Alex Gluchowski, zkSynch has raised more than $250m across 3 rounds - not counting the $200m ecosystem fund. The most recent $200m series C round was announced last November and was co-led by Dragonfly and Blockchain Capital. Previous rounds were led by USV, and a16z.

zkSync recently opened their mainnet beta up to developers after almost exactly a year in testnet. There are already hundreds of projects planning to launch on zkSync once they are able to - this includes Uniswap, Instadapp, Balancer, and most of the blue chip DeFi protocols.

Based on Vitalik’s taxonomy, zkSync would fall into the Type 4 zkEVM camp. This is because zkSync converts Ethereum bytecode into their own SNARK friendly instruction set. This brings advantages in prover times but loses quite a bit of compatibility. 

Polygon Hermez

Polygon Hermez is actually the result of an acquisition made by Polygon in August 2021. Led by Ethereum OG Jordi Baylina, the Hermez team was one of the first teams to deploy a zk Rollup (payments focused) on mainnet - even launching their HEZ token in early 2021. Hermez announced their plans for a zkEVM in July 2021 and the acquisition/merger was announced soon after. Polygon pledged 250m MATIC tokens to facilitate this merger - worth around $250m at the time of the merger and $335m in early 2023. This was part of Polygon’s zk strategy which allocated $1b from the Polygon treasury to go towards promoting zk within their ecosystem.

Since the acquisition, Polygon Hermez has made significant progress in building out their zkEVM infrastructure, launching a testnet version last October. They recently announced a mainnet launch coming on March 27.

Hermez falls somewhere between Type 3 - Type 2 on the zkEVM axis. Modifications were made to some of the more complex data structures which resulted in improved proving speed while remaining fully EVM equivalent. Any application which works on L1 will likewise work in the Hermez zkEVM. Since much of the inefficiencies in zk proving the EVM are still there, the proof times are still long relative to other zk proved VMs.

Scroll

Work on Scroll began in 2021 led by Sandy Peng with comp sci PhDs Ye Zhang and Haichen Shen. Scroll announced their first round of funding in April of 2021, raising $30m led by Polychain and Bain Capital Crypto.

Although getting a later start than the aforementioned zkEVM teams, Scroll has quickly closed the gap. They deployed their pre-alpha Testnet last July and have since upgraded the testnet to enable smart contract deployment. Their fully open Alpha Testnet is expected to launch sometime in the first half of this year.

Similar to Hermez, Scroll is aiming towards becoming a Type 2 fully EVM equivalent zkRollup. 

Takeaways

The winners in the zkEVM space are still TBD but considering the head start in terms of time and funding, it seems quite likely that the teams above will remain leaders. It will be interesting to see how the tradeoffs between the different ‘Types’ will play out. In terms of adoption, will proof times or compatibility matter more? At a high level, it would make sense that they converge on either Type 1 or Type 4 given the tradeoffs involved. 

From a venture perspective, as with their optimistic counterparts, most of the returns in this category are likely to come from investments that have already been made, although there is more room for technical breakthroughs here. 

EVM Alternatives

Although not as saturated as the market for EVM execution layers, a number of projects are experimenting with new languages and runtimes. 

Fuel

Fuel was started in late 2019 by longtime Ethereum contributors and ConsenSys researchers Nick Dodson and John Adler.  In total, Fuel has raised over $80m led by Blockchain Capital and Stratos . 

The Fuel team deployed the first (arguably only) fully functioning optimistic rollup to mainnet in December of 2020. While that initial deployment only supported transfers, Fuel has now built a fully programmable alternative to the EVM and their own domain specific language called Sway which borrows heavily from Rust and Solidity. 

Fuel is quite unique in the L2 world given its UTXO data model. This design enables extreme parallelization leading to significantly higher throughput than an account based scheme like the EVM. 

Through 2 public test nets since September 2022, Fuel has grown a meaningful ecosystem of developers - with dozens of projects building in Sway planning to deploy on Fuel as soon as it launches. A mainnet launch is expected sometime in 2023. 

Starkware

Founded by renowned computer scientist Eli Ben-Sasson and serial founder Uri Kolodny, Starkware Industries is an Israeli based company developing layer 2 scaling solutions - primarily Starknet and Starkex. They have raised more than $280m since 2018 led by Polychain, Pantera, Paradigm, Sequoia, and Coatue.

Starkware has developed a custom turing complete programming language called Cairo which was designed for efficiently writing Stark-provable programs. This makes Cairo based rollups far more performant than their EVM counterparts. Proving overhead is greatly diminished although still an issue. Cairo is the basis for Starkware’s core products - Starkex and Starknet.

Starkex rollups are app specific rollups custom built on Starkware’s tech stack. Starkex rollups have been functioning on mainnet since late 2020, with notable successes such as dY/dX, ImmutableX, and SoRare (although it’s worth mentioning that dY/dX is planning to move away from Starkex to Cosmos). While these Starkex rollups are permissioned and had to be custom built, Starkware has also been working on Starknet.

Starknet is a general purpose validity rollup designed as a permissionless L2. Starknet deployed an alpha version of their mainnet in late 2021. While this beta has gotten some usage from developers, the goal has not been to attract widespread usage. A major upgrade is expected soon with the release of Cairo 1.0 and subsequent upgrade to mainnet beta. This upgrade will enable much more fully featured dapps to deploy, thus opening the network up to users. With over 130 dApps already planning to deploy, Starknet can quickly become one of the most widely used execution layers.

Aztec

Aztec, founded in 2017 by Tom Pocock and Zachary Williamson, is a London-based project focused on privacy and scalability solutions for Ethereum.Aztec has raised over $50m in funding from well-known venture firms, including Paradigm, a16z, Variant, and others. 

Initially, Aztec developed a privacy infrastructure called Aztec Connect that served as the encryption layer for Ethereum. However, Aztec has recently announced the sunset of Aztec Connect, with the goal of shifting their focus towards a fully programmable encrypted zkRollup. In addition to the sunset of Aztec Connect, the entire protocol has been open-sourced.

Although Aztec Connect was an important milestone for the project, the team decided to discontinue its development and support in order to focus on the development of two major, interlinked products:

  • Noir, a universal language for zero-knowledge programming.
  • A next-generation encrypted zkRollup.

Noir is designed to be a simple and powerful way to write programs in zero-knowledge and will serve as the smart contract language for Aztec's next-gen zkRollup. The Aztec team plans to release more Noir features and a full set of specifications for their newest zkRollup in the first half of 2023. Similar to Fuel, Aztec has so far worked through the UTXO data model - which as mentioned earlier opens the door for significant parallelism.

As the project progresses, it will be interesting to see how it competes with other established players in the space and how it attracts new developers and users.

Takeaways

Relative to EVM execution layers, there is naturally quite a bit more open territory for alternative VM designs. Considering that to date just about every relevant smart contract is written for the EVM, there aren’t any clear winners yet when it comes to newer designs. But despite the uphill battle faced by every new design, in the long run, the performance advantages from some of these new designs will likely be too great for developers to ignore. 

From a venture perspective, there is a clear head start on this front from teams like Starkware and Fuel, but there is likely still quite a bit of optimization that can be done at the execution level - meaning that there still may be some interesting teams building in this category.

Conclusion

The share of transactions on rollups and other execution layers is poised to continue to grow in the years to come, especially as Etheruem and other base layers optimize towards serving these execution layers with secure data availability. The anticipation of this has led to the landscape becoming increasingly saturated.

Ultimately, business development and the ability to attract interesting applications will be crucial factors in determining the success of execution layers in the long run. The ability to create a robust ecosystem will be essential for driving growth in this highly competitive and rapidly evolving space. The performance gains from these modular execution layers should open the door for unique applications which would previously have been infeasible on existing L1s.

In total, well over $2b in venture dollars has gone towards funding rollups and execution layers - much of this went to early builders like Arbitrum, Optimism, Starkware, and zkSync. While there may still be room for new designs, it's likely that most of the venture returns will come from companies already in development. This is especially the case with EVM style execution layers where market leaders like Arbitrum, Optimism, and zkSync already have a strong foothold. The non-EVM territory is more interesting, with more opportunities for experimentation.

Acknowledgments/Disclaimer

With how quickly the space is evolving, we are certain there are companies we missed and details we left out. Please reach out to us on twitter or via email to eli@stratos.xyz.

Thank you to 0xDinoEggs, Filip Siroky, Pseudotheos, Ryan Sproule and others who provided feedback or answered questions around these ideas.

Disclaimer: Please note that we have previously invested in both Fuel Labs and Dymension, which are mentioned in this article. However, our objective in publishing this content is not to promote or endorse any specific investments.

About Us

Stratos is an early-stage venture firm that backs transformative blockchain companies. We apply our deep research expertise to make convicted bets on top founders at the earliest stages.

We have backed numerous top companies in the space — most often as a lead investor. Our investments include Fuel Labs, Space and Time, Dymension, Subspace, and many more. Our team brings together a wide range of experience in venture capital, entrepreneurship and technology.

Disclaimer
This post is for information purposes and does not constitute an investment recommendation, investment advice, an offer to sell or a solicitation to purchase any securities offered by Stratos Technologies or any entity organized, controlled, or managed by Stratos Technologies or any of its affiliates and therefore may not be relied upon in relation with any offer or sale of securities. Any offer or solicitation may only be made pursuant to a confidential private offering memorandum (or similar document) which will only be provided to qualified offerees and should be reviewed carefully prior to investing. The views expressed in this post are the subjective views of Stratos Technologies personnel, based on information which is believed to be reliable and has been obtained from sources believed to be reliable, but no representation or warranty is made, expressed or implied, with respect to the fairness, correctness, accuracy, reasonableness, or completeness of the information and opinions. The information contained in this post is current as of the date indicated at the front of the post. Stratos Technologies does not undertake to update the information contained herein. This document should not be relied on for, accounting, legal, or tax advice, or investment recommendations. Stratos Technologies and its principals have made investments in some of the vehicles discussed in this communication and may make additional investments in the future, in connection with such vehicles without further notice. Certain information within this post constitutes "forward-looking statements", which can be identified by the use of forward-looking terminology such as "may", "will", "should", "expect", "project", "estimate", "intend", "continue", "believe", or the negatives thereof or other alternative terminology thereon or comparable terminology. Due to various risks and uncertainties, actual events or results or the actual policies, procedures, and processes of Stratos Technologies and the performance of the Fund may differ materially from those reflected or examined in such forward-looking statements, and no undue reliance should be placed on these forward-looking statements, nor should the inclusion of these statements be regarded as Stratos' representation that the Fund will achieve any strategy, objectives, or other plans. Past performance is not necessarily an indication or a guarantee of future results.