Summary

Avalanche Community Proposal 77 (ACP-77) introduces a new type of validator, “Subnet Validators”, that lowers the barriers to entry to launch a dedicated blockchain by reducing both hardware requirements on subnet validators and upfront costs to register a BLS public key with the P-Chain for AWM interoperability.  

Below, Blockworks Research performs an analysis comparing the different dedicated blockchain solutions in the market today (Ethereum Rollups, Celestia Rollups, Cosmos appchains, and Avalanche Subnets) to help inform the community on the competitive landscape and where Avalanche fits. We then perform a scenario analysis to compare the total onchain operating costs of each using an example subnet, DEXALOT (ALOT).

Why launch a dedicated blockchain?

Before we dive into the analysis, we add some high-level context as to why one may want dedicated blockspace for an application or suite of products. Although mindshare has historically centered around a single execution environment housing a global shared state (Ethereum and more recently Solana), it has become apparent that these architectures likely cannot satisfy global scale and may not offer the best developer experience or value capture.

Performance

Ethereum mainnet places heavy costs on the end user during times of high demand, while in Solana’s case, the costs have been more lower, but congestion has rendered the chain near-useless where a majority of transactions fail to land. This lack of predictability for end users and developers may be fine for applications in the early stages of adoption but is not suitable for mainstream adoption.

To combat this problem, Ethereum has developed a rollup-centric roadmap that places Ethereum at the center of dedicated execution environments that leverage Mainnet for both “settlement” (fork choice rule and trust-minimized bridging) and “data availability” (data publishing). Solana, on the other hand, continues to work towards a global shared state machine by pushing the limits of commodity hardware while adopting multi-dimensional fee markets and other upgrades to optimize the stack.

Customizability

Global shared state machines also limit the ability of developers to customize their products for their end users. A recent example can be seen with the announcement of Uniswap v4 back in mid-2023, which required a new transient storage opcode (EIP-1153) on Ethereum mainnet to perform flash accounting that would drastically reduce gas costs for multi-pool swaps. Unlike app-specific chains, generalized blockchains like Ethereum and Solana must move slowly to accommodate all stakeholders of all sizes. To improve its own product, Uniswap (the largest application on Ethereum) had to lobby to get EIP-1153 included in the recent Dencun upgrade, nearly 10 months after the initial announcement of Uniswap v4.

Additionally, appchains allow developers to leverage additional design space at the node operator level to optimize the end-user experience. Examples of this can be seen more recently in the Cosmos ecosystem where chains can now enshrine an oracle network, a chain’s validators can run a decentralized in-memory orderbook, and even have full control of the block building process. Appchain/subnet architecture also enables the ability for a chain to accept a whitelisted set of assets to be acceptable forms of payment (including the subnet’s native token) or abstract gas fees entirely.  

Value Capture and Token Demand Drivers

Dedicated blockchains also enable applications to capture value at multiple layers of the stack and create new demand drivers for the application’s native token. Looking at Uniswap again, despite being one of the most successful applications in crypto and consistently the top gas consumer on Ethereum mainnet (consuming ~$230M in gas year-to-date), the UNI token has yet to capture any value from this usage and any additional MEV occurring on the platform.

On the contrary, subnets like DeFi Kingdoms have benefitted from its architecture choice by making its native token JEWEL the fee token on the chain, creating a demand driver that is a function of its success. DFK also distributes the gas fees to its various stakeholders, including burning 50% of the fees to create a supply sink. Over the past year, DFK chain users have paid nearly $900k in gas fees. In the future, DFK can also potentially enable staking of their native token JEWEL by leveraging the new subnet validator type introduced in ACP-77.

One of the additional strengths of an appchain is the ability to enshrine transaction ordering preferences in the protocol to optimize MEV internalization. Osmosis, a Cosmos appchain, identifies and executes backrun arbitrage opportunities on its platform that would normally leak value to external parties. Since the inception of this new feature, Osmosis has generated nearly $2M in MEV revenue.

While there are some notable limitations to launching a dedicated blockchain today, such as liquidity fragmentation, lack of atomicity, and additional developer complexity, developers looking to create the best end product for their users are likely to migrate away from global shared state machines over time as their product scales. This can be seen in dYdX’s move to a Cosmos appchain, MakerDAO’s plans to create their own chain, and most recently Aave hinting at an “Aave Network” in the future. Dedicated blockchains are an inevitability and will increasingly be created by a growing number of crypto app developers over time.

Why Avalanche Subnets?

When choosing a dedicated blockchain solution, developers should focus on an infrastructure solution that is interoperable, flexible, and performant. While Ethereum-based rollups and appchains built leveraging the Cosmos SDK have been the dominant solutions thus far, they both fall short in certain categories. Rollups currently lack a native interoperability standard, making every rollup reliant on third-party bridge solutions with varying degrees of trust assumptions. Additionally, the current reliance on a single sequencer and/or prover puts rollups at a higher risk of liveness faults. 

For the Cosmos SDK, although it is both interoperable, known to be relatively customizable, and battle-tested, there is tech debt that has not only hurt its flexibility, but has caused performance issues with many Cosmos chains susceptible to DDOS attacks due to a poor mempool design. Additionally, Cosmos chains have larger variable overhead for cross-chain messaging than Subnets do with AWM due to the nature of IBC light client verification versus Subnets only having to check participating signers on BLS multisignatures against the P-Chain. The tradeoffs here are higher fixed costs to sync with the P-Chain against higher variable costs for each IBC point-to-point connection. 

With the launch of AWM, the HyperSDK, and performance upgrades like Vryx and Firewood coming later this year, Avalanche is positioned to potentially become the best platform to build high-performance, interoperable blockchains. To learn more about how the Avalanche tech stack compares to Rollups and the Cosmos SDK, one can read our piece here.  

Ethereum and Celestia-based Rollup Economics

Onchain Costs

Both optimistic and zero knowledge (validity) rollups are execution environments in a modular stack that use other base layers for data availability and “settlement.” By leveraging fraud and validity proofs, rollups can operate under a 1-of-N honest minority assumption with a single sequencer and/or prover. This allows rollups to drastically reduce offchain costs with reduced hardware redundancy while still theoretically inheriting the base layer’s security, although there are new security risks around proving systems and current rollup sequencers being controlled by multisigs today. On the other hand, rollups must pay additional onchain overhead to these base layers to inherit their security.

Below is a chart that shows Optimism’s weekly onchain costs. One can see the drastic decrease in costs due to the Ethereum Dencun upgrade, which introduced blobspace via EIP4844. This EIP reduced rollup “L1 data costs” by ~99% and made rollup data costs have a separate fee market from Ethereum mainnet transactions.

One can see the unit economics of a rollup’s data costs ($/mb) drop significantly post-EIP4844 below. Today, Ethereum-based rollups are paying between $2/mb and $200/mb of data posted to Ethereum. Due to proof generation and verification overhead, validity rollups have higher onchain costs than optimistic rollup today, but should have higher unit economics at scale over time. One should also note that data availability unit cost is a function of demand for blobspace and will increase likely increase over time if Ethereum blobspace capacity doesn’t keep up with demand.

Rollups can also choose to post data to Celestia, an emerging L1 purpose-built for data availability that launched this past year and leverages data availability sampling to reduce rollup data availability costs further by an additional 92-99%. Below, one can see the unit economics associated with a Celestia-based rollup, with this specific rollup paying ~$0.20/mb currently. 

Offchain Costs (Excluding Labor)

Although rollups have reduced hardware redundancy today compared to normal blockchains, the costs of running a sequencer and/or prover are not insignificant. Recently, rollup-as-a-service providers like Conduit have found success abstracting away the friction of deploying an optimistic rollup and running the hardware that comes along with it. While hardware costs will vary greatly based on adoption of a given rollup, due to state growth and sequencer load, the current starting price to deploy your own rollup through Conduit is a flat $3k/mo. In addition to a sequencer, validity rollups must run a prover, which has been quoted at  ~$0.0019/tx in the past. One should note that proving costs are quickly becoming much cheaper and may drop another 10-50x in the coming year.** **Additionally, due to the lack of public information today, it is difficult to quantify the exact offchain costs of rollups.

Standardizing EVM Unit Cost Metrics

Paradigm recently released a piece on scaling Ethereum and is encouraging the space to stop using transactions per second (TPS) as a performance metric for the EVM and to instead use gas per second (GPS) as the standard moving forward. In this analysis, we use $/gigagas, where gigagas is equal to 1,000,000,000 units of gas, as a standard metric to compare different EVM dedicated blockspace cost structures.

Below we attempt to approximate the onchain unit cost ($/gigagas) of running different Ethereum-based rollups. We will use this metric later on in our analysis when comparing different dedicated blockspace solutions.

On a $/gigagas basis, OP-Stack chains like Optimism and Base are operating at the lowest unit cost of ~$1/gigagas, while Arbitrum and ZkSync are operating at ~$4/gigagas and $15/gigagas respectively. One should note that these unit costs represent rollups that use Ethereum for both data availability and “settlement”. For rollups that use Celestia for data availability (~$0.20/mb) while still settling to Ethereum, we estimate a unit cost of ~$0.38/gigagas. 

Lastly, both offchain and onchain costs of rollups may increase due to more hardware redundancy of sequencers and provers over time. While rollups can operate under a 1 of N honest minority assumption, they are mostly centralized today with single sequencers and provers typically run by the protocol teams. In the not-so-distant future, these protocols will likely incentivize permissionless validation of fraud and validity proofs, which may incur additional overhead. 

Cosmos Consumer Chain Economics (Leveraging Interchain Security)

Cosmos Consumer Chains share a similar architecture design to Avalanche subnets in a pre ACP-77 world where each chain has full flexibility to customize its stack but must leverage the validator set of the Cosmos Hub (similar to the Avalanche P-Chain). Although Interchain Security (ICS) is evolving into a more “opt-in” security model where validators can decide which consumer chains to support, today the entire validator set (180 validators) of the Cosmos Hub is forced to support every additional consumer chain as long as the agreement passes through the Hub’s onchain governance. Interestingly, ACP-77 turns the shared security model of the Cosmos Hub on its head and allows *any* validator set to secure *any* subnet as long as the validator(s) register with the P-Chain. This model may prove to be more scalable and cause less centralizing effects over time, although this remains to be seen.

Onchain Costs

The economics around Interchain Security has been a hotly debated topic in the Cosmos Hub community and still remains a grey area. Many in the community, including the potential consumer chains themselves, have no clear model to determine how much fees the Hub (and its validator set) should receive for the services rendered. Today, consumer chains pay the Cosmos Hub’s validators and stakers a percentage of token supply plus transaction fee and MEV revenue on its chain, with each consumer chain having a bespoke revenue share agreement. 

Neutron, a CosmWasm general smart contract platform and the first consumer chain of the Cosmos Hub, entered into a revenue-sharing agreement where 25% of all revenue is given to ATOM stakers. To align itself with the Cosmos Hub, Neutron airdropped 4% of the total 1B NTRN supply to ATOM stakers based on a snapshot performed on November 19th 2022. An additional 3% of total supply to ATOM stakers was given to those that voted on prop 72, which seeded the Neutron team with 50k ATOM to build the chain. 

Stride, the dominant liquid staking provider in the Cosmos ecosystem, entered into a revenue-sharing agreement with the Hub by sharing 15% of inflation, liquid staking rewards, transaction fees, and MEV revenue in exchange for the Hub’s security and 450k ATOM in the Cosmos Hub’s treasury. The 450k ATOM was liquid staked through Stride and deposited into a stATOM/ATOM liquidity pool on Astroport, an AMM on Neutron, and is solely owned by the Cosmos Hub as protocol-owned liquidity (POL).

Below is chart that shows both Neutron and Stride’s onchain costs paid to the Cosmos Hub, at  ~$50k and ~$250k year to date (~$12.5k/mo and $62.5k/mo), respectively.

As one can see, the total costs can vary greatly due to the structure of the revenue-sharing agreements being a function of a chain’s onchain adoption. Although Stride and Neutron are receiving the same amount of economic security in this scenario, Stride is paying much more due to having more revenue despite the Cosmos Hub having a lower take rate than it does with Neutron (15% vs 25%). This structure may create incentive misalignment whereby chains that find success decide to leave the Cosmos Hub’s security offering over time.

Offchain Costs (Excluding Labor)

For blockchains, offchain costs are largely a function of the amount of hardware redundancy the validator set has (i.e more validators = more redundancy). Similar to rollups, hardware costs can vary greatly from chain-to-chain based on a number of factors. Based on a survey of Cosmos Hub validators, the additional hardware cost of running an additional consumer chain is between $200 and $600/mo. With the current version of Interchain Security, where all 180 validators must secure each consumer chain, this amounts to between $3k and $9k/mo in offchain costs. One should note that with the upcoming launch of Partial Set (opt-in) Security, these costs will go down. To make a more apples-to-apples comparison of operating an Avalanche subnet, we will assume “opt-in” security is live on the Cosmos Hub and a given consumer chain has 8 validators (standard number of validators of an Avalanche subnet). Taking $400/validator, this would equate to ~$3200/mo in offchain costs to run a Cosmos appchain.

Economics of PAYG Subnet Validation (ACP-77)

Onchain Costs

ACP-77 proposes the idea of a continuous payment mechanism for subnet validators to register to the P-Chain, instead of a large upfront cost of 2000 AVAX (today) or 500 AVAX (referenced in ACP13), to lower the barrier of entry of subnet adoption. Similar to how rollups in the modular stack have onchain costs associated with using other layers for data availability and/or settlement, subnets would have to pay the P-Chain to maintain an up-to-date registration of its validator set. Although a deeper dive into the pricing mechanism of the continuous fee is required, ACP-77 alludes to a dynamic base fee that changes based on a target utilization of total subnet validators registered to the P-Chain, not based on the amount of economic activity on the subnet itself. This is an important distinction that differentiates Subnet Validation from other dedicated blockchain solutions today where onchain costs scale with demand of the chain, as outlined in the Cosmos section above. In ACP13, which was superseded by ACP-77, a minimum base fee of 512-4096 nAVAX/s is referenced based on the optimization work outlined here.

Below we perform a sensitivity analysis on a subnet’s monthly onchain costs assuming 8 validators at various AVAX prices.

Offchain Costs (Excluding Labor)

In addition to onchain cost savings, ACP-77 also reduces hardware requirements with subnet validators no longer needing to validate both the C-Chain and X-Chain. According to Ava Labs, subnet validation should lower the hardware costs from ~$250/mo to ~$80/mo using reserved pricing in an AWS EC2 instance, a ~64% decrease. With a validator set of 8, this would equate to ~$640/mo in offchain costs to run an Avalanche subnet.

DEXALOT Subnet Scenario Analysis

To properly compare the onchain costs of different dedicated blockspace solutions, we use real data from the DEXALOT subnet to quantify the daily cost in a scenario where DEXALOT is an:

Scenario 1: Avalanche Subnet post-ACP-77 with 8 validators

Scenario 2: Ethereum-based optimistic rollup using the OP-Stack

Scenario 3: Celestia-based OP-Stack rollup using Ethereum for “settlement”

Scenario 4: Cosmos Appchain with a 20% revenue share agreement

In Scenario 1, we take historical AVAX prices and a continuous base payment of 2048 nAVAX/s. For Scenario 2, although we do not have the amount of data that would be theoretically posted to a data availability layer from DEXALOT, we use the $1/gigagas unit cost for OP-Stack rollup onchain costs. For Scenario 3, we use the ~$0.38/gigagas referenced earlier for OP-Stack rollups that leverage Celestia for data availability and Ethereum for “settlement.” Lastly, for Scenario 4, we assume 20% of all DEXALOT subnet fees are going to a Cosmos Hub validator set of an arbitrary size.

Below we show the daily fees paid and gas consumed (in terms of gigagas) on the DEXALOT subnet since the beginning of the year as a point of reference,

While we reference approximate offchain cost figures for each solution in this piece, these figures can vary widely and may skew the final results of this analysis. We found it important to outline high-level offchain costs for each solution above for educational purposes only. We will strictly compare onchain cost across the various dedicated blockchain solutions.

As we can see above, unlike other dedicated blockspace solutions in the market today, subnet validation hypothetically decouples onchain economic activity from operating costs. In a real-world scenario, these costs would still likely fluctuate due to external factors like AVAX price and demand for SoV registration on the P-Chain, similar to blobspace demand from other rollups using Ethereum or Celestia for data availability.  

Ultimately, if ACP-77 was implemented today, onchain costs of subnet validation should be much more cost-effective than rollups that use Ethereum for both data availability and “settlement” while being competitive with rollup onchain costs using Celestia, the cheapest data availability layer on the market, plus Ethereum for “settlement.” 

Conclusion

As crypto applications mature and their development teams/communities seek more customizability of the product and create utility for the native token, applications will naturally gravitate to dedicated blockchains over time. With “OG” DeFi protocols most recently taking this next step in their product development cycle, this seems like the logical end state for most mainstream applications over time.

In order to capture the next wave of crypto app developers, dedicated blockchain solutions need to offer interoperable, flexible, performant, and cost-effective infrastructure. Not only is Avalanche positioned to become arguably the best platform to build high-performance, interoperable blockchains, our analysis shows that it is theoretically the most cost-effective way to launch a blockchain with redundancy to prevent liveness faults and native interoperability. Blockworks Research is heavily in favor of ACP-77, and we also advocate for future ACPs that focus on long-term cost structures for more mature subnets that resemble reserved instance pricing common in traditional web2 offerings.

 

This research report has been funded by Ava Labs Inc. By providing this disclosure, we aim to ensure that the research reported in this document is conducted with objectivity and transparency. Blockworks Research makes the following disclosures: 1) Research Funding: The research reported in this document has been funded by Ava Labs Inc. The sponsor may have input on the content of the report, but Blockworks Research maintains editorial control over the final report to retain data accuracy and objectivity. All published reports by Blockworks Research are reviewed by internal independent parties to prevent bias. 2) Researchers submit financial conflict of interest (FCOI) disclosures on a monthly basis that are reviewed by appropriate internal parties. Readers are advised to conduct their own independent research and seek the advice of a qualified financial advisor before making any investment decisions.