Written by Benjamin Funk
This article builds off of the previous MEV & The Evolution of Crypto Exchanges: Part I. While I recommend first reading through Part I if you haven’t yet, here’s a quick recap in case you’re short on time (or just want to ape into Part II).
Part I Recap & Reflections
Part I laid out the critical challenges and tradeoffs between between order matching systems in crypto today, including:
Limitations of futuristic AMMs
RFQs vs order books from a market microstructure perspective
The fundamental MEV-related tradeoffs and limitations of onchain order books
Design challenges for rollup-based exchanges
Batch auction analysis from a market microstructure perspective
Coming out of the other side, a few things become quite clear.
Many order flow aggregators using onchain order matching systems today generate, and will seek to internalize, MEV, as competition compels them to do so.
However, due to the technical challenges highlighted in Part 1, and the intensifying competitive landscape around exchanges, order flow aggregators new and old will accelerate the development and integration with platforms that leverage offchain order matching to enhance counterparty discovery across users. A few years down the line, the markets and assets that receive a stamp of approval from the global financial-regulatory-complex will leverage offchain order matching engines because they are a minimum viable architecture required for market makers and traders to engage with reliably.
As an industry often driven by short-term incentives, but also a pragmatic approach to solving today’s user needs, it’s easy for us to make assumptions about the evolution of the product landscape based on the current profiles of crypto users. However, we’re not even on the market's ground floor for crypto adoption or retail trading, let alone institutional participation. In five to ten years, the relatively price-insensitive MetaMask swappers are likely to be just a sliver of the millions of people that want to self-custody tokenized assets on blockchains and trade them efficiently.
As a result, we should be wary of the emergence of incumbents entering the crypto exchange market, but open to the opportunity to build and invest in companies that utilize offchain order matching engines while leveraging blockchains for settlement and self-custody.
While the order matching systems covered in Part I have their critical limitations, they also have many advantages for different markets. There will be markets for many valuable assets that the global financial-regulatory-complex won’t be able to compete in even if it wanted to, and onchain order matching venues will be the platforms to serve them. Moreover, the pace of innovation and diversity of thought taken by teams in building onchain financial products will create all sorts of interesting and also practical new paradigms that are worth exploring — as technologies to facilitate exchange and their impact on market structure and MEV.
After releasing Part I, I received a lot of feedback along the lines of “Great article; the challenges and tradeoffs in exchange designs are clear; how do we solve MEV though?”
While it’s difficult to answer that question, it’s worth considering that the more important questions might be how do we create the best exchanges and for what markets can the onchain exchange primitives we’re building be competitive? While the structure of MEV does shift according to blockchain design — such as mempool configuration, encryption schemes, and fee markets — the biggest contributor to the shape of MEV will be the exchanges where users trade crypto assets. With this in mind, we need to examine the impacts of new primitives around exchange design and their impacts on market structure.
Using our framework from Part I, this article will attempt to break down these questions by evaluating how new onchain primitives might impact market structure and MEV. Assuming much of MEV is downstream of adopted exchange designs, this will help give us a clearer picture of how the future of MEV might play out.
Beyond the LVR-minimizing AMMs and privacy-preserving batch auctions described in Part 1, there are new tools that plug into and build on top of onchain order matching systems. These tools don’t directly change how order matching *works*, but aim to help users and the dapps that aggregate their order flow better express and redistribute the value of the MEV they generate. In this section, I break down their unlocks and potential merits as means to improve the state of onchain exchange.
There are three contexts in which intents are discussed today:
UX-improving architecture separating offchain computation with onchain verification
New order matching system to compete with exchanges of today
Technological substrate to improve our ability to coordinate
For the scope of this piece, I’ll focus on just 1 and 2. However, intents as a technological substrate to improve our ability to coordinate, taken to their hypothetical limits, are one of the most exciting and wild frontiers in crypto today. In this context, we need to take a step back and see how intents can compose with the variety of cryptographic and financial primitives we’ve already created onchain.
The article that intent-pilled me on this is Christopher Goes’s Towards heterotopia - the prerequisite cultural and technological substrate for a return to a world of scale-free credit money. I’m pretty convinced we’d be better off redirecting techno-futuristic e/acc energy into investing time, energy, and capital to render the vision in this article alive than whatever we’re doing now. Alas, best to give this subject the scope it deserves.
With this framework, intents are a way for developers to design decentralized applications and deliver better UX for users. Intent-centric smart contract design leverages offchain computation to perform onchain actions that are constrained in the set of potential state transitions they can trigger on blockchain through some verification mechanism. As a result, dapps and DEXs can help users achieve their “intents” faster and at a lower cost in terms of price and gas.
My colleague Nick described what this looks like on blockchains like Ethereum today better than I could. Many dapps on Ethereum, such as Across, UniswapX, and CoW Swap, use an intent-centric smart contract design in that they leverage offchain computation and optimistic verification to minimize latency and costs while preserving safety.
This design pattern is especially helpful in the context of onchain exchange. For example, the resulting price improvements from leveraging intents on UniswapX on a single chain are the product of enabling fillers to leverage offchain logic, access offchain liquidity, and incorporate offchain prices into their quotes. This also allows fillers to quickly assess and provisionally fulfill users' intents for cross-chain swaps by lending funds, all while some oracle (like the UMA Oracle in Across’s case) ensures the security and finality of these transactions. Intents systems also lean on the gas savings that come with “coincidences of slots'' from submitting a collective batch of intents onchain, as opposed to individual orders. It’s cheaper to verify 1 batched transaction than verify that 100 different transactions followed a correct state transition.
New, intent-centric blockchain architectures and virtual machines also benefit from separating offchain computation and onchain verification. In this regard, their differentiation stems from how these new designs enhance multi-party coordination, and whether state transitions are verified through the use of invariants or oracles.
Taking things a step further, some believe that by operating outside of the constraints of the Ethereum transaction format, users will be able to more granularly specify conditions of exchange and receive better outcomes in terms of trade execution.
There’s good reason to look at the state of order books today, then look at an intent-centric system for exchange, and conclude that intents could represent a needed evolution from order books. Specifically, the fact that order books are characterized by an order to trade ratio of over 100-to-1 could be an indication that limit orders aren’t cutting it as a means of expressing traders’ preferences. Moreover, it’s possible that in leveraging a common language for expressing trading preferences across disparate financial domains, intent-centric exchanges could clear trades across broader available liquidity than a traditional exchange, ultimately improving the efficiency of counterparty discovery.
In the context of exchange, intent-centric architectures can be implemented to tackle this by allowing users to add arbitrary database transactions to their order, enabling them express their preference functions for how state (probably the state of their balance of an asset) can be updated in the future. These database transactions specify additional conditions of exchange beyond the quantity and price of the desired assets. Solvers then calculate the pareto-efficient frontier of potential new states expressed across multiple parties, but are constrained in how they update the distributed ledger serving as a settlement layer. This ensures that the solvers abide by the validity predicates that represent the constraints on state transitions specified by all users' intents. As a result, nodes are enabled to verifiably update their copy of the distributed ledger without having to engage in the computationally heavy task of solving intents in a pareto-optimal way.
There are a few reasons to be skeptical that intents represent the solution to the challenges mentioned above. The core reason behind the symptom, which is a 100-to-1+ order-to-trade ratio, is that public order books reveal traders' preference functions in the first place, forcing them to break up trades into smaller orders to conceal the implied information they may reveal. The biggest reason traders are forced to submit many orders to express their preference function isn’t generally because their demand function is changing, it’s because they don’t want to reveal their demand function to the market. By flashing multiple orders, traders can effectively keep their demand hidden from the market.
Moreover, intents aren’t a new way of order matching, and could even deepen the existing challenges in creating competitive onchain exchanges today. Intents can be applied on top of onchain or offchain order matching systems, including order books and RFQs, but don’t represent a new way of operating an exchange that can solve the core challenges of onchain order matching. Although they follow very different designs, both Anoma and SUAVE represent architecture conducive to building onchain order books. Both allow users to specify arbitrary database transactions on top of their limit orders, and enable developers to create applications that guide users to do so. However, they don’t overcome the critical challenges of onchain order books already highlighted in Part 1.
Long-story short, there is a fundamental and inescapable tradeoff in the type of MEV onchain market makers have to deal with between order book configurations where multiple leaders provide input into the sequence of orders or a single leader decides the sequence of orders.
In all intent systems, the computationally complex work of clearing intents is delegated to the solvers, who run heavy algorithms to clear these intent-based orders optimally. The higher the breadth and variety of database transactions takers can define, the more an intent system looks like a complex combinatorial auction — participants can place bids on combinations of discrete heterogeneous items, or “packages'', rather than individual items or continuous quantities. Ultimately, intent-based architectures create more computational complexity for solvers than a traditional order book, akin to the time taken to solve a combinatorial auction. This results in the issues core to the MafiaEV problem described in Part 1, whereby the liquidity and user uptake on the exchange would likely suffer from the slower updates to prices.
It’s worth pointing out that some intent systems, such as Khalani, allow developers to create composable modules for their users to specify intents, enabling them to be compiled into more fungible atomic units. As a result, these modules help direct the runtime search for solvers, decreasing the heaviness of the computation solvers must undergo. However, the more fungible intents are for solvers, the less expressive they are.
While there’s no reason to think order books represent some pinnacle of evolution for exchange that can’t be surpassed and enhanced by new technologies, there’s reason to be skeptical that specifying one’s preference function more granularly will change the fundamental problems that challenge the scalability of onchain order matching systems today.
Along with this, it’s unclear to what extent traders want to, or should, go beyond defining slippage tolerances and limit orders. There’s not much evidence to support the claim that retail traders want to express very complex preference functions in the context of financial transactions. Either Joe wants to buy an asset because he wants to gamble on something, or Joe doesn’t want to gamble and outsources thinking about his preference function to someone who does things for him. Within an institutional context, sophisticated trading firms likely have internal systems that add some version of database transactions on top of their orders today, and have developed their own internal infrastructure to do so.
While intents represent an exciting frontier for designing distributed systems, enabling net new applications that transform how people coordinate, it’s unclear what their architectural edge is in creating the onchain exchanges of the future to compete with their offchain counterparts. It is possible that intent-centric architectures may help reduce MEV for assets whose order matching needs to be onchain by empowering users to better define constraints around trade execution. Expressing those constraints as validity predicates may be a more effective way of preventing MEV than retroactively enforcing a punishment on a solver that has violated a user’s intent.
The opportunities and risks associated with different configurations of the mempool by which users submit intents is also another major factor to take into account when evaluating how they might improve or harm the quality by which intents get executed. This topic is covered and debated in great depth by @0xQuintus, @gakonst, and @cwgoes in the following pieces: Intent-Based Architectures and Their Risks and Towards an intent-centric topology.
Unlike batch auctions covered in Part 1, where a user submits a trade without specifying a price, order flow auctions (OFAs) offer the right to execute a transaction at a price the user has already specified. In an OFA, the searchers/solvers/fillers who won the auction can execute a trade if it's within the user-specified restrictions.
Reintroducing the key stakeholders in matching a trade:
The Buyer (Party A)
The Seller (Party B)
The Market Maker – the intermediary facilitating the interaction
Bidders and the OFA platforms that host them (onchain or offchain) make money from pushing Party A and B on the opposite sides of a trade closer to their breaking price.
Just as intents need to be delineated from the systems that match orders, OFAs compose with and do not represent a new, standalone way of handling onchain order matching. In practice, the bidders in many of today’s OFAs enable bidders to take on counterparty risk by serving as market makers. In this case, those OFAs are implemented on top of order matching systems like RFQs or order books, where the searcher/solver takes on some liquidity risk in filling the user's trade as well.
In this paradigm, the value OFAs extend to order flow aggregators and their users goes beyond what the dapp and user could ever do alone, making the business model of OFA providers more sustainable long-term. However, it’s worth assessing whether the solvers are market makers taking on liquidity risk, or arbitrageurs taking none, to assess their long-term competitive advantage.
In tradfi, OFAs have emerged as a function of retail-friendly brokers leaning into order flow segmentation as their core business model. Naturally, retail-friendly brokers like Robinhood have a high concentration of retail traders who are much more likely to submit trades uncorrelated with broader market movements.
As a result, market makers (such as Citadel) pay those retail-friendly brokers to access that flow (meaning Robinhood auctions it off), because it helps them delineate between uninformed or informed traders, and gives them better guarantees around the non-toxicity of the flow sent from those brokers. As a result, unsophisticated traders coming from these retail-friendly exchanges could be profitably served by market makers at narrower spreads than on a regular exchange.
If the retail-friendly exchange retrospectively observes that a market maker doesn’t provide prices that compete with the NBBO, the retail-friendly exchange will auction the order flow off to another market maker (this is a simplified explanation). On Robinhood, this OFA is paired with an RFQ as the underlying order matching system.
Decentralized exchanges could leverage, systematize, and accelerate order flow segmentation by opening up the access and development of smart contracts that delineate the extent to which a trader is considered informed. Onchain market makers (passive or professional) could then offer different spreads to traders depending on the smart contracts or wallets they initiate trades from.
To visualize how this works, let’s look at an oversimplified runthrough of a switch from pre to post-order flow segmentation. In this example, let’s assume that Exchange 1 is implementing order flow segmentation, while Exchange 2 is not implementing order flow segmentation.
1. Pre-order flow segmentation, let’s imagine the breakeven point for market makers to service aggregated informed and uninformed traders is 5bps.
2. The competition between market makers is so high that all flow is serviced at this breakeven price on both Exchange 1 and Exchange 2.
3. New smart contracts are created on Exchange 1 that enable retail users to identify themselves as uninformed. Now, those retail traders are served at 4bps, and any trader seeking onchain liquidity through other means is served at 6bps.
4. Exchange 1 will likely profit off of increased retail volume.
However, we also need to think about the stable equilibrium of order flow segmentation. In practice, it’s likely that order flow segmentation leads to a series of adjustments by market makers that ultimately increase spreads for all market participants. Coming back to our example:
5. Because Exchange 1 has implemented order flow segmentation, sophisticated traders migrate from Exchange 1 to Exchange 2 — 5bps on exchange 2 is better than 6bps on Exchange 1.
6. Now, market makers on Exchange 2 deal with an increased concentration of sophisticated traders, and widen their spreads as a result.
7. Observing this shift, market makers on Exchange 1, initially operating at narrower spreads for retail traders, begin to increase their spreads, not due to a change in the risk profile of their traders, but just because they can from a competitive standpoint.
The end result is a market where spreads are generally higher for all participants, contradicting the initial expectation of efficiency gains through segmentation.
It’s interesting to consider whether AMMs could make credible commitments to servicing flow from smart contracts at unchanging spreads over time, potentially overcoming the competitive dynamic above. It’s also possible that identity projects or third party data providers issuing credentials to wallets could be an alternative to relying on different smart contracts that delineate different order types or order size.
However, much more work needs to be done to evaluate whether any implementation of a model for order flow segmentation could ever be resistant to sophisticated traders “tricking the system” by masking their use of smart contracts, trade size, and wallet credentials.
If bidders in the OFA do not take on liquidity risk, it’s fair to characterize them as taking a fraction of every $1 of MEV created by the user <> dapp relationship. For this reason, these businesses could be characterized by potential races to the bottom and the risk of vertical integration by the order flow originators they partner with. Products having already aggregated demand (order flow) and supply (market makers), like UniswapX, have gotten over the cold-start problem and may end up being the service providers to external dapps long-term beyond hooks and V4.
However, even if the bidders in an OFA take on no liquidity risk, OFAs could be characterized by potential network effects — each additional dapp participating in an OFA should create additional revenue per dapp through higher bids from searchers wanting to capture MEV from flows across dapps. Moreover, there are ways in which OFAs can develop moats by overcoming the cold start problem that creates a barrier to entry for new OFAs in the market. We can see this by going through an example:
Consider a new OFA attempting to enter the market. Because they’re new, their customer base and order flow will likely be insufficient to sustain searcher interest and activity. As a result, competition between searchers in the auction suffers, as do revenues distributed to dapps.
Alternatively, suppose a new OFA integrates with a dapp, capturing 20% of auctionable MEV in a batch through an auction because there are too few searchers for the auction to be competitive. To attract traders and solve this problem, the OFA might offer 0% fees to undercut existing OFAs. Yet, an established OFA, having already built demand, has attracted a sufficient number of solvers to ensure competitive auctions. If this OFA captures 90% of MEV in such auctions for order flow originators, even a 5% fee on bids would yield 5 times more revenue for originators than the feeless newcomer.
The competitive dynamics mentioned above are already driving OFAs to become hybrids between OFAs and exchanges. They are intentionally seeking to attract solvers who are willing to take on liquidity risk, not just extract arbitrage.
Additionally, in the search for higher profits and sustainable moats, OFAs are shifting their focus to attract order flow aggregators (i.e., dapps) over individual users. This is driven by two main factors. First, order flow from individual users sophisticated enough to use a specialized RPC isn’t that attractive to monetize. And second, targeting order flow aggregators is a more effective way of achieving the network effects and moat around the cold-start problem mentioned above.
While RPC solutions are technically compatible with dapp-centric OFAs, the fact that they add additional steps for users in requiring them to change their RPCs to a new network has limited their adoption, given that dapps tend to prioritize smooth UX over optimizing execution quality. To overcome this, emerging SDKs and APIs are leveraging both off and onchain infrastructure, such as those developed by the likes of Flood and Atlas, as well as enabling order flow aggregators to monetize their flow without requiring users to switch RPCs. Along with this, these SDKs and APIs are maintaining composability with a variety of wallet architectures that users may utilize in the process.
Order flow aggregators, including exchanges, are compelled through competition to more effectively match Parties A and B. The better they are at matching the orders from these parties, the smaller the market for capturable MEV for OFAs to compete in. The notion that solvers and the marketplaces for them provide a sustainable, value-added service presupposes that the order matching systems used by their customers need to generate MEV in the first place. That will be the case for some segment of crypto assets, but is not an assumption that can be uniformly applied across all.
In current and potentially future enshrined implementations of PBS, arbitrageurs (and the proposers that benefit from including their transactions in blocks) capture most of the LVR generated on Ethereum. As a result of OFAs’ technical progress mentioned above, the MEV landscape is likely to see a shift whereby these profits are captured upstream of proposers and programmatically distributed to dapps and wallets stakeholders as they see fit. The fact that orderflow auctions are currently happening through MEV-Boost means that the proceeds are currently going to the proposers. However, the new OFA architectures described above can instead redirect the proceeds of those auctions to the stakeholders of the underlying order flow aggregator, such as LPs.
Creating an efficient way for arbitrageurs to compete on paying LPs to arbitrage them could be a more effective way to improve the returns for LPs than to enable them to anticipate toxic flow from those arbitrageurs, and position themselves ahead of time. This is because, in attempting to rebalance their liquidity to avoid adverse selection, LPs need to outbid the arbitrageurs who are also competing for the inclusion in the block. Arbitrageurs are funding their bids out of LPs’ profits, meaning that the system managing LPs’ capital needs to bid as much as the losses they are trying to avoid.
While it’s possible to implement tools that redirect this value to order flow aggregators, it’s also possible to design exchanges to minimize LVR in the first place. By leveraging systems that are cleared at the prices bid by arbitrageurs, such as batch auctions, new onchain exchange designs could play a role in minimizing the amount of money paid to arbitrageurs and proposers to achieve price discovery, improving returns for passive liquidity providers.
Nevertheless, given the existence of offchain order books, the theoretical and empirical research points towards a reality where batch auctions, no matter how fast, will not be the means by which price discovery occurs onchain. This ultimately results in lower levels of liquidity, lower volume, and higher relative payouts to arbitrageurs relative to offchain alternatives (see more in Part 1 on batch auctions).
For this reason, it’s unclear whether or not these exchanges (despite being more trust-minimized) will compete in attracting passive LPs who would have a choice to lend their assets to market makers operating on offchain venues. If this is true, we should see the average losses to LVR decrease significantly over the next few years, which should in turn decrease profits to block builders and proposers downstream of the MEV supply chain.
What does the future of onchain order matching look like?
At a high level, most AMMs and order books today are leveraging new scaling solutions to reduce the cost and speed of submitting and canceling trades for market makers and traders. Their main differentiators are the low-level details around the mechanisms by which asset holders can contribute to a market-making strategy — either directly into a pool or by lending them to a sophisticated market maker — and the mechanisms by which they bound the losses of the respective market maker.
As covered in Part I, in attempting to make AMMs sophisticated there is always a tradeoff between keeping them transparent enough for liquidity providers and trying to make them more competitive with their offchain counterparts. While onchain order books allow for more flexibility, they suffer from an inescapable tradeoff between MafiaEV and MonarchEV that challenges their trustworthiness and competitiveness relative to offchain exchanges. Following this, it’s worth considering how onchain order books and AMMs may differentiate themselves in the future.
While onchain order books face many challenges in competing with offchain alternatives across makers and traders, the risk/reward ratio for participating in them will persist long-term, given that they will host large markets that won’t be available to engage with in offchain environments.
As we start thinking about AMMs and onchain order books through the lens of the customers they serve, as opposed to the assets they host, we reveal the drawbacks of framing AMMs simply as hosts for long-tail assets.
AMMs can offer a differentiated value proposition to order books in the long term by serving asset issuers looking to manage their portfolios in 1) an optimized way against some critical part of their product or onchain organization and 2) an automated, low-overhead way.
We could see AMMs evolve to tailor their needs to asset issuers or large holders who want to optimize their portfolios for a different preference function than the classic market maker’s desire to take on liquidity risk efficiently. Our conclusions in Part I make it clear that AMMs are not necessarily the best exchange venues. However, they can be used as non-custodial asset management platforms, enabling asset issuers to design and engage in automated portfolio management strategies that could create a plethora of permissionless, sophisticated ETFs.
Here, LPs would commit to a predefined strategy defined by a custom bonding curve that deliberately incentivizes arbitrageurs to optimize the value of a specific portfolio that, for example, optimizes asset management by targeting a specific value for volatility. Taking this approach involves deliberately modeling and framing the expectations around LVR as a cost to achieving specific outcomes for LPs, who will likely be the asset issuers themselves wanting to choose their strategy flexibly.
In the future, onchain organizations will likely want to make their strategies composable with the smart contracts related to their core products in such a way that that optimizes their portfolio according to critical onchain metrics. In turn, onchain organizations could create systems that automatically adjust their financial position according to some element of their core business.
As a byproduct, they will create liquidity for arbitrageurs to pick up, much like how ETFs get arbitraged today. Moreover, these systems can coexist and be reinforced by the existence of onchain order books, which can be tailored to market makers’ needs in taking on liquidity risk effectively, while ensuring competition among them. This is the approach to AMM development taken by the Primitive team, who have done some fantastic R&D on this subject.
Market-making is hard, and digital goods issuers need help to optimize it to the extent of gaining an edge in making direct profits from managing liquidity. In this context, AMMs represent infrastructure that expands the scope of what can be tradable on an exchange that would have otherwise traded P2P, and are poised to achieve another differentiated value proposition to onchain order books by reducing the time, cost, and complexity of market making.
As a result, a new generation of asset issuers could (and have already become) market creators. Companies issuing tokens with utility inherent to their business, such as brand loyalty points or skins for in-game characters, will increasingly leverage blockchain’s ability to frictionlessly and permissionlessly create markets around those goods. While some asset issuers will want to reduce their customer’s ability to trade their goods efficiently and actively, other digital goods issuers will want to integrate markets as a core part of the product or service offering.
For example, if an asset issuer wants to minimize their capital at risk while creating markets for their assets, an AMMs’ ability to constrain LPs to a public market-making strategy could be seen as a feature, not a bug. By attracting external LPs to commit to a specific, transparent, and tokenizable liquidity provision strategy, asset issuers could layer complimentary yield primitives to increase the per unit return for these external LPs while minimizing their capital at risk.
As highlighted in Part 1, the extent to which this is sustainable depends on the magnitude of these layered incentives and, more importantly, the market structure. Will faster, smarter market makers pick off the AMM’s stale prices? There are a couple of reasons why the market structure may not always play out such that competing trading venues exist.
First, it’s possible that in serving these particular asset issuers, other issuers and external LPs could engage with an AMM without worrying about being adversely selected from professional market makers on onchain order books to begin with. These professional market makers may not care to take on the risks of doing so for a less liquid or less volatile asset.
Another factor to consider is that asset issuers within this category could have natural monopolies on the market-making process. If the asset with liquidity in the AMM could be created for free by the asset producer, or if the asset's production came at a cost to the issuing entity or network, they may be able to monopolize the role of market-making said asset. Assuming the market maker wants to be lazy with their strategy, it’s possible to see a world where the issuing entity is much better positioned to take on the cost of market-making this asset on their balance sheet than other market makers.
The issuer would also have a lower hurdle rate because their market-making strategy would drive value to their core business. If the digital asset issuer wanted to, their comparative advantage at market making could effectively enshrine a passive AMM as the de facto exchange compared to an order book.
The asset issuers within this profile may even want to throw their assets into an offchain bonding curve that stitches together an offchain database with traditional banking rails. Whether or not these asset issuers will choose one or the other will depend on 1) the extent to which they want to access onchain liquidity and minimize the amount of money they add into the pool and 2) whether or not the driver of the asset’s utility feels the need to bootstrap trust from the public blockchain enforcing the rules of exchange.
However, the critical factor will come down to the user and developer experience. As capabilities for developers to abstract interactions with blockchains increase and onchain compliance tooling improves, onchain AMMs may likely be adopted in this market because they will provide a better experience for users and developers than stitching together internal systems with global banking.
If any of these paradigms materialize, and the pressures on applications to minimize LVR decrease, the impacts on MEV could be significant. If AMMs were adopted as vehicles for portfolio management, then LVR would be explicitly reframed and modeled as a cost desired by participants, who would pay for it in return for a specific, complex portfolio management strategy. If AMMs were adopted as lazy markets, LVR would become less likely to occur as liquidity consolidated on the AMM where the asset issuer already has a comparative advantage in making the market.
Fundamentally, blockchains settle value for complicated ecosystems of intertwined financial markets. Each of these markets comprises different types of assets and, as a result, have different requirements. Traditional finance runs on T+2 settlement (soon to be T+1) because banks, hedge funds, market makers, etc., all need to reconcile trades across their databases at the end of the day and settle up. The only way to solve this is if everyone uses the same database, which can only be done with a distributed network. This is a perfect use case for public blockchains, which could also provide settlement-driven cost improvements as well as self-custody.
As blockchains become the settlement layer for an even wider world of assets, each market will likely evolve into a specialized ecosystem, catering to the unique needs of its underlying assets.
Exchanges require both the matching of orders and settlement. While we must recognize that there is no “one-size-fits-all” order matching system for all onchain assets, today’s onchain order matching venues face existential challenges that are likely to deter market makers and takers from adopting them over offchain alternatives.
Onchain, the battle will be between systems that enable market makers to reliably update their liquidity profile to frequently changing prices (dynamic AMMs and order books), systems that create competitive auctions for market makers to do it for them (OFA-exchange hybrids), and systems optimizing returns to passive LPs by clearing trades at uniform prices closest to the prices on offchain venues (batch auctions).
However, for dynamic AMMs and order books to compete, they will have to leverage a design that overcomes the critical challenge on Ethereum today — market makers need to outbid the arbitrageurs who are also competing for inclusion in the block. Moreover, as described in Part 1, the market design around onchain order books and some classes of AMMs are converging to allow for dynamic and programmatic inventory rebalancing, so the distinction between the two is likely to become less relevant over time. Still, the challenges across them highlighted in Part 1 will be large technical hurdles to overcome. In light of this, and given the fact that trust minimization as a feature in and of itself doesn’t represent a large market, it’s a strong possibility that trust-minimized solutions will lose market share to systems that enable lending those assets to market makers on offchain venues.
On the other hand, from our explorations into AMMs’ potential future customers, we can see another class of AMMs emerging tailored to asset issuers and holders with a different preference function than efficient market making. This includes AMMs characterized by custom bonding curves explicitly designed to pay arbitrageurs to manage their portfolio according to a specific goal or return profile. AMMs will also likely unlock value in creating new markets that wouldn’t have otherwise existed by pairing a lower overhead for creating more liquid digital goods markets with an experience for users and developers superior to web2.
Much of the MEV we are working towards mitigating as an industry stems from the existence of the same assets having liquidity both onchain and offchain, along with the arbitrage incentives this creates. While the development of complex new mechanisms built at the protocol layer must be created assuming that this persists, what Part 1 and Part 2 have attempted to show is that onchain exchanges and financial products will likely specialize in capturing markets very different than those that offchain exchanges do. As a result, we will likely see the liquidity for any given asset consolidate either onchain or offchain, which should naturally reduce much of the pressure MEV places on maintaining stable, credibly neutral blockchains.
As for the application layer, we should be wary of the potential impact that adoption of advanced onchain order matching systems might have on the incentives for specialization and centralization in the name of MEV extraction, potentially harming blockchain’s long-term viability as a successful settlement layer.
Special thanks to @soumyab8, @Autoparallel, @0xjepsen, @ThogardPvP, @fulminmaxi, @tylerinternet, @katiewav, @mountainwaterpi, @willkantaros, @AshAEgan, @DannySursock, and @dberenzon for their feedback and insights.
I also want to thank @cwgoes, @0xQuintus, and @gakonst whose insightful work has been referenced in this piece.
This post is for general information purposes only. It does not constitute investment advice or a recommendation or solicitation to buy or sell any investment and should not be used in the evaluation of the merits of making any investment decision. It should not be relied upon for accounting, legal or tax advice or investment recommendations. You should consult your own advisers as to legal, business, tax, and other related matters concerning any investment or legal matters. Certain information contained in here has been obtained from third-party sources, including from portfolio companies of funds managed by Archetype. This post reflects the current opinions of the authors and is not made on behalf of Archetype or its affiliates and does not necessarily reflect the opinions of Archetype, its affiliates or individuals associated with Archetype. The opinions reflected herein are subject to change without being updated.