MEV Conservation: Why Ordering Value Cannot Be Eliminated, Only Hidden

Author: Jascha Samadi

Compiled by DeepChain TechFlow

Shenchao Overview: Jascha Samadi dismantles the industry’s biggest illusion with a physics-like proposition: MEV can never be “designed away.” Whether it’s Ethereum’s public auction, Canton’s permissioned ledger, traditional finance’s “payment for order flow,” or “last look,” the value of ordering has never disappeared—it has merely changed its hiding place. For investors and practitioners, the real question has never been “whether MEV exists,” but rather: who are you giving the power to price order sequencing to?

In physics, energy cannot be destroyed: it only changes form. Transaction ordering possesses the same property. Wherever a ledger executes transactions against a shared state of value, the execution order carries economic value, and this value is conserved in every architecture we know. It can be auctioned, hidden, destroyed, or renamed—but it does not disappear. Ethereum turned it into a public market. Permissioned networks like Canton moved it behind contracts and nondisclosure agreements. Traditional finance spent a century packaging it under names like "payment for order flow" and "last look." Hyperliquid has recently begun destroying it.

Over the past several years, we have spent considerable time working within and around the transaction supply chain—the path a transaction takes from a click in a wallet to its final placement in a block, and the market structures that have spontaneously formed around this path. This article presents the structural perspective we’ve gained from this work. It traces the conservation of ordering value under four institutional frameworks: open markets, private ledgers, regulated intermediaries, and public auctions. The truly interesting question has never been “whether there is MEV or not,” but rather whether the price of ordering is public or hidden. Historical enforcement records from traditional finance vividly illustrate what happens when this value is denied rather than priced.

This article continues our block building series, which so far has covered the basics of the block space market, as well as demand-side and supply-side dynamics in Ethereum block building since 2024.

A public, permissionless blockchain is built on a few fundamental design choices:

• Anyone can run a node and independently verify the network's history.
• Anyone can submit a transaction.
• Participation in consensus (the right to produce blocks and extend the chain) is open, earned through capital or work, not granted by a gatekeeper.
• What underlies all of this is the most critical choice: every participant replicates the same ledger. There is a single, global state visible to all participants, and each node re-executes every transaction to verify that the rules are followed.

Global shared state is what enables these systems to transcend ordinary distributed databases. Since every application exists within the same state machine, any contract can read and synchronously compose any other contract. A lending protocol can consume a price oracle it never coordinated with, accept collateral tokens issued by a third party, and be wrapped by an aggregator approved by no one—all accomplished within a single atomic transaction. Integration is an import statement, not a business development negotiation. This feature—permissionless composability atop a jointly verifiable state—is the source of most true innovations in on-chain finance: open verifiability, trustless neutrality, and markets that can be assembled at deployment speed by unrelated parties.

The same choice also brings a second, less-publicized implication: if everyone shares the same state, transactions must be applied to that state in a specific order, and since many transactions interact with the same state, the order can alter outcomes. Two transactions targeting the same liquidity pool are not interchangeable: the one sorted first receives a better price. A liquidation is available only to the first caller: the one who arrives first. An arbitrage opportunity exists for only one transaction: the one placed ahead of all others. Position in the sequence is access to economic opportunity, and scarce, valuable access cannot remain price-free for long.

Ethereum is the clearest example of what happens when this value becomes identifiable. An industrialized supply chain—comprising searchers, builders, relays, and validators (see also our introduction to Ethereum’s block space market)—has spontaneously assembled along the path between user wallets and the final block. Searchers scan public mempools and state to identify extractable opportunities: cross-pool arbitrage, liquidations, and, in predatory cases, sandwich trades around visible user orders. Builders assemble candidate blocks, packaging and ordering transactions and bundles to maximize total block value. Under the Proposer-Builder Separation (PBS) mechanism, validators no longer build blocks themselves; instead, they auction off block-building rights to the highest-bidding builder via relays. Even further upstream, order flow itself has become a tradable asset, with wallets and applications selling the right to interact with user transactions.

People often describe this mechanism as parasitic, and some parts may indeed be so. A more accurate description is: PBS is what a market for a scarce resource looks like after the resource has been recognized.

Sorting has always been valuable; Ethereum's architecture makes this value visible, contestable, and measurable.

The subsequent research agenda—order flow auctions, MEV rebates that return withdrawals to users who create transaction value, encrypted mempools, and batch auctions—does not seek to deny this phenomenon, but rather to govern it: consciously deciding who captures ordering value and how much of it flows back to users.

Another approach is to completely remove the prerequisite of a public order book. This permissioned model now supports trading volumes of tens of trillions of dollars per month.

The most important permissioned network in production today has deliberately and consistently reversed each of the above design choices—but it eliminates the ordering market, not discretion over ordering. Canton has no global state and no global transparency. Its architecture is a network of networks: each participant runs a validator that holds only the contracts in which it is a party, and data is propagated strictly on demand. This privacy model extends within individual transactions: subtransaction privacy means that even a counterparty to one branch of an atomic transaction cannot see other branches—for example, a bank on the bond branch of a settlement-versus-payment transaction does not learn the commission terms of another branch within the same atomic submission. There is no public mempool: pending transactions are never broadcast, only delivered to their stakeholders. The ordering layer is blind: synchronizer operators order encrypted acknowledgments without seeing transaction contents.

Canton's stance on MEV is almost mechanically derived from these options, built upon four mechanisms:

• No mempool available for sniper attacks: transactions are never publicly broadcast, so the raw material for front-running (visible pending order flow) does not exist at the network level.
• The sorting is blind: operators cannot front-run what they cannot see.
• No ordering market: no builder or searcher auctions, no priority fee bidding for position, no PBS supply chain: ordering is a utility function performed by identified, contracted, legally liable entities, not by anonymous actors.
• Sorting rules exist at the application layer: they are embedded directly into the smart contract logic, enabling exchange-style fair rules—such as price-time priority—to be enforced by design, without relying on validator behavior.

The scale of adoption behind this design is substantial by any standard. Broadridge’s distributed ledger repurchase platform settles approximately $8 to $9 trillion in repurchase transactions monthly on the Canton infrastructure (about $354 billion per day), making it the largest live tokenized real-world asset settlement platform to date. Publicly tracked data shows around $345 billion in representative asset value on the network, with its claimed monthly total throughput nearing $9 trillion. The alignment with the flagship use case is genuine: interbank repurchase and collateral markets are precisely the domains where privacy, permissioning, and traceable finality matter more than censorship resistance. No bank would broadcast its funding positions to a public mempool (real-time exposure of who is borrowing overnight cash is not transparency—it’s a run accelerator), and no regulator would license a critical market infrastructure where a software failure could mean irreversible losses.

We believe this divergence is likely to be long-lasting: such permissioned networks are likely to continue succeeding in institutional back-office workflows (settlement, collateral movement, post-trade processing), where participants are known, transactions are pre-negotiated, and confidentiality and recourse are requirements rather than preferences. That market is large, and Canton is winning it through its design advantages. But the immediate question is narrower: does this architecture eliminate MEV, or does it simply do something more familiar to it?

The structural observation is: There are still people placing orders.

On Canton, ordering discretion has not disappeared; it has been shifted: to sync operators on the public layer, and to individual companies on private syncs, where the vast majority of today’s trading volume is actually settled.

Discretion still exists; it is non-competitive and non-observable, constrained by confidentiality agreements and terms of service rather than eliminated by cryptography. Therefore, a permissioned network can strictly claim far less under the banner of “no MEV” than this statement suggests: there is no permissionless MEV market, and there is no public evidence that anyone is extracting MEV. These are two distinct statements, and the distance between them encompasses much of the history of market structure.

The near-zero MEV observed on these networks has a second explanation, alongside design: use case. Canton’s flagship workflows—bilateral repurchase agreements and delivery-versus-payment settlement for pre-negotiated trades—involve almost no contentious shared state. In a repurchase agreement where price, size, and counterparty are fixed before the transaction reaches the ledger, there is nothing to sandwich; sorting a pre-negotiated settlement queue generates no economic value on any chain or under any design. Thus, the current record is better described as unfalsified rather than proven.

Verifiable tests will arrive when a network supports a truly competitive market structure: a shared central limit order book, a clearing engine, oracle-triggered margin calls, and a tokenized government bond market for scaled trading. At that point, the order of events reaching the synchronizer will once again be economically decisive, and the participants able to capture it will be those with latency advantages, physical proximity, or operator relationships. Rent will not disappear—it will migrate to privileged and opaque channels: priority connections never listed on fee schedules, information proximity never disclosed anywhere. This architecture reintroduces trusted intermediaries with ordering discretion on a single ledger and rebuilds the trust model of traditional finance.

Whether this is a stable endpoint depends on the track record of this trust model on this exact issue—and that record is highly instructive.

Traditional finance has operated for decades under a model of "trusted intermediaries with discretionary ordering power"; its track record shows that opacity has not prevented the capture of ordering value—in fact, opacity is precisely what makes such capture profitable.

Traditional market structures often view what we now call MEV as a system design issue—one already resolved through proper rules, responsible intermediaries, and enforcement. A structural perspective, however, argues otherwise.

MEV is the property of valuable shared state combined with sequential execution, and every exchange possesses both. From this perspective, traditional finance has never eliminated this issue; instead, it has prohibited it through rules, regulated it imperfectly, and pushed extraction into channels invisible to victims. Enforcement records align with this structural view.

Payment for order flow has turned retail orders themselves into a monetizable asset. In just 2021, U.S. brokers collected approximately $3.8 billion in payment for order flow—capturing the value of seeing and internalizing order flow through bilateral agreements between brokers and wholesalers, agreements invisible to customers whose orders are treated as products. Robinhood paid a $65 million SEC settlement in 2020 for misrepresenting the execution quality associated with this arrangement. Dark pools, marketed as protecting institutions from predatory order flow, have generated a series of相反 settlements: Barclays paid $70 million in 2016, Credit Suisse paid $84.5 million, for misrepresenting how their dark pools operated, including that high-frequency order flow customers were told had been excluded actually participated in trades; and ITG paid $20.3 million in 2015 for operating an internal desk within its own dark pool that traded using confidential customer order information. In the foreign exchange market, Barclays paid $150 million in 2015 for using “last look” (a dealer’s option to refuse a trade after observing price movement direction) and coordinating the sharing of customer order information prior to benchmark fixes; cumulative fines for such practices across the industry have exceeded $10 billion.

The pattern in these cases is consistent. The value of ordering is real. The extraction behavior is carried out by a trusted intermediary within the sequence. Affected parties cannot observe it in real time. The truth only emerges years later, through whistleblowers and law enforcement actions. This is the operational model of “MEV-by-design” introduced onto the ledger: ordering rents are determined by compliance departments and discovered by regulators, rather than priced by an open market.

Transparent alternatives are not hypothetical: Hyperliquid takes the same structural rents, puts them into an open auction, and directs the proceeds to a public pool.

Hyperliquid’s architecture (with no public mempool and ordering handled at the consensus layer) has long been described as MEV-resistant, and by standard mempool-based extraction criteria, it indeed is. However, execution advantages still exist, as they are structurally inherent: priority always goes to those who engineer the lowest latency. Established market makers win queue positions through infrastructure optimization and network proximity, functionally equivalent to colocation. This advantage has long existed, carries significant economic implications, and is accessible only to participants with the capital and channels to build it. Rent is being captured—it’s just not visible on anyone’s dashboard.

In April of this year, Hyperliquid formalized this dynamic by introducing priority fees: public, recurring Dutch auctions priced in $HYPE, auctioning off the two key advantages of low-latency execution: earlier access to incoming trade data and a higher position in the execution queue. The effect can be measured with exceptional precision: each basis point paid in priority fees improves end-to-end latency by approximately 45 milliseconds. What was previously an implicit, private advantage limited by engineering capability has now become an explicit, public, continuously re-priced market open to all participants.

This mechanism has two attributes worth discussing separately. First, it formalizes a paid priority tier, which can indeed be criticized: a visible queue-position tax now exists where none existed before. However, the relevant comparison is not a market without priority tiers, but rather the same tier previously allocated invisibly through capital expenditures and connections at a price of zero.

A transparent and priced advantage is more honest and easier to correct than an implicit and free one.

A visible market can be measured, debated, bounded, and redesigned, while an invisible market can only be discovered after the fact.

The second attribute is the destination of the revenue—the aspect we consider truly novel in market structure. These fees are burned. The execution priority price is not paid to the platform operator, not shared with privileged counterparties, nor embedded in any bilateral agreement; it is removed from supply (economically equivalent to a proportional distribution to every holder of the network asset $HYPE). The contrast with how equivalent value flows in traditional markets is straightforward: colocation fees go to exchange shareholders; payment for order flow is allocated by wholesalers and brokers under contracts unseen by end clients; informational rents from dark pools and last-look rights accrue to those with privileged access, until enforcement recovers some portion. Hyperliquid exposes this same structural rent—the value of position in the sequence—to the open, enabling a permissionless auction to discover its price and directing the proceeds to a public pool.

The queue still exists; what has changed is that its price is public, and its economic benefits are shared with every holder, rather than settled within a closed bilateral structure.

This is not claiming that ranking value has been designed out of existence. It is acknowledging that it cannot be eliminated, and then deciding to price it on open markets and reallocate the proceeds.

Meaningful comparisons between chain designs have never been about "with MEV versus without MEV"; any place that orders valuable state transitions inherently generates ordering value.

The issue that truly distinguishes market design is narrower and easier to answer: who holds the ordering power; whether it is contestable or assigned; whether it is observable or hidden; and whether its rents are priced and redistributed in open markets, or quietly captured by those closest to the sequencer.

Permissioned networks address these questions through governance: clear operators, legal accountability, and institutional guarantees. For pre-negotiated settlement processes, where ordering carries no value from the outset, this answer is sufficient—and the adoption data reflects this. For truly competitive markets, evidence from both the brief history of crypto and the long history of traditional finance points in the same direction: denied ordering value is often captured in the shadows, while acknowledged ordering value can be priced, audited, and returned to the market that created it. In this sense, ordering value is conserved across designs; the difference lies in whether its price is public.