MEV Does Not Disappear. The Question Is Where It Goes.

1. The question has changed

For the first years of DeFi, the conversation around MEV was mostly about prevention. Encrypted mempools, private transaction channels, commit-reveal schemes, mempool obfuscation. The goal was to stop extraction from happening.

The intuition was understandable. Front-running and sandwich attacks look like attacks. If you can hide the transaction, the attack cannot happen. If you can randomize ordering, the attacker loses the edge. Prevention seemed like the right frame.

It was the wrong frame.

MEV is not an attack on a system with a clean alternative. It is a structural property of any system that executes ordered transactions competitively. When multiple parties compete to interact with shared o

n-chain state, some orderings are worth more than others. That value exists whether or not anyone extracts it.

Preventing one actor from capturing it does not make it disappear. It shifts it to whoever can reach it next, or it sits unrealized, which is a different kind of waste.

This is why encrypted mempools close one layer of MEV and leave three others intact. It is why every mitigation the field has deployed has been followed by adaptation from the extraction side. The game does not end. The players adjust.

The protocols that understand this have moved past the prevention frame entirely. The question they are asking now is not "how do we stop MEV?" It is "who captures this value, and can we design our protocol so that it stays inside our ecosystem instead of flowing to external operators?"

That question has two answers. The first is internalization: the protocol captures the value itself. The second is redistribution: the value is returned to the users who generated it.

Every DeFi transaction creates MEV. A swap moves the price. A large order creates a sandwich opportunity. That value originates from user activity. Returning it to users is a legitimate design goal, not just a secondary option.

That shift, from trying to eliminate MEV to deciding where it goes, is the frontier of protocol design today.

2. How much MEV are protocols losing

MEV loss rarely appears as a line item. It shows up as LP underperformance, as worse user execution prices, as liquidation bonuses that leave the ecosystem. The sources are distinct and each requires a different response.

Liquidity providers lose through loss-versus-rebalancing (LVR): every time an arbitrageur corrects an on-chain price against an external reference, the surplus goes to whoever has faster infrastructure and CEX access. LPs are the counterparty. Estimated cost: 5-7% of capital deployed annually, with no redistribution mechanism in most AMMs.

Lending protocols lose through liquidation MEV. The liquidation bonus was designed as an incentive for healthy participants. In practice, most of it flows to bots competing in priority gas auctions. Two mechanisms address this directly: Chainlink SVR wraps price feed updates in an auction, redirecting proceeds to the protocol; Oval does the same as a protocol-agnostic wrapper around any oracle. Together they have identified over $500 million in cumulative oracle extractable value across DeFi. Aave's integration of Chainlink SVR has recaptured over $1.1 million in its first months, with a capture rate approaching 90%.

Users lose through sandwich attacks on their own transactions. MEV Blocker returned 4,079 ETH to users in a single year, value generated by user activity that would otherwise have gone entirely to searchers.

A note on Substrate-based chains. Chains built on Substrate have a structural advantage that EVM chains do not: direct control over block construction at the protocol level. A Substrate chain can execute specific logic before any user transaction at each block: liquidating undercollateralized positions, rebalancing internal pools, correcting price discrepancies between its own AMMs.

This is not an external bot racing to capture the opportunity. It is the chain itself acting as the first mover, capturing value that can then be redistributed to users, stakers, or the protocol treasury. It is one of the most underexplored advantages of building on this stack.

3. Four ways to keep the value

There is no single solution. The right mechanism depends on the protocol type, the source of MEV, and how the captured value should be distributed. The landscape organizes into four categories, each operating at a different layer.

Ordering and sequencing Change who controls transaction ordering and under what rules.

Order flow auctions (OFAs) auction the right to backrun user transactions, returning proceeds to users or the protocol. Batch auctions settle all orders in an interval at a uniform price, making transaction order irrelevant within the batch. Intent-based systems let users express desired outcomes rather than raw transactions, shifting MEV burden to competing solvers. Application-specific sequencers give protocols direct control over ordering within their domain.

A distinct variant: TWAMM (Time-Weighted AMM) breaks large orders into infinitely small pieces executed continuously across many blocks, making large trades significantly harder to sandwich. It does not internalize MEV directly, it reduces the surface available for extraction. Most relevant for protocols handling large programmatic orders: treasuries, DAOs, rebalancing mechanisms.

Fee and pricing Change the economics of extraction without redesigning the user interface. Dynamic fee mechanisms spike swap fees when aggressive directional trades arrive, capturing value for LPs instead of letting it leak to arbitrageurs. Priority fee taxes (Balancer v3) charge a 99% tax on searcher priority fees, routing approximately 90% of competitive MEV back to liquidity providers.

Information hiding Hide transaction intent until inclusion is certain. Commit-reveal schemes separate commitment from execution across two transactions. Threshold-encrypted mempools (Shutter Network, live on Gnosis Chain) encrypt transaction payloads until a distributed set of keypers decrypts them post-inclusion. These approaches close reactive MEV. They do not address structural MEV or LVR, which require no mempool visibility.

Oracle-specific capture For protocols where MEV is tied to oracle price updates rather than on-chain order flow. When an oracle update triggers liquidation eligibility, bots race to act on that window. Chainlink SVR and Oval both wrap this process in an auction: the right to backrun the update is sold, with proceeds redirected to the protocol instead of external searchers. Relevant for any protocol with oracle-triggered mechanisms: lending, derivatives, structured products.

4. Why design is the hard part

Knowing the toolkit exists is not enough. Each mechanism creates a new equilibrium in the ecosystem of bots, LPs, and users. Getting that equilibrium wrong can be worse than doing nothing.

An OFA with low auction participation becomes a monopoly. If only one or two searchers bid, the winner captures nearly all the value. The protocol has added complexity without improving outcomes.

A priority fee tax calibrated too aggressively disincentivizes legitimate arbitrage alongside predatory extraction. AMMs need arbitrage to maintain price efficiency. Tax all of it equally and LP returns are damaged from a different direction.

An intent-based system with concentrated solver supply creates censorship risk and execution quality problems. The mechanism is only as good as the competition it generates.

The mechanism is not the answer. The mechanism correctly calibrated for your specific context is.

That requires modeling how the market will respond, not just implementing a smart contract. MEV internalization is a game-theoretic design problem as much as an engineering one. The technical part is often the easier half.

Closing

MEV does not disappear. The value that searchers extract from your protocol exists because your protocol exists. The question is only where it goes.

Protocols that do not design for this are already answering the question: they are giving it away. The tools to change that answer exist and are increasingly mature. The hard part is choosing and calibrating the right one for a specific context.

If you are building a DeFi protocol and thinking about this problem, reach out.