Where Is MEV Heading? The Future of DeFi From the Trenches

Why This Question Won't Leave Me Alone

I spend most of my days knee-deep in transaction bytes, custom error codes, and the kind of low-level Solana plumbing that makes even seasoned developers reach for another cup of coffee. But when you stare at the same problem long enough, the bigger questions start tapping you on the shoulder. What is MEV actually for? Who benefits? Who pays? And where is all of this going?

The Solana MEV journal I've been keeping was supposed to be a workbench log — a Tuesday-night kind of project. Yet every infrastructure paper that lands in my feed, every prosecution that hits the wires, every protocol announcement nudges me to ask the same thing: am I building inside a market that's maturing, or inside one that's about to be re-engineered out from under me? This piece is my attempt to answer that — not as a forecast, but as a present-tense snapshot of a debate that, by mid-2026, feels impossible to ignore.

What MEV Has Quietly Become

MEV — maximal extractable value — is the Ethereum Foundation's way of describing "the maximum value that can be extracted from block production in excess of the standard block reward and gas fees by including, excluding, and changing the order of transactions in a block," per ethereum.org. Translated for the rest of America: it's the spread between what a transaction is worth and what the block producer can pocket by choosing how, when, and whether to include it.

That sounds small until you look at the dollar signs. A May 2025 paper from Jake Rubin and colleagues totals "over $1.3 billion already extracted via MEV-related strategies" across the public chains. The European regulator ESMA, in its July 2025 TRV Risk Analysis, reports that 2025 MEV transaction volume on Ethereum alone reached roughly $561.92 million, with sandwich attacks accounting for $289.76 million of that — about 51.56%.

Meanwhile, on Solana, Jito CEO Lucas Bruder's 2025 retrospective reports tips generated through the Jito stack jumped from $3.52 million in 2023 to $674 million in 2024. Of that 2024 figure, $550 million flowed to stakers. Bundle volume processed through Jito grew from roughly 16.1 million in December 2023 to about 590 million in December 2024 — a 36x jump in twelve months. Whatever MEV is, it is no longer fringe.

This isn't the side hustle of a few quants in a Brooklyn loft. It's a parallel financial infrastructure, and the question of what it should be allowed to do is now squarely on the table.

The Ethics Question Nobody Can Quite Settle

The most uncomfortable thing about MEV is that it isn't one thing. It's a family of activities, and the family contains both the auto mechanic and the pickpocket.

On one end, you have arbitrage and backrunning. Rubin et al. estimate arbitrage at roughly 60% of MEV activity, and a16z's primer describes backrunning as the kind of trade that contributes to "keeping AMMs' prices up to date." If a Uniswap pool and a Binance order book briefly disagree on the price of ETH, somebody has to close that gap. The bot that closes it earns a margin, but the side effect — accurate prices, working liquidations, solvent lending markets — is what everyone else relies on. It's the financial-plumbing equivalent of the guy who comes through with a snowplow at 4 a.m. so the rest of town can get to work.

On the other end, you have sandwich attacks. The a16z piece describes the choreography clearly: an attacker buys before a victim's purchase to push the price up, then sells afterward at the inflated rate, leaving the victim holding the bag. Rubin and his co-authors write that, from the victim's point of view, this is "indistinguishable from being scammed." That language matters. When the experience on the receiving end is identical to fraud, it stops being a sterile technical debate.

The specifics get rough. On March 23, 2025, the Rubin paper documents a Raydium sandwich attack where a user bought 100 SOL worth of a token — about $13,100 — with unlimited slippage. A bot pocketed 95 SOL from the single trade, and the user took roughly a 95% loss against the token's market price. A Shutter Network op-ed by Loring Harkness flags an even larger single-trader loss of about $800,000 in January 2025, and points at the prosecuted Peraire-Buneo case in which two brothers allegedly extracted roughly $25 million by exploiting an MEV-Boost software vulnerability.

The math splits cleanly. The Rubin paper estimates that the harmful tail — frontrunning and sandwiching — accounts for somewhere between 10% and 15% of MEV. The GoGol Substack survey lands in the same neighborhood, with about 60% arbitrage, 30% oracle-extractable value (liquidations, price feed updates), and 10–15% in the explicitly harmful category. So most MEV is something like the boring back-office of DeFi. But the 10–15% sliver is what shows up at the dinner table when a retail user asks why their swap on a meme coin cost them everything.

From where I sit, that split is the whole story. A regulator looking at the harmful 10–15% sees fraud. A market-maker looking at the helpful 60% sees plumbing. Both are correct. The hard part is designing a system that lets the second exist without enabling the first.

The Infrastructure Arms Race

The infrastructure layer is where the abstract debate gets concrete, and it's moving fast.

On Ethereum, the dominant pattern is Proposer-Builder Separation (PBS), implemented in practice via MEV-Boost. The Flashbots timeline traces the journey: MEV-geth open-sourced in November 2020, Flashbots Alpha and Relay in January 2021, and by April 2021 "mining pools representing over 84% of Ethereum hashrate" had adopted MEV-geth. MEV-Boost for proof-of-stake shipped in November 2021 and activated 17 epochs after the Merge in September 2022. By mid-2025, aggregated industry reporting cited in various analyst primers puts roughly 90% of Ethereum blocks as being built via MEV-Boost.

That success comes with a footnote in red ink. Bruder cites the now-familiar statistic that on Ethereum, two builders construct roughly 96% of blocks. The 2024 Flashbots data he references suggests the top five searcher entities capture around 80% of MEV. PBS was supposed to be a decentralizing move, separating block-builder from block-proposer. The current state looks more like Walmart and Target dividing the shelves — efficient, but not exactly the small-town main street the white papers promised.

Flashbots has responded by spinning down their own central builder. The Flashbots site records that in November 2024, BuilderNet had its first release, jointly operated with Beaverbuild and Nethermind. In December 2024, Flashbots "migrated all builders, orderflow, and refunds to BuilderNet" and stopped operating any centralized block builders on Ethereum. BuilderNet v1.2 followed in February 2025. The Flashbots Protect product, as of October 2024, reported serving 2.1 million unique Ethereum accounts, shielding $43 billion in DEX volume, paying out 313 ETH in MEV refunds, and handling more than 30 million daily requests.

On Solana, the story rhymes but doesn't repeat. Bruder's piece reports that the Jito-Solana validator client went from 48% of network stake at the start of 2024 to 94% by year-end. He describes Jito as moving away from "short-term solutions like privileged transaction lanes" toward enhanced privacy primitives and more efficient transaction ordering, with the Block Assembly Marketplace (BAM) launching in July 2025 — nodes running inside Trusted Execution Environments. The peak numbers he cites for Solana are American Black Friday-scale: $10 billion in daily trading volume and $27.6 million in daily fees during the high points of 2024.

The pattern is the same on both chains. A first generation of MEV infrastructure solved the worst spam-and-uncertainty problems and concentrated power in a handful of operators. A second generation, now arriving, is trying to undo that concentration without giving up the efficiency gains. Whether it works will define the next several years of DeFi.

"It's Not Physics" — The Philosophical Fight

There is a real, ongoing argument about whether MEV is gravity or whether it's a bug.

The inevitability camp argues that any transparent ordering system on a public ledger creates extractable value. Once you can see pending transactions and reorder them, somebody will pay to do exactly that. The a16z framing supports this: MEV is "fundamental to how blocks get built on nearly all blockchains." From this angle, fighting MEV is like fighting friction. You can reduce it, channel it, redistribute its proceeds — but you can't make it zero.

The anti-inevitability camp pushes back hard. A Medium piece by pmcgoohan puts it bluntly: "MEV is not a fundamental law of the universe, it is the exploitation of a network vulnerability that is our responsibility as developers to fix." CoinTelegraph's framing — "MEV is a choice blockchains must confront" — leans the same way, casting MEV-enabling tactics as "deliberate manipulations that subvert user intent." The Shutter Network position, predictably given their commercial focus on encrypted mempools, is similar: there is "nothing beneficial about a system that forces users to either accept systematic exploitation or pay additional fees to intermediaries."

The middle ground is where most working engineers seem to live. The GoGol survey summarizes one of its five myths as "MEV can be fully eliminated," then patiently explains that it can't — but it can be mitigated through MEV-Boost-style relays, private order flow, and fair sequencing. That sounds about right. The honest answer is that you can't get to zero without giving up properties of the chain that people value, but you can choose where to set the dial.

From the cockpit of my own bot, I notice that the philosophical position you take changes which problems you find worth solving. If MEV is gravity, you build better aerodynamics. If MEV is a bug, you patch it. Most teams hedge: they ship aerodynamics today and contribute to the bug bounty on the side.

The Regulators Walk In

For most of MEV's history, the response from Washington and Brussels has been a polite silence. That's no longer true.

The ESMA report from July 2025 is, to my knowledge, the first formal regulatory analysis of MEV from a major jurisdiction. ESMA places the discussion under the EU's MiCA (Markets in Crypto-Assets) framework, which requires entities to adopt systems to detect market abuse and tasks ESMA itself with developing regulatory strategies for MEV. The headline numbers in that report — $561.92 million in 2025 Ethereum MEV volume, with sandwich attacks at $289.76 million or 51.56% — are no longer just industry trivia. They are the dataset a regulator will reach for.

In the United States, the Shutter Network piece cites the prosecution of the Peraire-Buneo brothers, who allegedly extracted roughly $25 million by exploiting an MEV-Boost vulnerability. The Rubin paper notes that as of April 2025, "there is currently no legal consensus, but if the Trump Administration and the SEC conclude that sandwiching/front-running constitutes market manipulation, MEV bots and searchers may face legal repercussions." Translation: the SEC has not yet declared sandwiching to be market manipulation, but the door is open, and the precedent from prosecuted cases shows that at least the worst exploits will land squarely inside the criminal code.

What I find interesting about this moment is how poorly the existing frameworks fit. Securities law was designed for human market makers and identifiable counterparties. The MEV ecosystem is a global tangle of pseudonymous searchers, builders, relays, and validators. A bot in Berlin can sandwich a user in Boston via a builder operated by a Luxembourg entity, settled on a chain whose validator set spans every continent. Pretending that the Securities Act of 1934 is going to neatly classify this is a stretch. So the most likely outcome is something gradual — a few high-profile prosecutions, some MiCA-style market-abuse rules, and a slow drift in which infrastructure operators take on more compliance-style duties whether they want to or not.

What I'm Watching From The Cockpit

A few threads in the research stand out as the ones I keep coming back to, the way you keep checking your dashboard on a long drive.

First, redistribution mechanisms. The Rubin paper points to CoW Swap returning roughly 90% of MEV profits to users through their order flow auction. The a16z list of mitigations explicitly highlights order flow auctions (OFAs) that redistribute MEV back to users through bidding. A May 2025 arXiv benchmark of private MEV protection RPCs puts the share of Ethereum transactions now flowing through private RPCs at around 80% — and many of those private routes feed OFA-style auctions. If even a meaningful fraction of MEV profit gets rebated back to the end user instead of flowing entirely to searchers and builders, the political economy of the system changes overnight.

Second, encrypted mempools and trusted execution. The a16z mitigation list mentions "encrypted mempools with random transaction shuffling using randomness beacons." Shutter Network is one obvious example. Jito's BAM marketplace runs nodes inside TEEs. The Rubin paper notes that protected order flow with TEEs is one of the more credible cryptographic mitigations available today. The basic idea is the same in each case: if the next transaction's content is hidden from the orderer until the moment of inclusion, the sandwich-attack family of strategies stops working at the technical level — not just at the policy level.

Third, the protocol-asset path. Some Ethereum researchers want to internalize MEV as a protocol asset — execution tickets, attester-proposer separation (APS), inclusion lists treated as "one-way valves." The a16z piece lists single-slot finality among its mitigation toolkit (which prevents subsequent proposers from forking to claim MEV). This is the deepest kind of fix because it doesn't try to outwit the extractors — it changes the rules of the game so the extraction surface is smaller to begin with.

Fourth, the on-chain footprint of MEV revenue. Bruder's claim that Jito tips reached about 50% of Solana network revenue in 2024 is a number you can't unsee. If half of what a chain earns is MEV-related, then MEV isn't a side feature, it's a core economic pillar. The same incentives that make MEV profitable are also funding validators. Any cleanup that undermines MEV economics will need to deal with the budget hole it leaves behind.

Fifth, the developer side. The Solana fee level — Rubin et al. quote typical fees "less than 10 cents," against Ethereum gas during peaks running into "hundreds of dollars" — keeps reshaping where the action lives. A chain where a failed attempt costs the price of a pack of gum is a chain where small operators can iterate. That's a structurally different environment from one where each test is the cost of a steak dinner. As long as that gap persists, MEV innovation will keep pushing toward cheaper execution layers.

And sixth, the question I can't shake: who is this market actually for? The Ethereum Foundation's documentation notes that searchers "may pay 90% or more of their total MEV revenue in gas fees to validators" for highly competitive opportunities. The arXiv survey reports roughly 440,000 ETH extracted in MEV pre-Merge and about 180,000 ETH in the eight months after. If most of the value flows from users to searchers to validators, with a small middleman cut for builders and relays, then the redistributive question becomes the political question. And political questions, in my limited experience, don't stay quiet for very long.

What This Means Going Forward

If I had to compress the picture into a thesis, it would be this: MEV is in the awkward adolescence between "unexamined market behavior" and "regulated infrastructure." The data is now public enough — ESMA's tables, Bruder's transparency, Flashbots' open product timeline — that the era of "nobody really knows the numbers" is over. The next era will be defined by which mitigations the ecosystem chooses to ship at scale, and which interventions the regulators choose to bring.

The most likely path forward, in my read, is messy. Encrypted mempools will keep growing share in segments where users actively opt in. OFAs and protected order flow will keep clawing back chunks of MEV for users. PBS and BuilderNet-style decentralized building will keep nudging the builder market away from its current top-two concentration. Regulators in the EU will use MiCA to standardize disclosure and abuse-detection requirements, while the U.S. picks off the most egregious exploiters one prosecution at a time. None of this is the dramatic single-strike fix that the inevitability skeptics call for. But layered, it might be enough to make the harmful sliver shrink and the helpful core keep functioning.

What I don't think will happen — and this is the part I'm least certain about — is the wholesale disappearance of MEV. The math is too good, the incentives are too aligned, and the infrastructure is too entrenched. A future without MEV would require either a fundamentally different ordering model on every major chain or a regulator with the appetite to ban a class of behavior whose harmful share is real but minority. Neither feels like the base case from where I'm standing.

Key Takeaways

• MEV is no longer fringe. Cumulative extraction has crossed $1.3 billion per Rubin et al. (May 2025), and per the ESMA July 2025 report, 2025 Ethereum MEV transaction volume alone reached about $561.92 million.
• The ethics question splits the activity into halves. Around 60% of MEV is arbitrage (market plumbing), per Rubin et al. and the GoGol survey; the 10–15% sliver of sandwiching and frontrunning is what drives the headlines and the prosecutions.
• Infrastructure is consolidating, then trying to de-consolidate. Per Bruder, two builders construct roughly 96% of Ethereum blocks; Flashbots' BuilderNet (live since November 2024) and Jito's BAM (July 2025) are explicit attempts to walk that back.
• Regulators have arrived. ESMA's July 2025 report, MiCA's market-abuse hooks, and the prosecuted Peraire-Buneo $25 million MEV-Boost case (per the Shutter Network op-ed) mark the start of an actual regulatory perimeter.
• The future is layered mitigation, not a silver bullet. Encrypted mempools, OFAs, TEEs, and protocol-asset designs (per a16z and the Rubin paper) are each chipping at different parts of the problem. The combination, not any single one, is what will define the next chapter.