I pulled the relay dashboards at 06:00 Melbourne time. The pattern had barely shifted in months. Two visible relays carrying the majority of public blocks. A small cluster of builders producing the bulk of supply. An unknown slice routed through private channels. The concentration data is public. The response is silence.
Here is what the data says: the neutral block producer is gone. The MEV supply chain — searchers, builders, relays, validators — has consolidated into a structure that functions like a cartel. Private auctions. Hidden relays. Validator collusion. Three mechanisms. One outcome: the actors who control transaction ordering capture the rent, and everyone else pays the difference.
I didn't need a leaked memo to see this. The blocks showed me. The market, busy pricing the next narrative, hasn't looked.
MEV — maximal extractable value — is the profit embedded in transaction ordering. Arbitrage. Liquidations. Sandwich attacks. For most of DeFi's history, it was a bot problem: fragmented, opportunistic, individually small. DeFi Summer 2020 turned it into an industry. I traced one $4.2 million arbitrage flow on Compound that year, reconstructing the transaction sequence with Python scripts and Etherscan logs. The flash loan mechanics were clean. The settlement logic was clean. The ordering — who got to trade first, and why — was a coordination game wearing a market costume.
This should be the moment for hard questions. The market is in a bull phase, and bull phases are terrible auditors. Capital is flowing into AI x Crypto narratives, infrastructure tokens, and any project that can spell "decentralized" in a whitepaper. The inconvenient data — the concentration metrics, the hidden relays, the colluding validators — gets filed under "known issue, fix later." Later is now.
The industry's proposed cure was Proposer-Builder Separation (PBS). The theory was sound: separate block production from block proposal. Force builders to compete for the right to propose. Let validators pick the highest bid without seeing the contents. Neutrality by construction. No single actor controls what enters a block.
The reality that shipped was the architecture with a broker layer bolted on top. Builders compete, but the competition is gated. Private auctions admit only vetted searchers. Hidden relays forward blocks with no disclosed selection logic. Validators coordinate to extract ordering rent collectively rather than remain neutral.
The external shell still looks decentralized. Thousands of validators sign every slot. But economic power — who gets included, who gets front-run, who pays the slippage — sits inside a black box. This is what the phrase "MEV cartel" describes. Not a protocol. Not an exploit. A systemic market structure.
Let me walk the supply chain step by step, because the failure modes only make sense in sequence.
Step one: a searcher detects an arbitrage opportunity in the mempool. Step two: rather than broadcasting the transaction publicly, the searcher sends it to a private auction — a venue where only selected participants can bid for exclusive access. Step three: a builder assembles blocks from these private orders, bundles them, and submits the bundle to a relay. Step four: the relay — which may publish no rules — forwards the block to validators it trusts. Step five: validators sign and propose. The user's original transaction is included, eventually, at a worse price. The entire sequence is valid. Nothing on-chain is false. The extraction happens entirely in the ordering.
Audit each component individually and it passes. The smart contracts are sound. The relay code is functional. The validators are signing valid blocks. The failure is emergent — it lives in the interaction between components, not inside any single one. This is the difference between a bug and a systemic risk. A bug has a patch. A systemic risk has a political economy.
Now the failures.
Failure one: private auctions broke the neutrality assumption.
Block space was supposed to be an open market. Anyone can bid. Best price wins. In practice, the most profitable order flow never touches the public mempool. Searchers with privileged channels see pending transactions first. They pay for that access, and the payment is the point: the auction exists, but the bidder pool is closed.
This is where most commentary stops, labeling it "centralization risk." To stop there is to misread the engineering. The consensus layer is fine. The chain settles on time. The bottleneck wasn't throughput. It was accountability. When order flow is private, no one can verify whether the winning bundle was the best bid or the best-connected bid. That unverifiable gap is the rent.
Flash loans don't create this problem. They industrialize it. Before flash loans, arbitrage was capital-bound; you needed inventory to execute. After flash loans, anyone with a clever contract can borrow the entire pool, front-run the victim, and repay atomically. Efficiency improved. So did extraction. The Compound case I traced in 2020 demonstrated it perfectly: a loan materialized from a single call, a sequence of swaps executed ahead of the victim, a clean profit, a valid transaction log. Nothing violated the protocol's rules. Nothing needed to. The ordering was the attack surface.
Failure two: hidden relays broke accountability.
Relays sit between builders and validators. They receive block proposals, validate them, and forward the winner. A public relay publishes its rules. You can inspect its accept/reject behavior and its decision logic. A hidden relay discloses none of that. It is a black box with veto power.
Operationally, that means a relay can censor a transaction without explanation. It can privilege one builder over another. It can skim a percentage of the block's surplus and call it an infrastructure fee. You cannot audit it. You cannot fork its relationships. You can only trust it.
The relay's only weakness is its fear of being traced. On-chain, everything leaves footprints eventually. But the relay's selection logic executes off-chain. It does not need to explain itself. The traceability gap is the product.
Failure three: validator collusion broke governance.
Validators are supposed to be independent guardians of protocol rules. Collusion changes the calculus. A lone validator captures only the MEV flowing through its own blocks — a fraction of the rent. A collaborating group shares order flow, coordinates block allocation, and splits the proceeds. The per-member cost of control drops as the group grows. Defection becomes irrational because defectors lose access to the shared order flow.
This is what "decentralization is a pay-to-play game" means, concretely. The consensus layer is distributed. The economic layer is not. Validators still sign blocks — but they sign blocks they didn't build, containing orders they didn't see, priced in a market they don't disclose.
From a systems perspective, this is a principal-agent failure. The protocol delegates block production to validators. Validators delegate it to the cartel. The user's interest — fair, neutral inclusion — is the one interest no actor in the chain is incentivized to protect.
The downstream is an invisible tax.
Every AMM swap pays it. Slippage, front-running, failed transactions. The user sees a worse fill. The cartel sees revenue. No invoice is issued. Quantify it and the scale becomes uncomfortable. Every block with private order flow carries hidden costs absorbed by liquidity providers and retail traders. The mechanism is mundane: a user submits a swap, a searcher spots it, a bundle places the user's transaction at the exact wrong position. The user's slippage tolerance is the searcher's revenue. On a busy AMM, this repeats thousands of times per hour. Individuals lose cents. The aggregate is millions.
This is why I now include a rough MEV concentration score in protocol assessments — an estimate of how much ordering value flows through opaque channels. It is a crude metric. It is also a surprisingly strong predictor of where the risk hides.
The transmission is broad. Wallets and aggregators feel pressure to integrate MEV protection simply to retain users. DEXs competing with protected aggregators lose flow. Infrastructure providers carry a reputational tax for touching hidden relays. The entire user-facing stack is being forced to compensate for a failure at the production layer.
The regulatory framing writes itself. In traditional markets, front-running is market abuse. On-chain, it is "maximal extractable value." Same activity, different vocabulary. If a group of validators coordinates to extract value from user orders, the behavior pattern matches market manipulation and antitrust theories. The CFTC has already signaled interest in crypto market conduct. Relays, builders, and colluding validators sit squarely in that crosshair. The statute hasn't caught up, and that gap creates its own risk. Either the behavior remains unregulated and extraction continues, or it gets regulated retroactively and the exposure lands on the most visible actors — the relays and builders who left a paper trail.
I have seen this movie before. In 2022, dissecting the Wormhole bridge hack, I found the Guardian Network's multi-sig threshold looked adequate on paper. The flaw was deeper: the model assumed validators would act independently. They did not. The MEV cartel rests on the same broken assumption. Complexity, in both cases, was a cover for insecurity — more moving parts, more places to hide.
So what do you watch?
Three signals. Builder concentration: if the top three builders produce more than half the blocks, the auction is a formality. Private relay share: the percentage of blocks routed through hidden channels tells you exactly how much of the market you are not allowed to see. Governance pressure: watch what happens when inclusion lists or enshrined PBS move from research to a formal EIP. The cartel's response will tell you how exposed it feels. And user migration: if MEV-protected interfaces grow, the market is voting with its feet.
The bulls get something right, and it deserves a fair hearing: the cartel is unstable.
Rent attracts competitors. That is the one law of markets that still works reliably on-chain. Every period the cartel extracts more, the incentive to build a transparent alternative grows. The margin embedded in block space is the investment thesis for a transparent alternative.
The anti-MEV narrative also gets something wrong. It implies transparency alone solves the problem. It does not. An open auction is still an auction. The winner still extracts. The user still pays. Legibility matters — but it is not salvation.
And the "MEV is just a market" position contains a kernel of truth. Block ordering is a scarce resource. Pricing it is not inherently evil. The evil is the opacity: the unaccountable intermediary capturing the spread between what the market would pay and what the buyer can see. That is the structural sin.
I didn't say the fix is easy. I'm saying the fix is already being assembled inside the broken system. MEV-Share redistributes a slice of the rent back to users. Inclusion lists limit what builders can hide. Enshrined PBS, if it ships, eliminates the relay as a trust point. Decentralized builder networks are in production trials. None are perfect. All are better than the current state.
The bottleneck was never the technology. It was the market structure. And market structure can be attacked with better incentives, better data, and better defaults. None of this excuses the cartel. It just means the panic should be calibrated: the risk is real. The catastrophe narrative is not yet earned.
You don't fix a cartel by asking it to be kind. You make its coordination expensive. Watch the concentration metrics. Watch the private relay share. Watch who fights the transparency proposals hardest.
The next block is being built right now. The question is whether anyone will be allowed to see its price.
