Why On-Chain Auctions Can't Solve MEV

On-chain auctions are incomplete. They only capture value after a transaction is public, missing the lucrative pre-publication phase where searchers and builders operate. This is the domain of private order flow and off-chain agreements.

The MEV supply chain is fragmented. Value accrues to off-chain entities like Flashbots and Jito Labs, not the on-chain auction itself. The auction is the final, commoditized step.

Protocols like UniswapX demonstrate the shift. They move complexity off-chain via intents and fill auctions, proving the on-chain settlement layer is too slow and transparent for optimal execution.

On-chain auctions leak information. Any bid placed in a public mempool reveals user intent, enabling front-running and sandwich attacks before the auction resolves. This creates a zero-sum game where searchers compete to exploit, not just execute, the revealed transaction.

The settlement layer is the bottleneck. Protocols like Flashbots Auction or CowSwap's CoW AMM must finalize on-chain, making the auction's outcome itself a target for last-look manipulation. This reintroduces MEV at the final, inescapable step.

Real-time competition is impossible. The block-time latency of Ethereum or other L1s prevents the continuous, sub-second bidding required for efficient price discovery. This structural delay guarantees inefficiency compared to off-chain systems.

Evidence: The dominant MEV-Boost model on Ethereum proves the point; its proposer-builder separation is an off-chain auction. The winning bundle is only published on-chain after the auction concludes, explicitly avoiding on-chain bidding.

On-chain auctions leak information. Protocols like Flashbots Auction and PBS reveal bid data on-chain, creating a public signal for other searchers to front-run. The winning strategy is to observe the auction and submit a marginally higher bid, not to compute the optimal value privately.

Private mempools are a band-aid. Solutions like Flashbots Protect or Titan delay information leakage but do not eliminate it. The transaction's existence and its final execution price are still revealed, allowing for latency arbitrage and time-bandit attacks in subsequent blocks.

The fundamental constraint is state. A blockchain is a public state machine; any auction mechanism that finalizes on-chain must publish its outcome. This creates an unavoidable information delta between the proposer, who sees bids first, and the network, creating new MEV vectors like proposer extractable value (PEV).

Evidence: Ethereum's PBS rollout shows that in-protocol proposer payments simply shift MEV from validators to builders. The builder market is now dominated by a few entities like Jito Labs and BloXroute, centralizing the information advantage the auction was meant to democratize.

Auctions formalize MEV extraction. They don't eliminate the value; they create a market for it. This transforms a hidden tax into a visible fee, but the economic cost to users remains. Protocols like Flashbots Auction and CowSwap's CoW AMM demonstrate this.

Auction winners centralize. The highest bidder wins the right to order blocks, which favors the largest, best-capitalized searchers or builders. This creates permissioned block building, contradicting decentralization goals. The PBS (Proposer-Builder Separation) model on Ethereum exemplifies this risk.

Latency determines outcomes. Even in a fair auction, the fastest network connection wins. This advantages proximity to validators, a physical form of centralization. The time-bandit attack vector shows how latency arbitrage undermines auction fairness.

Evidence: Ethereum's post-merge MEV-Boost dominance shows >90% of blocks are built by a handful of professional builders, proving auctions consolidate, not distribute, power.