Understanding MEV on Sui: Why Architecture Matters

Most discussions of MEV assume Ethereum's architecture: a global mempool, sequential execution, and an account-based state model where every transaction competes for the same shared state. Sui takes a different approach on all three fronts. Understanding how MEV shows up on Sui means understanding why these architectural choices matter.

Object ownership and parallel execution

Sui organizes state around objects rather than accounts. Each on-chain asset (a coin, an NFT, a liquidity position) is a discrete object with a unique ID and an explicit owner. Transactions declare upfront which objects they intend to read or modify. When two transactions touch entirely different objects, they can execute in parallel with no ordering dependency.

This is a meaningful departure from Ethereum, where every transaction against Uniswap V3, for example, touches the same pool contract storage slots. On Ethereum, the ordering of those transactions is the primary lever for MEV extraction. On Sui, transactions touching different objects are causally independent. There is no ordering between them to exploit.

However, when transactions touch the same shared objects, such as a shared DEX pool, they still require consensus ordering through Mysticeti. In those cases, transaction ordering matters again, and MEV dynamics re-emerge around shared state contention.

The consequence: traditional frontrunning, where a searcher observes a pending transaction and submits a higher-gas version ahead of it, does not have a straightforward analogue on Sui. There is no public mempool to observe pending transactions. And even if there were, the parallel execution model means many transactions simply have no meaningful ordering relationship. That said, validators and full node operators do see transactions before finalization, so the attack surface is different rather than nonexistent.

Where MEV does exist on Sui

The absence of traditional frontrunning does not mean the absence of MEV. Value extraction on Sui appears primarily through cross-pool arbitrage. Sui's DeFi ecosystem includes multiple concentrated liquidity AMMs (Cetus, Turbos, FlowX) alongside the DeepBook central limit order book. Price discrepancies across these venues create arbitrage opportunities.

The mechanics are familiar to anyone who has studied fragmented equity markets: a large swap on one venue moves the price. That price is now stale on other venues. A searcher detects this and submits a corrective transaction. Same strategy as statistical arbitrage in TradFi, different infrastructure.

What makes Sui interesting is how searchers find these opportunities. Without a mempool, searchers cannot observe pending transactions before execution. Instead, they appear to monitor finalized state. Sui produces checkpoints roughly every 300 milliseconds. Searchers watch for pool state changes and respond as fast as possible.

What we see at Birdai

We have been analyzing MEV activity on Sui and building tools to make this landscape more transparent. Our analysis has identified a significant number of transactions that exhibit MEV patterns: multi-hop swaps across venues, arbitrage cycles, and backrun patterns following large swaps.

The distribution of this activity reveals some interesting structural features of Sui's MEV landscape, which we explore further in our companion piece on searcher concentration.

Implications for protocol design

Sui's architecture makes certain MEV strategies structurally harder and others easier. Sandwich attacks, the most user-harmful form of MEV on Ethereum, are structurally difficult on Sui because they require inserting transactions around a victim's transaction, which demands mempool visibility and ordering control that Sui's model does not readily provide.

Cross-venue arbitrage, however, is alive and well. This form of MEV is generally considered benign since it corrects price dislocations. But the competitive dynamics of that arbitrage matter. When a small number of participants consistently capture the majority of these opportunities, the result looks more like monopolistic extraction than healthy competition.

Sui's execution model provides structural protection against certain attack vectors, but it does not eliminate value extraction. Understanding where and how MEV occurs is a useful first step toward building infrastructure that channels it productively.