MEV-related strategies now make up the biggest share of Ethereum priority-fee spending, accounting for about 66% to 80% of weekly outlays, according to a new market-structure study from Greenfield Capital. The firm says Telegram bots, once the dominant force in retail-driven fee bidding, have been overtaken as block-building economics on Ethereum shift toward MEV-heavy order flow.
The report also flags a change at the wallet layer. Greenfield says user incentives may be getting distorted by rebate mechanisms tied to order flow auctions, with MetaMask users paying an average priority fee of about $1.20 per transaction, up from roughly $0.15 in early 2024. GMGN users are paying more than $2.10 on average, while Wintermute sits near $3 per transaction, though the latter is tied to profit-seeking MEV activity.
How Greenfield frames the Ethereum fee market
Greenfield says the goal of the research is to identify which types of order flow are driving priority-fee demand on Ethereum and which participants inside those categories are paying the most to win ordering and inclusion. The study groups activity across MEV, Telegram bots, smart contracts, DEX frontends, wallet-based trading and solver-driven systems.
The firm describes the block-building market as one of the key places where value gets priced, routed and allocated after Ethereum’s shift to proposer-builder separation through MEV-Boost. Since the Merge in September 2022, validators have increasingly monetized execution-layer ordering rights by selling block construction to specialist builders in auctions. Under EIP-1559, base fees are burned, so the execution-layer revenue that comes from block-building auctions is the only validator income stream directly linked to blockspace demand and not funded by dilution.
That matters, Greenfield argues, because the structure of the builder market affects more than who wins the right to build blocks. It also shapes validator revenue sustainability, the efficiency of blockspace allocation and the execution quality users receive.
The report defines priority fees broadly. It includes both native priority fees and direct builder payments sent through coinbase.transfer, then aggregates both at the transaction level as a single measure of what order-flow originators are willing to pay for state access, transaction ordering and inclusion.
Looking at spend by originating address, Greenfield says Ethereum’s priority-fee market has changed sharply over time. The market was once heavily concentrated, with jaredfromsubway and Banana Gun accounting for a large share of spending. It is less concentrated now, and total spending on priority fees has also fallen materially since early 2024 in both dollar and ETH terms. The report links that decline to weaker market conditions and less retail speculation on Ethereum mainnet, which reduced the number of profitable opportunities worth bidding for.
After applying Flashbots labels available on Dune to classify MEV-related flow, Greenfield says the broad market has shifted from retail-facing Telegram trading bots and terminals, which dominated at the start of 2024, to MEV-related activity, which now leads the market. The firm adds that the current labels likely still undercount MEV to some degree because searchers frequently rotate strategy contracts and maintaining complete coverage is difficult.
At the same time, the report says Telegram-bot fee spending had fallen as token sniping opportunities dried up on Ethereum, but has shown renewed strength in recent weeks. It also identifies a steady shift in retail-facing order flow away from DEX frontends and toward wallet-based trading, especially MetaMask Swaps.
Smart-contract interactions still absorb meaningful fees
Greenfield separates out a “smart contract” category covering priority fees paid by users who interact with major token contracts such as USDC and USDT for actions like transfers and approvals.
That category has remained relatively stable in nominal terms throughout the year, with weekly spending running between $100,000 and $200,000. As nominal spending in other categories declined, its share of total priority-fee spending has risen.
The report says these transactions generally are not competing for priority state access or faster inclusion, so there is little obvious reason for them to carry extra priority fees. Yet because they involve some of the most frequently used contracts on-chain, millions of transactions each week paying only slightly more than necessary add up. Even with average priority fees 99% lower than those seen in other categories, the segment still represents 24% of all priority-fee spending.
Top spenders show a steep power-law distribution
Greenfield then examines the top 100 priority-fee spenders over the last 45 days and finds a strong power-law distribution. Two MEV-related addresses, jaredfromsubway and 0xbdb3, account for a large share of the total, spending more than $2.4 million in priority fees over that period. After the top 15, spending drops off sharply.
On that sample, MEV strategies account for 69% of total priority-fee spending, down from around 80% at the start of the year. The biggest non-MEV spenders are Telegram bots, including GMGN Terminal, Maestro and Banana Gun. Greenfield highlights GMGN in particular, saying it ranks third overall and has nominal fee spending in the same range as jaredfromsubway and Wintermute. Major token contracts such as USDC and USDT, along with MetaMask wallet trades, appear next.
A more detailed label breakdown shows a changing frontend market. Large Telegram bots such as Banana Gun and Maestro had declined, while MetaMask trading rose into the top tier of frontend fee spenders. More recently, Telegram bots have rebounded, led by GMGN Terminal after its expansion to Ethereum mainnet. Greenfield says GMGN now tops both the frontend label category and the router label category, accounting for 41% and 37% of total priority fees within those buckets.
The report says GMGN’s growth appears to have added new users to Ethereum rather than simply taking share away from incumbent products. In that reading, the total amount spent in the frontend category increased, while MetaMask’s nominal priority-fee spending stayed flat but lost share as the category expanded.
Among meta-aggregators, Greenfield says solver-based systems such as 1inch Fusion, CowSwap and Uniswap X have been effective at keeping priority-fee spending low because solvers handle execution complexity on behalf of users, and CowSwap can batch orders. In router labels, MetaMask Swap Router had led since mid-2025 but was recently overtaken by Maestro Router. Uniswap, 0x, Aave, 1inch and Banana Gun Router follow at similar shares.
For MEV labels, which Greenfield says are incomplete by nature, the report still shows a resurgence by Wintermute in marked contracts and a notable return by jaredfromsubway in recent weeks.
Wallet rebates and the rise in observed user fees
One of the harder issues to measure from on-chain data alone is the spread of priority-fee rebates through OFA systems. Greenfield says OFAs do more than sell backrun opportunities to the highest-bidding searcher and return that bid to the order-flow provider. Some now offer fixed-percentage rebates on all priority fees.
Under that model, transactions entering through an OFA RPC are shared with all eligible builders, and the bundle goes to the builder willing to rebate a fixed share of the priority fee. The report says that share is often 90%.
Different OFAs use different rebate logic and business models, and the rebates usually are not returned to the originating address in the same block. They are commonly settled on a periodic basis later on. That makes it impossible, using on-chain data alone, to attribute the rebate to any specific transaction with precision.
Greenfield says this can distort conclusions about who is really paying for priority access and how much they are actually willing to pay, because some of the fees visible on-chain may eventually flow back upstream to order-flow sources. Even so, the firm says the broader conclusions still stand. For MEV-related traffic, bids are tightly calibrated already, and from a builder’s perspective, receiving a 90% rebate can be economically similar to lowering the bid by 90% for a given opportunity.
The report argues that consumer applications are particularly well placed to benefit from OFA rebates because they control RPC selection for users and often handle relatively uninformed order flow where priority fees are set loosely and imprecisely, leading to frequent overpayment. Greenfield says this may explain the high priority-fee levels seen in wallet-originated flow, including the roughly $1.20 average paid by MetaMask users.
Because these rebates often are not passed on to MetaMask users, the report says a conflict can emerge between minimizing user costs and maximizing wallet revenue. Wallets encourage usage, typically recommend default priority fees and can influence what most users end up paying.
Greenfield also points to consolidation at the OFA layer. The report names Blink’s acquisition of Merkle and the purchase of MEV-Blocker by Consensys/SMG, the operator of SERVO MEV protection, from CoW DAO. Those relationships, it says, increasingly affect order-flow routing and who wins blocks, even though much of the associated economics remains hard to observe directly on-chain.
Given these attribution problems, Greenfield says measured priority fees in consumer apps should usually be treated as an upper bound on what users are truly willing to pay for priority access and inclusion, not as a precise estimate of their net cost.
Telegram bots lost ground, then started to recover
Greenfield’s data shows a broader migration in retail flow. Telegram bots were a defining source of fee pressure in 2024, but their spending dropped as token-sniping opportunities on Ethereum became less common. That opened the way for MEV-related strategies to take over as the dominant source of priority-fee demand.
The shift has not been one-way. In recent weeks, the report says, spending from Telegram trading bots and terminals has grown strongly again. GMGN stands out among newer entrants. After bringing its product to Ethereum mainnet, it quickly rose to the top of both frontend and router label categories, suggesting that fresh retail activity is still being pulled onto Ethereum through these interfaces.
Three MEV markets, three distinct structures
To break down the MEV side of the fee market, Greenfield groups activity into three common categories on Ethereum: statistical arbitrage, sandwich MEV and atomic arbitrage. It reuses classifications from its prior work for statistical arbitrage and relies on Dune user @hildobby’s categorizations for sandwiching and atomic arbitrage.
The report notes that MEV searchers often have to spend a large share of extractable profits on priority fees to beat rivals for the same opportunity. In earlier work, Greenfield used Binance marks to estimate margins and assumed average statistical arbitrage margins of about 10%, implying that roughly 90% of value goes to priority fees. The biggest statistical-arbitrage opportunities can reach margins of 30%, the firm said in that prior framework.
Greenfield stresses that priority fees are not the same thing as realized MEV profit. Different strategies and different searchers have different average margins depending on market structure and economies of scale. Even so, the firm considers priority fees a useful and reliable proxy for understanding the market.
Across both DEX volume and priority-fee spending, statistical arbitrage is the biggest of the three categories. Greenfield says 70% to 85% of MEV-related DEX volume and usually about 50% of MEV priority fees come from stat arb.
Atomic arbitrage, by contrast, shows more pronounced fee spikes and has been rising since early 2025. In recent weeks it has often accounted for more than 25% of total MEV priority fees. Greenfield says one possible explanation is that a gas-limit increase in November 2025 made multi-hop atomic arbitrage more competitive, a change that coincided with a larger share of DEX volume tied to atomic arbitrage.
Sandwich-related fee share has been on a long decline from the start of 2024, when it made up more than 60% of MEV-related priority-fee spending. Greenfield links that shift to more retail flow bypassing the public mempool as MEV-protection infrastructure matured, including private RPCs and dApp-level solutions such as CowSwap batching. The report says 92% of DEX volume has been routed through private mempools in recent months.
That trend has shown a reversal lately. In recent weeks, sandwich-related spending appears to have climbed again, with roughly 35% to 51% of MEV priority fees coming from the strategy.
Fee intensity differs sharply by MEV strategy
Greenfield compares the three categories by average priority fee per transaction and by priority fees as a percentage of DEX volume. Statistical arbitrage has the lowest fee intensity on both measures, which the report attributes to more frequent opportunities with lower extractable value per trade.
Atomic arbitrage is much more fee-intensive. For each $1 in priority fees spent, it corresponds to only about $10 in DEX volume, and on average 0.4% of DEX volume tied to atomic arbitrage is paid out as priority fees.
The report also says about 38% of current Uniswap volume comes from statistical-arbitrage-related trades. That share was as low as roughly 5% in early 2026 and above 45% in early 2025. Greenfield says the share of Uniswap volume tied to stat arb is strongly correlated with Deribit’s BTC DVOL index, and even more strongly correlated with ETH 12-second realized volatility if measured block to block. In other words, the opportunity set for stat arb rises and falls with volatility across assets traded on centralized exchanges and AMMs.
Across all three identified MEV strategies, Greenfield estimates they account for about 49% of total Uniswap volume today, down from as much as 56% in early 2024.
Stat arb: Wintermute faces a new challenger
Within statistical arbitrage, Greenfield tracks DEX volume, total priority fees paid and relative fee share across participants. The relative distribution is where the market structure becomes most visible.
From early 2024 to September 2024, Wintermute and SCP/beaverbuild together controlled about 50% of the market, with the rest spread across a long tail. After September 2024, a searcher labeled “Kayle” in Greenfield’s earlier work quickly gained share.
By April 2025, though, Wintermute and SCP had regained and even expanded their positions, squeezing out competitors and taking more than 75% of the market between them. After SCP gradually moved from beaverbuild to BuilderNet, its share began to slip from December 2025 onward. In the first three months of 2026, Wintermute controlled 70% of the segment, SCP sat near 5% and 0x7cDa58 followed behind. Greenfield says that was the most concentrated period inside statistical arbitrage.
More recently, an address labeled 0xbdb3ba has emerged as a challenger. By priority-fee spending, it now commands a share roughly similar to Wintermute’s, and the two together account for 83% of total fee spending in the statistical-arbitrage category.
Greenfield says the contract uses an unusual approach, bundling all stat-arb swaps into a single transaction that can contain more than 50 one-sided swaps. The team behind the contract is still unclear. Manual inspection suggests that, like Wintermute, it appears to favor routing to Titan, but that observation alone does not tie it to any known builder or indicate vertical integration.
Sandwich MEV: jaredfromsubway still dominates
The sandwich market looks different. Greenfield sees a larger drop in sandwich-related fee spending over the long run, but the most striking feature is how little jaredfromsubway’s dominance has changed. Its share of relative priority-fee spending has hovered between 70% and 85% for years.
The report notes that the data shown here was collected before more than $15 million was stolen from jaredfromsubway’s main contract. How that event will affect sandwich competition in the months ahead remains an open question.
When jaredfromsubway is manually excluded, the remaining market looks fast-moving. Participants rise and fall quickly, much faster than the long-tail behavior seen in statistical arbitrage when Wintermute and SCP are removed. Greenfield’s reading is that the top of the sandwich market is stable, while the layer below it is not.
Recently, 0x01FD has established itself in second place behind jaredfromsubway, accounting for about 20% of total sandwich-related priority-fee spending and holding that position for several weeks. Greenfield says it appears to be the most credible challenger jaredfromsubway has faced in a long time.
As for the sharp jump in sandwich-related fee spending over the past three weeks, where the strategy represented 45% to 60% of MEV fee spending, Greenfield says the cause cannot be known with certainty from the available data. One plausible explanation is a priority-fee war between jaredfromsubway and 0x01FD as competition intensified, compressing margins. Those two are the only meaningful drivers behind the increase, according to the report.
The same shift appears in fee intensity. Greenfield says jaredfromsubway is now spending about three times as much priority fee per $1 of DEX sandwich volume as it was four weeks ago, possibly in response to pressure from 0x01FD. The report adds that 0x01FD appears likely to be run by the same team behind BobTheBuilder.
Atomic arbitrage: the most fragmented MEV segment
Atomic arbitrage presents another picture. The market is somewhat smaller overall, but the opportunity set is more variable, producing larger spikes.
Greenfield says that fits the mechanics. Large atomic-arbitrage opportunities often come from a single user making a highly price-impactful swap in an AMM pool, whether by mistake or by choice. That temporarily pushes AMM prices out of line with one another and creates room for atomic arbitrage to restore alignment, even if centralized exchange prices have not diverged in the same way.
The report points to the AAVE frontend swap incident referenced earlier as an example. A trade that swapped $50 million into just $35,000 triggered the largest atomic-arbitrage priority fees observed since the start of 2024. Earlier in the report, Greenfield also said the week of March 9 recorded both the highest total priority fees and the highest single-transaction priority fee in recent years, driven by a user on the AAVE frontend who accepted a high-slippage trade, lost $50 million and triggered a $34 million spike in reverse-execution priority fees.
Such events are usually rare and often linked to user input errors. Statistical arbitrage is different because it arises whenever CEX and DEX prices diverge over the 12-second block interval, while sandwich opportunities persist as long as some users continue to send unprotected trades into the public mempool. That, Greenfield says, explains why atomic-arbitrage fee patterns are more volatile.
It is also the only one of the three MEV segments where competition is highly dispersed and no single player regularly controls the market. Greenfield says that may reflect the segment’s smaller overall size and weaker economies of scale, especially because flash loans reduce the capital barrier.
That is a very different setup from statistical arbitrage, where latency, inventory and exchange fee tiers matter and where Wintermute holds advantages, or from sandwiching, where jaredfromsubway’s large inventory position creates similar scale effects that support persistent dominance.
Who is really paying for block value
Greenfield closes the first part of its broader study by saying Ethereum priority-fee demand is not one unified market. It is a collection of separate order-flow submarkets with very different economic drivers.
Since 2024, the composition of valuable order flow has changed materially both by category and by participant. The contrast is especially visible inside MEV, where statistical arbitrage, sandwich attacks and atomic arbitrage differ sharply in concentration, opportunity frequency and fee intensity.
This part of the research focuses on the demand side: who creates valuable order flow and why they pay priority fees. The next part will move to the supply side, examining how builders compete for that flow, how builder surplus is generated and how increasingly specialized order-flow providers and their service requirements are reshaping the block-production market.
Greenfield disclosed that it has investments in Gattaca, the parent company of Titan, as well as CoW Protocol and NuConstruct. The firm says the analysis is based entirely on public on-chain data obtained through Dune Analytics, reflects independent research views and should not be treated as investment advice or as an endorsement of any project, protocol, builder, searcher or other market participant mentioned in the study.