[Constitutional] AIP: Transition Arbitrum One ordering policy to Priority Gas Auctions (PGA)

[offchainlabs]

Evolving Arbitrum’s Transaction Ordering Policy: From Timeboost to a More Open and Competitive Sequencing Market

Update June 11 2026: This AIP and the proposal to transition Arbitrum One to use priority fees for transaction ordering does not introduce new forms of MEV or create MEV. Just like it always has been, Arbitrum One will continue to have a private mempool at the sequencer. This means that the order of transactions are published only after the transactions have already been sequenced by the sequencer. Additionally, the ordering of transactions cannot be manipulated to extract MEV after the sequence has been published publicly. This subtle but important detail means that MEV-related sandwich attacks and frontrunning were not and are not possible on Arbitrum One.

Abstract

This Constitutional AIP proposes to disable Timeboost on Arbitrum One and replace it with a Priority Gas Auction (PGA) mechanism, an ordering policy that’s more familiar for actors who are willing to pay for transaction priority, allowing more market participants to be a part of Arbitrum’s next phase of growth. In addition, it would sunset Timeboost on Arbitrum Nova.

Timeboost was the first foray into experimenting with MEV via an express lane auction mechanism on Arbitrum One. Deploying a new transaction ordering policy (PGA) represents the next step in the journey to position Arbitrum as the premier platform for the programmable economy.

This version of a PGA-based ordering policy will first sort transactions based on priority fee in 125ms PGA rounds before publishing a block at 250ms. There will also be an anti-starvation mechanism which works as a primitive form of inclusion guarantee to help ensure that transactions that didn’t get included in a round because of a low or null priority fee can be included within a couple of blocks, with the exact number determined by the number of PGA rounds.

If approved, this AIP would:

• Disable Timeboost’s express lane auction on Arbitrum One and Arbitrum Nova following the successful passing of this proposal in an onchain vote.
• Enable the collection of priority fees added to transactions via an arbOwner call.
• Activate PGA on Arbitrum One, enabling open, permissionless gas-price priority bidding for transaction ordering.
• Establish an on-chain revenue distribution contract that routes the new priority fee proceeds: 97% to the ArbitrumDAO Treasury and 3% to the Arbitrum Developer Guild.

Explicitly, this AIP only proposes sunsetting Timeboost on Arbitrum Nova; it does not propose adopting PGA as an ordering policy for Nova, in line with the Nova Minimization AIP [Constitutional] AIP: Minimize Arbitrum Nova

Motivation

Timeboost was launched on Arbitrum One in April 2025, a novel transaction ordering policy designed to put user fairness and spam reduction at the forefront while, for the first time, creating additional revenue to the ArbitrumDAO via Timeboost auction fees. While Timeboost has delivered meaningful benefits including user fairness, and a new revenue stream for the DAO, some structural considerations have also come to light that are worth addressing:

• Timeboost’s architecture represents a large barrier of entry with emerging DeFi primitives such as propAMMs, which require cheap and frequent, low-latency, priority-ordered inclusion to update onchain parameters.
• Timeboost participation remains limited as many builders that entered the auction at launch exited. Additionally, the overhead to build a custom tool that would participate in an ahead-of-time auction increased the difficulty of adoption.

A fundamentally different model is needed and the more familiar ordering policy: Priority Gas Auctions (PGA) stands out as a clear option. PGA addresses the challenges above by making ordering determined by per-transaction priorityFeePerGas bids rather than a periodic, ahead-of-time express lane auction:

• Competition is distributed across every transaction and every block, making collusion structurally harder.
• PGA mechanics are well-understood by CEX-DEX arbitrageurs and MEV searchers operating on other EVM chains (Ethereum, Solana, Base, OP Mainnet, Unichain), and propAMMs, lowering the barrier to entry for these participants.
• Additionally, we introduce an anti-starvation mechanism that helps facilitate that non-priority paying or low priority transactions get included within a couple of blocks. The exact wait time for a transaction is dependent on the selected parameters for the PGA round and the priority fee expressed in a transaction.

While ArbOS 60 Elara introduced the capability for Arbitrum chains to collect priority fees (via the maxPriorityFeePerGas field in EIP-1559 type 2 transactions), it deliberately did not enable this feature on Arbitrum One or Nova. It was clearly noted that enablement would require a separate Constitutional DAO vote. This proposal would enable the collection of priority fees.

Rationale

A new transaction ordering policy offers the ArbitrumDAO an opportunity to continue to capture additional revenue that should not come at the expense of users, with the introduction of an anti-starvation mechanism that helps facilitate transactions with 0 priority fees or low priority fees get included. Transactions remain subject to paying the latest base fees.

Rather than capturing arbitrage opportunities by having the fastest hardware or bidding ahead-of-time, participants can win these opportunities by bidding in a just-in-time auction to gain priority ordering.

The priority auction is permissionless and participation is open to everyone, where transactions with the highest bids get included first within a 125ms PGA round, and ties are broken by arrival timestamp.

The Arbitrum DAO can configure all aspects of PGA, including enabling or disabling it, the auction’s design, and how to handle proceeds.

Key Terms

Priority Fees: are user-defined incentives paid in addition to the network’s base transaction fee. Priority fees are denominated in the Arbitrum’s native gas unit and are paid directly to the Arbitrum DAO Treasury.

Priority Gas Auction: an auction in which multiple participants simultaneously submit transactions with their specified priority fees in order to secure preferential ordering by the sequencer.

Specifications

1. Priority Gas Auctions (PGA)

We are formally introducing our Priority Gas Auction design, which will enable Arbitrum One users to affect their transaction ordering within a block by specifying a maxPriorityFeePerGas in each transaction. We added some modifications in the ordering mechanism to help facilitate valid transactions with low or null priority fees are included in a timely manner, by adding a anti-starvation boost to all transactions that remain in the mempool after a PGA round.

1.1 How PGA Ordering Works

1. The sequencer receives a transaction and places it into a private mempool tagged with: (a) a priority value p equal to its priorityFeePerGas bid, and (b) a fine-grained arrival timestamp (at the sequencer) in milliseconds for tie-breaking.
2. Based on the arrival timestamp (at the sequencer), transactions are slotted into PGA rounds. Transactions arriving during a running PGA round are held and enter the sequencer’s mempool only at the start of the next PGA round. Depending on the block time of the chain, which we call B, the PGA round time is B/K, for some K chosen by the chain operator with permission from the DAO. On Arbitrum One, given the 250ms block time and a proposed value K = 2, the nominal PGA round duration is 125ms.
3. At the start of each round, all queued transactions are ranked in descending order of priorityFeePerGas, with ties broken by earliest arrival timestamp.
   1. The priorityFeePerGas is calculated as follows: priority_fee_per_gas = min(transaction.max_priority_fee_per_gas, transaction.max_fee_per_gas - block.base_fee_per_gas)
   2. Note: The sequencer collects, the computed the priority fee per gas, according to EIP-1559.
4. Transactions are appended to a block greedily until: the 32 Mgas target is hit, the calldata limit of 95,000 bytes is reached, or the round expires. The block hard limit is 64 Mgas; individual transactions are capped at 32 Mgas (L−T), where L is the block size limit and T is the Target size.
5. After each PGA round, if the mempool is non-empty, all waiting transactions receive a virtual priority boost (see section 1.2 for more details).
6. The sequencer will publish a regular block at 250 ms intervals and publishes ordered transactions within every 125ms round, via a new fast sequencer feed. In the case where a block is full before the 125ms are completed, the sequencer will start building the next block. Note: There is a limit of 8 blocks per 1 second that can be produced.

1.2 Anti-Starvation: Priority Boost Mechanism

This mechanism helps facilitate that transactions with null or low priority fees will get included within a reasonable time. The mechanism works as follows:

• At the end of each round, all remaining transactions receive a virtual priority fee or boost, which gets calculated by adding p / (2K) to all waiting transactions in the mempool, where p is the priority of the last transaction included in the previous PGA round, and K is the number of rounds per block.
• This boost is a virtual ordering artifact only — it does not change the fee charged to senders.
• During the following PGA round, any transactions that were received in the previous PGA round that were not included in that round will be sorted by descending priority, taking into account the boost, before being appended to a block. This way, transactions that didn’t get included in a previous PGA round due to a low/null priority fee get a virtual boost to facilitate their inclusion in the following PGA rounds.
• The sequencer must maintain a nonce cache for transactions with future nonces in the same fashion as its currently done today, preserving their original priority for restoration when they re-enter the mempool. Transactions in the nonce cache must NOT receive the boost.

1.3 Configuration Parameters

Parameter	Default (Arb One)	Configurable By	Notes
K (rounds per block)	2 → 125ms rounds	Chain Operator, with permission from the DAO	Higher K = faster rounds up to a limit. Otherwise it becomes FIFO
Anti-starvation boost divisor	2K	Implicitly set via K	Lower K = faster anti-starvation

This AIP proposes granting the current sequencer operator the below rights to make the following adjustments from time to time for a period of two (2) years. The rights described below are expected to only be exercised in circumstances where doing so would enhance PGA’s long-term stability, preserve or improve the user experience for those using Arbitrum One, increase the security posture, resiliency, or stability of the chain, and/or otherwise help increase revenue for the ArbitrumDAO:

• The right to change the K round parameter, a value which determines the number of rounds per block. The starting default is 2 but adjustments could be to any value between [1 & 10], inclusive.

2. Timeboost Sunset

Upon passage of this AIP the following steps will be applied to both Arbitrum One and Arbitrum Nova:

• Timeboost users that have funds locked in the Auction contract will be able to remove their deposits anytime.
• The Timeboost autonomous auctioneer will cease accepting new bids.
• The sequencer will cease enforcing the 200ms non-express lane delay for normal transactions (note that this delay only exists when there is a Timeboost Express Lane owner).
• The express lane endpoint will be decommissioned. Any transactions submitted to this endpoint after decommissioning will be rejected.
• Any collected funds from the auctions can be flushed to the ArbitrumDAO’s L2 Treasury Timelock contract at anytime, by anyone. Note that this is possible today as well.

Steps to Implement

The following implementation sequence is proposed:

1. A post to the ArbitrumDAO Forums (this post) to initiate productive, collaborative discussions with the DAO and its members, followed by governance call(s) to answer any questions.
2. Temperature Check (Snapshot): This forum post will advance to a temperature check vote on Snapshot sometime after the forum post is live. The temperature check will gauge DAO sentiment on both the Timeboost sunset and a subsequent PGA activation.
3. OnChain Vote (Alt Arbitrum Gov): Following a successful Snapshot temperature check, this proposal will proceed to an on-chain constitutional vote for ~14 days. The ArbitrumDAO will be notified on this thread and via official channels before the vote goes live.
4. PGA Activation: On the same day the vote is passed and executed onchain, PGA will be activated on Arbitrum One. Timeboost’s express lane and associated delay logic will be simultaneously decommissioned.

---

Overall Cost

The ArbitrumDAO has full control over how to spend the proceeds from the PGA auction, with 3% of auction proceeds to be set aside for the Arbitrum Developer Guild, which helps fund core Arbitrum development.

This proposal comes at no additional development cost to the ArbitrumDAO given it is already covered by an OCL and AF contract through January 2027. The engineering work will be owned by Offchain Labs in its capacity as an Arbitrum Aligned Entity.

[nir]

Wouldn’t this open the possibility for users to be frontruned and sandwiched?

[MconnectDAO]

Risk Management Gaps in the PGA Transition

Thank you to Offchain Labs for this detailed proposal. As a DAO governance researcher, I’d like to raise a few concerns regarding risk management that I believe should be addressed before this moves to Snapshot.

1. MEV/Frontrunning Risk Unaddressed
   The proposal does not explicitly address how users will be protected from frontrunning and sandwich attacks under PGA. User @nir raised this concern but has not yet received a response. Unlike Timeboost’s sealed Express Lane auction, PGA makes transaction ordering publicly visible via priorityFeePerGas bids this creates a significantly larger MEV attack surface. Chains like Ethereum mainnet have seen billions lost to sandwich attacks under PGA-style ordering. Arbitrum should not replicate this without a mitigation plan.

Suggested mitigations to consider:

Optional private mempool / MEV Blocker integration for DeFi users

On-chain MEV monitoring post-launch

Clear user-facing documentation on MEV risks

2. K Parameter — Unchecked Centralization
   The proposal grants the chain operator authority to modify K (rounds per block) between 1–10 for 2 years without DAO vote. While bounded, this still allows significant unilateral control over ordering behavior. A transparency mechanism (e.g., mandatory on-chain announcement before K changes) would strengthen governance oversight.

3. Anti-Starvation ≠ Anti-MEV
   The anti-starvation mechanism protects low-fee transactions from exclusion but it does not protect against ordering manipulation. These are two separate problems and the proposal conflates them in its risk section.

I support the intent of this proposal and the revenue routing to the DAO Treasury. However, I’d request Offchain Labs to respond to the MEV vulnerability question before this proceeds to an onchain vote.

[offchainlabs]

There will be an open call to discuss this proposal. Details below:

Thursday 11 Jun · 15:00 – 15:45 UTC
Video call link: https://meet.google.com/ayd-mrub-emr

[leyiang]

I’m totally all for this, it’s great news!

I said long ago that Timeboost would ultimately be used by only a tiny handful of whales, making Arbitrum even more centralized, and yet it took you two years to believe it. There were clearly many mature and simple alternatives. Yes, I am the one who was driven out of the Arbitrum ecosystem by the Timeboost whales. I will try to come back if this AIP effective
Constitutional AIP: Proposal to adopt Timeboost, a new transaction ordering policy - #105 by leyiang

[offchainlabs]

Thanks for your feedback and questions @MconnectDAO and @nir - we have added an update at the top of this post that we hope clarifies your questions.

[Entropy]

Entropy is supportive of this proposal and believes transitioning from Timeboost to a PGA-based ordering policy is the right move for Arbitrum at this stage.

Over the past several months, our team has spent significant time analyzing Timeboost’s performance, Arbitrum’s fee mechanism relative to competing chains, and the revenue implications of ArbOS 60. This research has been done in collaboration with Offchain as part of an ongoing advisory dialogue on ordering policy, and we’re glad to see this proposal move forward.

Since launch, Timeboost has generated ~$7.46M in cumulative fees, which is a meaningful contribution to Arbitrum’s monthly revenues. That said, the express lane has not become the default execution path for onchain MEV. 87% of atomic arbitrage and ~98% of liquidation activity on Arbitrum bypasses Timeboost entirely, and despite being live for over a year, auction participation has remained limited. As Offchain outlined in the proposal, the limited participation is largely a result of the high barrier to entry and the overhead associated with building custom tooling to participate in ahead-of-time auctions. This has led to three entities account for ~97% of all wins:

In February 2026, Kairos emerged as a secondary marketplace, which furthered compressed competitive bidding and lowered annualized revenue. From our research on revenue capture specifically (conducted in March 2026), Timeboost was capturing just 1.74% of atomic arbitrage revenue on Arbitrum, compared to 28.83% on Ethereum via MEV-Boost and 7.34% on Base via priority fees. Much of this gap comes down to Timeboost’s design where a single winner controls the express lane for an entire 60-second round and the previously mentioned dedicated bidding infrastructure to price expected MEV across that full window. The change in auction dynamics due to Kairos has pushed annualized revenue down to approximately $2M.

Switching to PGA addresses the competitive dynamics that have limited Timeboost. As Offchain highlighted in the proposal, this is a mechanism that searchers across the EVM ecosystem are already familiar with, which should broaden participation meaningfully compared to Timeboost’s custom bidding infrastructure requirement. Overall, our team views this as an important structural improvement for the network’s revenues and we look forward to supporting the proposal through the offchain and onchain votes.

[leyiang]

You have a very good point, but I don’t think the bidding tool is the biggest obstacle for Timeboost. The real problem is clearly the design of the mechanism itself. At least for me, competing with Wintermute is impossible: when they win a one-minute slot, they can deploy a huge amount of capital and execute a massive volume of arbitrage trades, while I simply don’t have enough capital to execute that many trades. So in terms of expected profit, I can never beat them. That means my expected return in the bid has no advantage at all, and the only rational choice is to give up bidding entirely. In any case, it’s a good thing that the community has noticed.

[offchainlabs]

Hello! This AIP now has a live off-chain temperature check vote here.

[Zeptimus]

Voting FOR, i like this one. This opens it up so anyone can pay for priority. way healthier. 97% of the fees go to the treasury at no dev cost too. I also like the anti-starvation piece keeps normal cheap transactions flowing, so were not squeezing regular users.

[maxlomu]

Voting FOR on offchain and onchain votes. It’s a reasonable change that opens up the priority fee to a more competitive and standardized market.

Anti-starvation is a clean mechanism for normal users to get included.

[Abel189]

I support this proposal.

Timeboost was an innovative experiment and successfully introduced a new revenue stream for the DAO while improving transaction ordering fairness. However, the participation model remained relatively specialized and created a higher barrier to entry for many market participants.

A PGA-based approach is easier to understand, easier to integrate with existing infrastructure, and more familiar to builders, arbitrageurs, market makers, and emerging DeFi primitives. This should help broaden participation while allowing Arbitrum to continue capturing value from transaction priority demand.

I particularly appreciate the inclusion of the anti-starvation mechanism. Maintaining a reasonable experience for users who do not actively bid for priority is important, and the proposed boost system appears to provide a practical balance between market-based ordering and transaction inclusion fairness.

Going forward, I would encourage close monitoring of user transaction costs, ordering dynamics, and the impact of future adjustments to the K parameter. Transparency around these metrics will help the community evaluate whether the new system is achieving its intended goals.

Overall, this appears to be a reasonable evolution of Arbitrum’s transaction ordering policy and a positive step toward broader participation and sustainable DAO revenue generation.

[Griff]

Vote: FOR

This mostly sounds good… glad we are continuing to experiment with how we charge for blockspace, Timeboost was a good experiment but if it didnt get adoption, i’m glad we can recognize that, learn from it, and pivot.

But one thing bugs me… wtf is the Arbitrum Developer Guild? I looked it up, sounds like a sort of protocol guild thing for Arbitrum… but is it really active? Who is leading it?

[cornellbc.eth]

Cornell Blockchain is voting FOR this proposal because we support any initiative that lowers the barrier of entry for smaller parties to compete with bigger players, and are against the monopolization of Timeboost by 3 parties. Since Timeboost captures ~2% of atomic arbitrage revenue, with the overwhelming majority of arb and liquidation activity bypassing it entirely, then switching to PGA to bring more funds to the DAO makes sense. We also appreciate the clarification that the private mempool will protect users from sandwich attacks under PGA. Additionally, we second @abel189 comment that the transaction costs should be closely monitored, especially the choice of the k parameter, to ensure that the anti-starvation measure is working as intended.

[Manugotsuka]

The following reflects the views of L2BEAT’s governance team, composed of @krst and @Manugotsuka, and is based on their combined research, fact-checking, and discussion.

We voted FOR.

We see this proposal as a sensible evolution of Arbitrum’s transaction ordering policy. Timeboost was a valuable experiment: it helped Arbitrum explore a new ordering model and introduced an additional revenue stream for the DAO. At the same time, it also came with complexity, and that complexity seems to have limited broader participation.

Moving to a Priority Gas Auction model feels like a more straightforward path. PGA is easier to understand, easier to integrate with existing infrastructure, and more familiar to the participants that actively compete for transaction priority. This should make the market more open while still allowing the DAO to capture value from priority demand.

We also appreciate the inclusion of the anti-starvation mechanism. A more competitive priority market should not mean that regular users are pushed aside. The proposed boost mechanism seems like a practical way to help low or zero-priority-fee transactions get included within a reasonable time.

Before voting, we asked our Research Team to review the proposal and validate the implementation path. The team did not identify any material issues and agreed that the proposed transition is worth supporting.

[Arbitrum]

Hey @griff,

To answer your question, the Arbitrum Developer Guild was established to recognize and support long-term contributors whose work would strengthen the Arbitrum technology, tooling, and infrastructure. Currently, the ADG receives the following allocations for its operation: 2% of the Net Protocol Revenue share from Arbitrum chains, 3% of Timeboost fees and community donations. During the initial phase, the Arbitrum Foundation acts as a facilitator for the Guild.

For a more detailed update on the ADG, see page 27 in the AF’s 2025 Transparency Report.

[Flook]

Merlyn Labs is voting FOR this proposal to transition from Timeboost to Priority Gas Auctions because it should open up the priority market to broader, more standardized competition. Routing 97% of fees to the treasury at no dev cost and preserving a fair experience for regular users through the anti-starvation mechanism are big pluses as well.

[shogun]

The following reflects the views of GMX’s Governance Committee, and is based on the combined research, evaluation, consensus, and ideation of various committee members.

Thanks OCL for clarifying the introduction of PGA is performed in a private sequencer mempool, to avoid frontrunning or sandwich attacks, which is important for perps users.

Timeboost was an instrumental revenue driver for Arbitrum, delivering ~$7.46M in cumulative fees, and annualised to $2M after Kairos as showcased by @Entropy. With PGA superseding Timeboost, there continues to be many use cases across perps, dex, and lending environments for a stronger usage of a maxPriorityFeePerGas. For GMX we observe these changes for traders, the new 125ms, private mempool based ordering mechanism is well-suited for traders interacting with GMX’s platform. The revenue split of 97% to the DAO Treasury is clear to us.

Elevating the network’s revenue with this AIP is a good direction for Arbitrum, and therefore voted FOR this proposal on Snapshot.

Our only remarks are to extend comments from @Griff:

• Who can participate in the @Arbitrum Developer Guild?
• Whose currently is in the Arbitrum Developer Guild?

[SEEDGov]

Voted FOR;

We were strong supporters of Timeboost as an innovative approach to transaction ordering. For us, it demonstrated Arbitrum’s willingness to experiment with different mechanisms instead of simply following existing industry standards, while creating a meaningful source of revenue for the DAO.

With that said, we believe this proposal reflects a pragmatic next step: adopting a PGA brings Arbitrum closer to the transaction ordering model already familiar to builders and other sophisticated market participants, reducing friction for adoption while providing a more scalable foundationfor growth.

We also support the proposal because it balances economic incentives with user protections. The anti-starvation mechanism helps ensure that users are not required to compete in priority fee markets simply to have their transactions included, while the new model creates a sustainable source of revenue for the DAO.