Inside the PropAMM Router's first billion dollars
In June, the PropAMM Router handled $154,288 across 375 swaps. We were still testing on mainnet. On November 16, less than six months later, it passed $1 billion.
USD volume through the router since its first swap on June 1, 2026. Hover or tap to read any day.
Swapped events, on-chain. USD value is taken from the stablecoin leg of each swap. Days are UTC; data through Nov 16, 2026, 22:23 UTC.Most of that growth came in bursts. New makers joined, larger orders started arriving, and weekly volume increased almost sixfold in the middle of August. Then one of the largest venues stopped filling, and volume fell for several weeks. November has since brought our busiest week yet, at $117.2 million.
As that flow grew, we worked with makers, builders and aggregators to get more orders filled at the prices makers were offering. That meant getting fresh updates into blocks, investigating failed quotes and following contract redeployments. It also gave us a closer view of how the router's execution model works in practice, as orders pass from an aggregator's initial quote to the trade that actually lands on-chain.
The billion is all volume through the router. About $110 million of it went to the Uniswap fallback; the other $899 million was filled by the eleven PropAMMs behind the router. In total, those PropAMMs have filled about $13 billion, so about 7% of their volume arrived through the router. Makers receive substantial flow through other routes.
The same distinction applies to our July introduction: the $546 million reported there covered the PropAMM ecosystem. The router itself had handled about $12 million then.
From a quote to an executed swap
A PropAMM lets a market maker supply a pool's prices rather than deriving them from a curve. The maker streams signed price updates to block builders, who put them on-chain ahead of the swaps that use them. Prices can change several times within a 12-second block, so the maker can follow the market closely and offer tighter quotes.
For an aggregator, those tighter prices are a reason to send orders to PropAMMs. The difficulty is that the price can change between simulating a trade and getting it included in a block. A venue that offered the best quote when the aggregator planned its route may no longer be the best place to execute it. As we explained in our introduction to the router, that is why we requote on-chain.
When a swap executes, the router asks the whitelisted PropAMMs for quotes again and selects the best available one. The venue is chosen using prices at inclusion time, so an order can benefit if another maker now offers a better price.
This also helps protect against quote spoofing. A venue could advertise an attractive price to draw an aggregator's order, then worsen it before inclusion. Requoting makes that venue compete against the other PropAMMs again: if another maker now quotes better, that maker gets the trade. We covered this protection in the September audit update.
The selected venue still has to deliver. The router checks the recipient's actual balance change against the user's minimum output. If the venue reverts or under-delivers, or no PropAMM can meet that minimum, it attempts a Uniswap V3 swap in the same transaction. The minimum applies there too. The router itself takes no fee.
The example below follows an order through those cases. With fresh prices in the block, a PropAMM can quote and fill it. If execution fails, Uniswap gets an attempt. Without fresh prices, the router may have to use Uniswap from the start.
Pick a case and watch what the router does when the swap is included in a block.
swapV1 runs on-chain: requote, pick, execute, and fall back to Uniswap V3 in the same transaction if the chosen venue reverts or under-delivers.Integrators reach all of this through one contract, rather than maintaining a separate integration with every maker. We started with Bebop, FermiSwap and Kipseli on the whitelist. There are now eleven venues, and those makers have already used nineteen different contracts. Five have changed contracts at least once. We update the whitelist as those changes happen, while aggregators continue calling the same router address.
That maintenance became more consequential once substantial flow started arriving.
The August increase and September slowdown
Regular flow began in the week of June 29. Through July, only two makers were filling through the router, and FermiSwap accounted for 87% of the volume going to PropAMMs.
In August, the available liquidity expanded. Tempest and TaurusFi were listed on August 4, followed by Metric on August 17. Orders were getting larger too: the median rose from $354 in July to $947 in August, then to $2,332 in September. Weekly volume jumped from $16 million to $93 million in mid-August.
It had taken until August 21 to reach the first $100 million. The next $150 million took ten days. By September 18, the router had passed half a billion.
Metric was a major part of that growth. At its busiest, it filled $44 million in a week. As its activity declined through September, the effect showed up in the router's volume. It stopped filling on September 28 and returned on October 15 with a new contract, after seventeen days away.
Weekly volume fell from an early September peak of $102 million to $33 million in the week Metric left. The decline in Metric's fills accounts for 62% of that drop. Accumulating the next $250 million took 37 days, more than twice as long as the previous quarter-billion.
The router could keep checking every venue, but with less liquidity behind those quotes, it was handling less volume. Adding makers had helped orders get filled in August; losing a major source of liquidity had the corresponding effect in September.
Weekly volume through the router by the venue that filled it, from launch week to Nov 9–15.
Since then, volume has recovered and more venues are contributing. In the week of November 9–15, seven venues each filled over $1 million, and the largest accounted for half of the volume going to PropAMMs. The weekly chart includes El Zorro, the venue built by Zonda Markets based on our PropAMM Factory, alongside the five largest venues and the remaining makers.
The last $250 million took 22 days. November is averaging $13.5 million a day so far, ahead of September's $11.7 million, our strongest complete month. November 11 set the daily record at $21.8 million.
Getting the price into the block
Having more makers available helped, but orders still needed to reach them with a usable price. Early on, this failed often enough that most router swaps in launch week ended up on Uniswap V3.
The fallback has been useful throughout. By the reporting cutoff, it had filled $109.5 million across 14,899 swaps, all at or above the users' minimum outputs. That is 18% of swaps and 10.8% of volume. It lets orders complete when the PropAMM path cannot fill them and Uniswap can still meet the minimum.
When we trace those fallbacks, the most common finding is that no fresh maker price was available in the block. The maker needs to get its update to the builder, the builder needs to place it before the swap, and the taker needs to route to a venue with an available price. A missed connection anywhere along that path can leave an order without a PropAMM quote.
Solving this has meant working with the participants on each part of the path. Builders improved the placement of price updates ahead of swaps. One of the biggest changes came in late July, when the fallback rate in blocks built by Titan, BuilderNet and Quasar fell from 27% to 10% in a week.
We worked with makers to get their updates to every builder and investigate quotes that failed on-chain. For integrators, we built SDK quoting against the prices makers are streaming, so their routing decisions can use a more current view of liquidity than stale on-chain state.
New flow can expose integration problems again. The increase in fallbacks in the second half of October came with a new aggregator whose orders initially fell back more often than existing flow. The rate declined as that integration settled.
Across the router, the fallback rate has fallen from 76% in launch week to 10.1% in the week of November 9–15. Builder coverage explains much of the remaining gap. In that last complete week, the rate was 6.9% for Titan, BuilderNet and Quasar, compared with 59% for everyone else. Those three builders account for the blocks containing 92% of all router swaps.
Share of router swaps filled by the Uniswap V3 fallback each week, and the same rate split by who built the block.
In July, almost every router swap in a block built by someone other than Titan, BuilderNet or Quasar ended up on Uniswap V3. By November 9–15, that share had fallen to 59%. More builders are now including fresh maker prices before swaps, allowing PropAMMs to fill orders that previously had to use the fallback.
The flow we are building for
Almost all router swaps arrive through another contract. Aggregators, solvers and arbitrage bots account for 99.3% of them, in transactions sent from 26,268 distinct addresses. For the person making the trade, the router can be an intermediate step they never interact with directly.
That makes the work on venue maintenance and execution reliability useful beyond the integrations themselves. An aggregator can keep its connection to the router while we add makers, follow redeployments and work on the conditions that let their quotes execute.
Order sizes also help explain where we are putting our development effort. The median swap is $1,200, and nearly half of swaps are below $1,000. But orders of at least $10,000 account for 89% of volume. Twenty-seven swaps have exceeded $1 million; the largest was $1.72 million.
Each trade-size band as a share of the number of swaps and as a share of the dollars.
Most swaps trade ETH against USDC or USDT. Stablecoin trades are less frequent but much larger: USDC/USDT accounts for 37% of volume from only 10% of swaps, with an average size of $46,000. ETH pairs average $8,000. WBTC and cbBTC against stablecoins contribute another 7% of volume.
What comes next
The numbers point at where the work is.
- More venues, and deeper ones. Volume followed the venues, up and down. If you run a market-making desk, PropAMM Factory takes you from nothing to a live venue, and we are happy to help with the listing.
- Cheaper swaps. Requoting every venue on-chain costs gas, and a venue that does not quote your pair still costs a revert. The deployed router already has
swapViaSelectedVenuesV1, which lets an integrator pass only the venues that quote the pair. We will publish the measurements in a follow-up post. - Larger orders. We are working on split orders: one order filled across several venues, with the split planned and executed on-chain in the same transaction, so that a large order is not limited by the depth of a single venue.
The contracts and the Rust, TypeScript and Python SDKs are in the repository. If you want to integrate the router, list a venue or run one, reach out to us.
How we measured
- A swap is a
Swappedevent emitted by the router at0x4ddf368080cd7946db5b459ad591c350158175e1. Reverted transactions are not counted. - USD volume is taken from the stablecoin leg of each swap. 67 swaps had no stablecoin leg and count as zero, so volume is a slight undercount.
- A fallback is any swap filled by Uniswap V3, whatever the reason.
- Total PropAMM volume counts every fill at the venue contracts we index since June 3, 2026, through the router or not.
- Days and weeks are UTC, weeks run Monday to Sunday, and weekly charts stop at the last complete week.
- Data through Nov 16, 2026, 22:23 UTC.