How Much of Solana's DEX Volume Is Actually Arbitrage?
I have been digging into this question for a while now, and the numbers are pretty wild.
Quick primer: arbitrage on Solana happens in a few different flavors. Atomic arbs execute within a single transaction—> buy low on one DEX, sell high on another, pocket the difference, all in one go.
Bundled arbs spread across multiple transactions in the same block to achieve the same result.
And here is something that might surprise you: Lot of bots aren't running custom programs. They are routing through aggregators like Jupiter and DFlow.
So I built some queries on Dune to actually quantify this.
the numbers.
Starting with atomic arbitrage on Jupiter, the results were eye opening: at least 40% of Jupiter's volume is pure atomic arb activity.
Let's do some quick math. Aggregators handle roughly 60% of all Solana DEX volume, and Jupiter dominates that space with ~90% market share.
So: 0.6 × 0.9 × 0.4 = ~22% of total Solana DEX volume is just atomic arbs running through Jupiter.
When I added bundled arbs (multiple transactions across the same block to net a profit on the initial asset), Jupiter's arb share jumped from 40% to 50%, pushing the total DEX figure to around 27%. Factor in DFlow and other aggregators, and we're looking at roughly 30% of all Solana DEX volume coming from arbitrage that I have been able to track through aggregators alone.
Note : we took 2 consecutive txns to spot bundled arbs to avoid insanely heavy computation on dune
What we are missing
Here is the thing—> I am pretty confident we are only capturing maybe 40-60% of actual arb activity. The rest is harder to track with current resources, but based on hundreds of transactions I've analyzed manually:
>> Profit margins outside aggregators are significantly higher. Makes sense—>custom routing gives you more room to optimize.
h/t @0xGhostLogs for making me realise this
>>There are other types of arb strategies we haven't captured yet.
The takeaway
Conservative estimate: at least 50% of Solana's DEX volume is arbitrage on avg and on certain days, it's probably closer to 60-70%.
>> Dune's 30-minute compute limits have been a constraint, but it's still the best tool out there for this kind of analysis. If you have got ideas on how to improve the methodology or want to poke holes in my assumptions, I am all ears.
I have built a dashboard to specifically track these numbers : https://t.co/POk0BgvXNB
cc: @mert@DefiIgnas@toly@niteshnath@DanJablonski
Our friends at @Coinbase are on a mission to create the best possible trading experience for @Solana tokens
That's why they came to DFlow to power swaps ✍️
I’m excited to announce that DFlow has been acquired by MoonPay.
DFlow has consumed almost every waking moment of my life since I started it in 2021. An incredible amount of work has gone into it over the years with tangible growth results that reflect the massive effort of the entire team.
Today, DFlow is integrated almost universally across trading apps on Solana. Transactions constructed by DFlow are in almost every block produced on Solana. We’ve served millions of traders across the world. We’re constantly on the frontier of the technology powering markets, working with the largest and most cutting edge businesses in next-gen finance like Phantom, Kalshi, Coinbase, and hundreds more.
I’m proud of what we’ve built so far and excited to work with Ivan and the MoonPay team on the next phase of insane growth.
Why We Ship: @DFlow
From JPMorgan's floor to building DFlow in Miami. @DanJablonski is working on opening capital markets for everyone. The 2008 crisis showed him why it matters.
The hard part of building a Dune dashboard was never the SQL. It's understanding the dataset underneath.
Most broken dashboards on Dune aren't broken because someone wrote bad SQL —> they're broken because the person didn't understand what the data actually represents. A protocol upgrades its program, the account structure changes, a new instruction gets added etc..and the query that worked last month is now silently returning wrong numbers. No errors, no warnings. Just wrong data that looks right.
AI can write the SELECT statement. It can't tell you that the table you're querying stopped being relevant two weeks ago because the protocol shipped a v2. That's the part that actually matters, and that's the part most dashboards get wrong.
AI is moving at an insane pace and the gap is closing fast. But as of today Claude + Dune <> instant analyst.
Say hello to the newest way to trade @Kalshi on @Solana.
Prediction Markets are now live on the DFlow frontend.
Choose a market.
Take a position.
Track it in real time.
Predict politics, sports, crypto, culture — everything.
All from your favorite onchain trading UI.
@moonshiesty Great read.
"we could compute the bid-ask spread" —>> that might be tricky since most blocks only have trades in one direction from what I've seen.
Say hello to DFlow Proof 🔍
A new way for users to verify real-world identity on @Solana, without requiring applications to build or maintain KYC infrastructure.
The end goal is seamless onchain identity.
Learn more and get started below 👇
Two OG giants of Solana DeFi are teaming up once again
DFlow is now live in Orca for swaps 🐳
@Orca_so is Solana’s leading liquidity layer and DEX for fast, efficient trading
Orca improves capital efficiency and liquidity depth while offering swaps, near-instant execution, permissionless pools, and a seamless UI