How MEV Bots Make Money on Ethereum (and Who Keeps the Profit)

People ask me this all the time, usually right after seeing a screenshot of some bot making a crazy profit in a single transaction. "How does it make money? Where does that money come from?"

The short answer is that MEV bots make money from ordering. Every block is a list of transactions, and whoever decides the order can capture value: buying before a price moves, selling right after someone else moves it, or being the first to grab a reward a protocol pays out. Bots, which the industry calls searchers, compete to get their transactions into the right spot.

The longer answer is more interesting, and the part almost nobody tells you is how little of that money the bot actually keeps. Let's go through both.

Where the money comes from

DEX arbitrage

The classic. The same token trades at slightly different prices on two pools, say Uniswap and SushiSwap, or two Uniswap pools with different fee tiers. A bot buys on the cheap one and sells on the expensive one in the same transaction. If the prices don't line up enough to pay for gas, the transaction reverts in simulation and the bot never sends it.

Where the price difference comes from: other people's trades. Every large swap moves one pool's price and leaves the others behind. The arbitrage bot brings them back in line. That's actually useful, it's how DEX prices stay consistent.

The problem is competition. Simple two-pool arbitrage on major pairs is one of the most crowded games in crypto. I explained why, and what to do instead, in how to build a profitable MEV arbitrage bot. The short version: go for multi-hop, long-tail routes where fewer bots are looking.

Backrunning

A backrun is arbitrage aimed at a specific transaction. A bot sees a big swap pending, knows it'll move a pool's price, and places its own arbitrage right after it in the same block. The bot doesn't touch the victim's trade at all. It just cleans up the price difference the trade leaves behind.

Backruns are also what powers the user refunds you see in Flashbots MEV-Share and MEV Blocker. Users send their transactions privately, bots bid to backrun them, and part of the bid goes back to the user.

Liquidations

Lending protocols like Aave and Compound need someone to close loans when the collateral is no longer worth enough. They pay a liquidation bonus to whoever does it: you repay part of the debt and receive the collateral at a discount. Bots monitor every borrower's position and race to be first when one goes underwater. It's a whole topic on its own, so I wrote a separate post on how liquidation bots work.

Sandwiching

This is the one people hate, and they have a reason to. A bot sees a pending swap with loose slippage, buys right before it to push the price up, lets the victim buy at the worse price, then sells right after. The victim pays the difference.

There's a technical twist that matters for anyone thinking of doing this. A sandwich isn't atomic: the bot's first trade and its last trade are separate transactions around someone else's. If anything goes wrong in between, the bot holds the bag. The ethereum.org MEV page mentions "salmonella" tokens, which were built specifically to punish sandwich bots that didn't check what they were buying. I explain how users avoid this in what is a sandwich attack.

Sniping

Buying a new token in the same block that trading opens. It's not MEV in the textbook sense of reordering other people's trades, but it uses the same tooling: mempool monitoring, simulation and bundles. I covered it in how Ethereum sniper bots work.

CEX-DEX arbitrage

When the price on a centralized exchange moves, on-chain pools are suddenly stale. Bots trade on-chain to bring the pool in line and hedge on the exchange. This is where a lot of volume is today, and it's dominated by firms with serious capital, exchange accounts and low-latency setups. Not really a solo game.

The long tail

Everything else: weird tokens, small pools, NFT mispricings, protocol-specific opportunities like oracle updates, auctions, rebases or reward claims. Each one is small, but nobody else is looking. That's where I recommend solo developers start, because you can actually win.

So who keeps the profit?

Here's the part the screenshots don't show you.

Since the Merge, most Ethereum blocks are built by specialized block builders and delivered to validators through MEV-Boost. Searchers don't get into blocks by paying high gas. They send bundles to builders with a payment attached, and builders take the bundles that pay the most.

So when ten bots spot the same 1 ETH arbitrage, what happens? They bid against each other. One bids 0.5 ETH to the builder. Another bids 0.7. Another bids 0.9. The one willing to keep the least wins. The ethereum.org MEV docs note that for competitive arbitrage, searchers may end up paying 90% or even more of their total MEV to the validator.

That's why:

A big profit on-chain doesn't mean a big profit for the bot. Often the builder and validator kept most of it.
The real competition is about margin. If your code runs a bit faster or uses less gas, you can bid higher and still make money.
Order flow is worth a lot. If you see transactions nobody else sees, you don't have to bid against anyone. That's why wallets and apps sell or route their order flow, and why refund programs exist.

The costs nobody puts in the screenshot

Even on the trades a bot wins, there are costs:

Infrastructure. Your own node (or several), low-latency connections to builders, and servers close to them. Public RPCs are too slow for competitive strategies.
Gas on failed attempts. With bundles, failed transactions mostly don't land, which is the main reason everyone uses them. But anything you send publicly can still fail and cost gas.
Inventory risk. Non-atomic strategies like CEX-DEX or sandwiching can leave you holding tokens if the market moves.
Contract bugs. Your bot contract holds or moves funds. A bug, or a token designed to exploit bots, can wipe out months of profit in one block.
Development time. The biggest one. Strategies decay. What made money six months ago probably doesn't anymore, because others found it too.

Is it still possible to make money?

Yes, but not the way YouTube tells you. Those "deploy this MEV bot contract in Remix and fund it with ETH" videos are scams, by the way. The contract just sends your ETH to the scammer, as SentinelOne documented in detail.

From what I've seen, the bots that make money today usually have at least one of these:

A niche. A protocol, chain or token type that the big searchers ignore.
Better information. Private order flow, or faster data than everyone else.
Better execution. Cheaper gas, faster simulation, smarter bidding.
Combined strategies. One transaction that does an arbitrage and a liquidation, or a backrun with several hops, which simpler bots can't compete with.

If you want to learn this properly, the best way is to pick one strategy, study historical examples on a block explorer, and test your bot on a mainnet fork before you put real money in. If you'd rather start with tools that already work, MEVDAO gives you the Sniper Bot, Premium Sniping Group and Token Rescue Bot in one subscription.

If you have a specific strategy in mind and want a custom bot built, message me on Telegram. And join the MEV Secrets email list right below for weekly tips.