Uniswap

The contracts have never been drained. In seven years and across 48 chains, with billions in liquidity sitting in them continuously, there has been no core exploit — which among DeFi protocols of this size is close to unique, and it is the single strongest thing that can be said about the protocol.

The risk sits somewhere else, and it is where people actually lose money: phishing sites, malicious token approvals and fake front-ends. None of these break Uniswap. They persuade you to sign something, and a signature you authorised cannot be reversed by anyone. Reaching the interface through a bookmark rather than a search result, and reviewing what an approval grants before signing it, removes most of this exposure.

The second real risk is the token you are swapping into rather than the venue. Anyone can list anything on a permissionless exchange, which is the point of it and also means the pool you are buying from may hold a contract designed to stop you selling. Uniswap does not vet listings and does not pretend to.

What you do not carry is custody risk. There is no account, no balance held by a company and nothing to freeze — you trade from your own wallet and the funds never leave it except into the trade. That is a different risk model from an exchange, not a safer version of the same one: nobody can seize your funds, and nobody can help you recover them.

Traders paid 0.1762% on average across thirty days, not the 0.30% that gets quoted. We computed it the only way that means anything: $87,995,089 of protocol fees divided by $49,950,245,141 of volume. Every DEX comparison in existence quotes the fee tier, and the tier is a number almost nobody pays, because volume concentrates in the cheaper pools.

Since December 2025 that fee is split. The UNIfication proposal turned on the protocol fee, so v2 liquidity providers now take 0.25% where they previously took 0.30%, with 0.05% going to the protocol. Selected v3 pools route a share depending on their tier, and Governance Proposal 100 in July 2026 extended the same arrangement to v4 pools across seven networks. The protocol now captures roughly one-sixth of fees.

Gas is the part the fee rate does not include and often the part that dominates. A swap on Ethereum mainnet can cost more in gas than in protocol fees at small sizes, which is why the same trade on an L2 deployment is a different proposition entirely. Compare the total, not the percentage.

For large orders the constraint is depth rather than fee. $3.46bn of liquidity turning over 14.4 times in thirty days is a hard-working book, and a trade big enough to move a specific pool pays the price impact regardless of what the tier says.

UNI holders waited years for the fee switch. It arrived in December 2025, the vote was effectively unanimous at 125 million to 742, roughly $23.15m of protocol revenue has accumulated since, and over 100 million UNI has been burned. The token then made a new cycle low.

That sequence is worth sitting with, because it is the clearest natural experiment available on a claim the whole sector makes. Burning supply does not create demand. It removes tokens from a float, and if nobody is bidding for what remains, the price does what it was going to do anyway. We reached the same conclusion looking at an exchange token that retired around 40% of its supply to no visible effect; Uniswap is the second independent case, and a cleaner one, because the revenue behind the burn is real and verifiable on-chain.

The design choice underneath it is deliberate and defensible. Revenue buys and burns UNI rather than being distributed to holders, because a direct payout would make the token look like a dividend-bearing instrument, with the securities-law consequences that follow. Uniswap chose the structure that keeps the token furthest from that characterisation.

What the fee switch genuinely changed is that the protocol has revenue and a mechanism connecting it to the token. That is a real improvement on governance rights alone. What it did not do, and could not do, is manufacture buyers.

All three versions run simultaneously and hold liquidity at the same time. v2 is the original constant-product design, simple and still carrying real volume. v3 introduced concentrated liquidity, letting providers place capital in a price range rather than across the whole curve, which is why capital efficiency jumped and why providing liquidity became an active job rather than a passive one.

v4, live since 2025, is the largest architectural change since the protocol started. Its core idea is hooks — plugins that run at defined moments in a pool’s lifecycle, letting a developer attach custom logic such as dynamic fees or on-chain limit orders to a pool without forking the protocol. It turns Uniswap from an exchange into something closer to a platform others build exchanges on.

Hooks also widen the attack surface in a specific way worth understanding: the core contracts remain unexploited, and a hook is third-party code you are trusting when you trade through a pool that uses one. The safety record described above belongs to Uniswap’s contracts, not automatically to everything now attached to them.

Unichain, Uniswap’s own L2, matters for a reason unrelated to trading: its net sequencer fees route into the UNI burn, so its volume is a direct input to the token’s deflation. Which version and which chain you are on changes your fee, your gas and, with hooks, who wrote the code executing your trade.