Uniswap is a place to trade one token for another, and there is nobody on the other side of the trade.

That sounds like a contradiction. On a normal exchange, when you sell, somebody buys. There is an order book, a list of everyone willing to buy and everyone willing to sell, and the exchange matches you to one of them. If nobody wants what you are selling at the moment you want to sell it, you wait.

Uniswap works a different way. People deposit pairs of tokens into a pool, and the pool trades with you automatically at a price set by a formula. The formula only knows two things: how much of each token is in the pool. Buy one of them and it gets scarcer, so the price it quotes goes up. Sell it back and the price comes down. Nobody approves the trade and nobody has to be awake.

The people who put the tokens in are called liquidity providers, and they are not doing it as a favour. Every trade pays a fee, and the fee goes to them. That is the whole arrangement: depositors supply the inventory, traders pay to use it, and the software does the matching.

It runs on Ethereum and on a number of other blockchains. The core contracts cannot be changed once they are deployed, and the code is published openly, which means anyone can copy it. Many people have.

Here the honest answer is uncomfortable, and Uniswap's own documentation gives it plainly.

UNI is the governance token. Holding it lets you vote on proposals about how the protocol is run: what fees are charged, what gets built, what the treasury spends. That is a real function and it is not nothing.

What it does not do is pay you.

The documentation asks the question directly and answers it in one word. Do UNI holders earn a share of protocol fees? No. Protocol fees are only used to permanently burn UNI, and holders have no individual or pro-rata claim on protocol revenue and there are no direct distributions of any kind.

So the value a holder gets is indirect. The protocol takes a cut of trading fees, uses it to buy UNI off the market and destroy it, and everyone still holding owns a slightly larger slice of a slightly smaller pile. If the burning is big enough and keeps going, the price should reflect it. That is the argument, and it is a weaker claim than a dividend, which is owed to you by right rather than arriving through the market.

That burn started on 28 December 2025. It is nine months old.

Uniswap earned $913,650,840 in trading fees over the year to 15 September 2026.

Every day of that year is measured. The record runs unbroken from 3 November 2018, which is the week Uniswap launched, and there are no gaps in it. That matters more than it sounds. For most projects a fee figure has to be estimated from a partial window, and an estimate can be stretched to flatter. This one is a total.

The direction is the more interesting number.

Monthly fees bottomed at $42 million in March 2026. Then they went up: $96 million in July, $104 million in August, and $117 million in the first fifteen days of September alone. At that pace September would close near $230 million, which would put it within reach of the highest month Uniswap has ever had, the $251 million it earned in May 2021 at the top of the last cycle.

A business earning close to its record five years later, after a full collapse in between, is a different kind of thing from one that peaked and faded. The trailing-year figure of $913 million understates where it is right now, which is why the pace measured over the last ninety days works out to about $1.35 billion a year.

Trading fees are set per pool. Most charge 0.05%, 0.30% or 1.00% of the trade, and the newest version lets whoever creates a pool set anything at all, including zero.

On the version where the split is published, 83% of that fee goes to the liquidity providers who supplied the tokens and 17% goes to burning UNI. Before the fee switch was turned on, the providers got everything.

So the order is: traders pay, depositors take most of it, and a slice goes to destroying the token. $46,463,740 has gone to the burn since it started. Against $913 million of fees in the same rough window, the burn is the small end of the split, and it is the only end a UNI holder touches.

There is also a contract called the FirePit. Anyone can burn 4,000 UNI through it to claim the fees that have accumulated, and governance can change that threshold. It is a mechanism for someone to do the work of collecting, not a payment to holders.

One thing to hold on to: several other flows also fund the burn, including 85% of the sequencer revenue from Uniswap's own chain. None of those are measured anywhere public, so nobody outside knows how large they are. Unmeasured is not the same as zero, and it is also not the same as evidence.

UNI's market value is $3,962,209,584, at $6.36 a token on 16 September 2026.

Set against fees running at $1.35 billion a year, that is a price of about 2.9 times the annual takings. Measured against the year actually completed, $913 million, it is about 4.3 times.

Put it in ordinary terms: buying the whole of Uniswap at today's price costs roughly three to four years of what passes through it in fees. For comparison, a wireless network project once traded at about 87,000 times its annual takings, and the most expensive thing assessed here trades at about 4,600 times. Three is not in the same conversation.

That is the lowest ratio of price to earnings anything on this site has produced, by a wide margin.

The catch is in the sentence above it. Those fees are not Uniswap's income in the way a company's revenue is its income. Most of that money is paid straight out to the people who supplied the tokens. The part the protocol keeps is $46 million, not $913 million, and that part goes to a burn rather than to you. The multiple is genuinely low, and it is low against a number that mostly belongs to somebody else.

Uniswap is routinely described as the largest decentralised exchange. The interesting part is that its lead has not been steady.

In August 2025, Uniswap handled about 25% of all the trading volume going through decentralised exchanges. By January 2026 that had fallen to 17%. Competitors, many of them running copies of Uniswap's own openly published code, were taking business.

Then it came back. 19% in June, 27% in July, 23% in August, and 28% so far in September.

The code being copyable is the central risk to this business and always has been. Anyone can run the same formula. What they cannot easily copy is the depth of the pools, which is what determines whether a large trade moves the price against you, and the habit of every other application routing its trades through Uniswap by default. The eighteen months above suggest both of those are real and neither is permanent.

There is one number that has not recovered, and it is a big one.

The value sitting in Uniswap's pools peaked at $10.24 billion on 12 May 2021. Today it is $3.52 billion. That is a fall of about two thirds, and five years later there has been no return to the old level.

Fees have come back. Deposits have not. Those two things can diverge because fees follow trading activity while deposits follow how attractive it is to supply the tokens in the first place, and supplying has become harder and more competitive work than it was in 2021.

For a holder it cuts both ways. Less capital in the pools means less depth to defend the position with. But earning near-record fees on a third of the deposits means the capital that is there is working much harder than it used to.

A token's release schedule is where a project can quietly take value from the people already holding it. Issue more, hand them out, and everyone who already owned some owns a smaller share without anyone announcing anything.

Uniswap's schedule is done.

A billion UNI was created in September 2020 and the whole allocation has now vested: 43% to the community treasury, 15% given away to people who had already used the protocol, 21.3% to the team and future employees, 18% to investors, 0.7% to advisors. Nothing is pending. There are no cliffs left to fall.

Since December 2025, the supply has been shrinking rather than growing, because of the burn. About 892.9 million UNI were in circulation as of July 2026, down from the billion, and 100 million were destroyed from the treasury in one go.

Governance does keep the standing authority to mint up to 2% a year. It has never used it. The older 2020 documentation describes that 2% as a perpetual inflation rate starting after four years, and the current documentation describes it as an unexercised power. Both statements are still published, and they do not agree with each other.

The contracts cannot be changed. The organisation around them can.

UNI holders vote, but the voting has a history worth knowing. In 2026 the DAO voted to take back roughly $42 million of UNI it had loaned to delegates, on the stated grounds that the tokens had served their purpose. That vote happened while the project was dealing with criticism that its governance was less decentralised than it appears. Independent write-ups of Uniswap's governance describe it as concentrated, and that criticism is old rather than new.

The regulatory chapter is closed. The SEC issued a Wells Notice to Uniswap Labs on 10 April 2024, which is the formal warning that an enforcement action may follow. On 25 February 2025 the investigation was closed with no action. Separately, a court ruled in Risley v. Uniswap that the authors of automated smart contracts are not liable under federal securities law for fraud committed by third parties using them, which is a broader result than one company winning one case.

There is one security event in the record and it is widely misreported. In April 2020, an attacker used a reentrancy flaw in a particular token standard to drain funds from two protocols at once. The headline figure of $25 million was almost entirely a loss at Lendf.Me, which lost $24.5 million. Uniswap's share was between $300,000 and $1.1 million.

The code itself has been audited repeatedly: formal verification and review by six engineers for version 2 in 2020, ABDK Consulting and Trail of Bits for version 3 in 2021, and published reports for version 4.

One test in this assessment returns a Fail, and it is a disagreement about strategy rather than a finding about conduct.

The argument is that exchanges specialise by what they trade. Stablecoins want a design built for deep liquidity in things that barely move. Newly issued assets want something coupled to the issuance. Heavily traded assets want an order book. On that reading, a venue that tries to serve everything with one mechanism loses each category to somebody who built for it.

Uniswap is deliberately one mechanism for everything. Its newest version adds hooks that let a pool customise its behaviour, but the customisation is available to every asset type equally rather than designed per type.

Whether that is a weakness is a genuine question, and not one the numbers settle. If you think the specialists win each niche, Uniswap is structurally exposed. If you think the default venue with the deepest pools wins regardless, it is not. Nothing else in this assessment found Uniswap extractive, dishonest or misrepresented.