Exchange Data Says Binance Has 18% of the Market. That Number Is Not Usable.

Across the 61 centralized exchanges reporting spot volume to DefiLlama, $33.3 billion changed hands in twenty-four hours.¹ Binance accounted for 17.9% of it. The Herfindahl–Hirschman index across those venues reads 641 — well below the 1,000 that a ten-way even split would score. Pulled 4 August 2026.

Read literally, that says centralized crypto trading is less concentrated than ten identical exchanges splitting the market evenly. We do not believe it, and we are not going to publish it as a finding just because the arithmetic runs.

Why the number fails

Look at the ranking rather than the index. Second and third place go to BTCC at 10.5% and Pionex at 7.5%. BitMart is fourth at 4.0%. Coinbase ranks twelfth, at 2.8%¹ — behind BitMart, WEEX, LBank, Bybit, CoinW, OKX, Toobit and Gate. Coinbase is a listed company that reports audited transaction revenue to the SEC each quarter. Any dataset that places it twelfth by spot volume is measuring something other than what it claims to measure.

We are not alleging misreporting by any venue named here. The point is narrower and worse: this dataset is assembled from figures exchanges report about themselves, there is no independent settlement layer to check them against, and so it cannot distinguish a large venue from a venue that says it is large. A concentration statistic built on top of that inherits the problem and hides it behind a decimal point.

The denominator is incomplete too. Thirty of the ninety-one exchanges tracked report no spot volume at all, Robinhood among them. Every share above is therefore a share of the venues that chose to answer.

What we are publishing instead

No concentration figure for centralized exchange trading. This is the second measurement in our concentration series and the first that has failed, and the failure is the result. Stablecoin issuance could be measured because supply settles on a public ledger and can be counted by anyone. Exchange volume cannot, because it does not settle anywhere observable. That distinction is the whole difference between a number and a claim.

What would change this

A volume dataset with a published cleaning methodology and a stated rejection rate would let us try again. So would order-book or trade-print data from the venues themselves, audited or otherwise attested. Absent either, the honest position is that the question is open, and anyone quoting exchange market share to one decimal place — including us, in an earlier draft of this piece — is quoting a self-assessment.

Next in this series: on-chain trading volume, which can be counted, and what its share tells us given the denominator problem described above.

---

How exchanges actually work

This note is a measurement and its result. The mechanics behind it — how a venue is built, what it really charges once spread and withdrawal fees are counted, what a proof-of-reserves attestation does and does not prove, what the terms of service say about your assets in an insolvency, and what to check before trusting a venue — are in our standing reference, Crypto Exchanges Explained.

Revision history

• 6 August 2026. Trimmed back to the measurement and its method. The evergreen material added on 6 August was moved to the exchanges reference, which this note now links up to. The finding and its figures are unchanged.
• 4 August 2026. First published. Concentration measurement attempted against the DefiLlama centralized exchange endpoint and withdrawn as unusable.