← All guides

Exchange proof of reserves explained: what it proves, and what it cannot

Market basics · 9 min read

Financial disclaimer, not financial advice. This article is general information and education only. It does not recommend buying, selling, or holding any crypto asset, fund, or product, and it is not personalised financial, tax, or legal advice. Crypto assets are volatile and speculative and can lose all of their value. Consider your own circumstances and, where appropriate, speak with a qualified professional in your jurisdiction before making any financial decision.

When an exchange holds your crypto, a simple question sits underneath every other one: if every customer asked for their assets back at once, could the exchange return them in full? Proof of reserves is the industry's attempt to answer that question with verifiable evidence instead of reassurance. It became a standard expectation after the FTX collapse in late 2022 revealed an estimated eight-billion-dollar gap between claimed and actual reserves. But “proof of reserves” names a family of reports of very different strength, and the phrase on a webpage guarantees none of them. What matters is which halves of the proof a given report actually contains.

Half one: proof of assets

The asset side asks whether the exchange controls the crypto it claims to hold. Because holdings sit on public blockchains, this half can be shown directly: the exchange publishes its wallet addresses, and it proves control of them cryptographically, typically by signing a message with the private keys for those addresses. A signature is stronger than a screenshot of a balance because only the key holder can produce it. The published wallets can then be inspected by anyone, and their balances added up. That establishes a floor: at the moment of signing, those assets existed and that exchange held the keys. It establishes nothing yet about whether the assets are enough, enough is a comparison, and a comparison needs the other side of the ledger.

Half two: proof of liabilities, and the Merkle tree

The liability side is the total the exchange owes its customers, the sum of every customer balance. Publishing that list directly would expose every customer's holdings, so the standard construction is a Merkle tree. The exchange captures an anonymised snapshot of all user balances at a specific moment, hashes each balance, and combines the hashes upward in pairs until a single hash remains: the Merkle root. That one fingerprint represents the total owed, and changing any single balance in the tree would change the root, which makes the published root a tamper-evident commitment to the whole liability figure.

In its strong form, then, proof of reserves proves two things: that the exchange controls a certain amount of crypto in its wallets, and that this amount is greater than or equal to everything it owes customers. The summary number is the reserve ratio, verified assets divided by total liabilities, which should sit at or above 100%. Some implementations strengthen the liability half with a zero-knowledge proof (a zk-SNARK), which can prove that no balances in the tree were negative or omitted and that the total is honest, all without exposing individual figures. Where an independent third party performs or attests the reconciliation, collecting the wallet addresses, confirming the signatures, and comparing signed reserves against the Merkle root, the result is an audited proof rather than a self-published one. The difference between those two is not cosmetic.

Checking your own inclusion

The Merkle construction has a property most financial statements lack: each customer can personally verify that their own balance was counted. The exchange provides each user with a Merkle proof, a small cryptographic path of hashes connecting that user's balance up through the tree to the published root. Recomputing the path either reproduces the root or it does not. On an exchange that offers this check, the procedure is short:

  1. Find the current report and its snapshot date. A proof describes one moment; note which moment, and how the exchange says the report was produced.
  2. Retrieve your inclusion record for that snapshot. The exchange provides your leaf entry and its Merkle path for the report in question.
  3. Run the verification the exchange provides (an on-page tool or a published verification method) and confirm your balance hashes up to the same Merkle root the exchange published for that report.
  4. Confirm the report covers both halves. Inclusion in the liabilities tree only matters if the same report also proves assets, signed wallet control, at or above the liability total.

What success looks like: your own balance verifiably sits inside a published liability total, and a signed, dated asset total in the same report meets or exceeds it. If an exchange publishes wallet balances but offers no liability tree and no inclusion check, what it has published is half a proof, the half that was never in doubt in most exchange failures.

What the proof cannot show

Blind spotWhy the proof misses itWhat it means for a reader
TimingThe proof is a snapshot at a single momentAssets could be borrowed briefly to pass the check and returned afterwards; frequency and recency of reports matter
EncumbranceControl of a wallet says nothing about claims against its contentsProved assets could be secretly borrowed or pledged as collateral elsewhere
Off-chain obligationsThe tree covers customer balances on the platformFiat holdings, loans, and off-balance-sheet liabilities sit outside the proof entirely
Exclusive controlA signature proves a key was used, not that no copy of it existsKeys could in theory be duplicated or shared; the proof cannot rule it out

There is a fifth gap that follows from the others: the proof says nothing about the days between reports. It cannot capture hacking, seizure, or management decisions taken after the snapshot. That is why the honest summary of proof of reserves is narrow: it is a meaningful, checkable answer to one question, assets versus customer liabilities, at one moment, for the assets it covers, and treating it as a general guarantee of safety asks it to prove things its design never touches.

A five-point report check (the decision framework)

  1. Both halves present? Signed proof of assets and a Merkle-tree proof of liabilities. Assets alone is not a solvency proof.
  2. Reserve ratio at or above 100%? Verified assets divided by total liabilities, stated for the same snapshot.
  3. Dated and recurring? A snapshot date on the report, and a publication cadence, one old report proves one old moment.
  4. Independently attested? A named third party reconciling the two halves, rather than the exchange attesting to itself.
  5. Can you verify your own inclusion? A Merkle proof for your balance that resolves to the published root is the strongest single signal the system offers an individual customer.

What success looks like: all five questions answered from the report itself, with dates. A report that answers only the first half of question one has told you the exchange held some crypto once, which was never the question.

FAQ

Is proof of reserves the same as an audit? Not exactly. A full proof combines cryptographic evidence any user can check with, in stronger implementations, an independent third party attesting the reconciliation. A self-published graphic carries neither property, whatever it is labelled.

Does a 100% reserve ratio guarantee my funds are safe? No. It states that verified assets met or exceeded customer liabilities at the snapshot. It says nothing about encumbered assets, off-chain obligations, or anything that happens after that moment.

Why do some exchanges only publish wallet addresses? Because the asset half is the easy half: the wallets are already public and signing is cheap. The liability half requires exposing a complete, checkable accounting of what is owed, which is precisely the half that failed customers needed in past collapses.

Reminder: Crypto Investing is educational content only, not financial advice. Crypto assets are volatile and can lose all value. Product terms, fees, and rules change, read the current prospectus, terms, or manufacturer documentation for any product you are considering.