What is proof of reserves, and what can't it prove?
Proof of reserves shows the coins an exchange holds on one date. How the Merkle tree check works, what it leaves out, and how to read a report.
In November 2022, the exchange FTX halted customer withdrawals and filed for bankruptcy in the United States within days. Its customers learnt that the balances shown in their accounts were not all backed by coins. Within weeks, several of the largest remaining exchanges began publishing what they called proof of reserves.
Those reports set out to answer one question: does the exchange hold the coins its customers think it holds? This piece covers how proof of reserves works, the gaps it leaves, and what to look for in one.
What is proof of reserves?
Proof of reserves is a report in which a crypto exchange, or custodian shows that it controls enough on-chain assets to cover the balances its customers hold with it. A full report answers two questions. What does the exchange have? What does it owe the people who keep their money there? A report that answers only the first is like someone showing you their savings but not their loans: you see the money, not whether it covers what they owe.
How does an exchange show what it holds?
An exchange shows its holdings by publishing the blockchain addresses of its wallets, and proving it controls them. Since blockchains are public, anyone can read those balances on a block explorer (opens in a new tab). Proving control usually means signing a message with each wallet's private key or moving a small, pre-agreed amount from it. Without that step, an exchange could point at someone else's wallet, and count the coins as its own.
How does an exchange show what it owes?
An exchange usually shows what it owes with a Merkle tree: a structure that combines every customer balance into a single total, and a single fingerprint, called the root. Each customer can check that their own balance was included without seeing anyone else's account. The more customers check, the harder it becomes to leave balances out unnoticed.
The plain Merkle tree has known weak spots. An exchange could pad it with negative balances that shrink the reported total, and the tree can leak hints about other accounts. Ethereum co-founder Vitalik Buterin described a fix in a November 2022 post: wrap the tree in a zero-knowledge proof that shows every balance is non-negative, and that the balances add up to the stated total, without revealing any single account.
What can proof of reserves not show?
Proof of reserves cannot show that an exchange is solvent unless it counts liabilities as carefully as assets. Even a report that answers both questions still has four blind spots.
First, it is a snapshot. The report holds true on one date, and coins could be borrowed for that day and returned afterwards. Second, it sets on-chain assets against customer balances only. Loans the exchange has taken, and other debts outside the customer tree may not appear at all. Third, it rarely covers every asset: a report may include bitcoin and ether, but leave out smaller tokens that customers also hold.
Fourth, it does not show who would have the stronger claim on those coins if the exchange failed. That turns on the platform's terms of use, and the law where it is based, as the Celsius bankruptcy showed. Our note on crypto lending (opens in a new tab) covers that case.
Why did an accounting firm stop producing these reports?
In December 2022, the accounting firm Mazars paused its proof-of-reserves work for crypto firms. Its clients had included Binance, KuCoin, and Crypto.com. A Mazars spokesperson told Decrypt, the pause reflected concerns about the way these reports are understood by the public.
The concern makes sense once the format is clear. Most proof-of-reserves reports are attestations: an accountant checks specific items on one date, and reports what it found. An audit goes further and forms an opinion on a company's whole financial position. Our guide to reading a reserve attestation (opens in a new tab) sets out that difference for stablecoins.
What should a reader look for in a report?
Five questions carry most of the signal. Does the report cover liabilities as well as assets? Who prepared it, and is that firm independent of the exchange? What date does it cover, and how often is it repeated? Which assets does it include and which does it leave out? Can you check your own balance in the tree? A report that answers all five is easier to read. It is still a dated snapshot rather than a guarantee.
Where Northtape fits
Northtape runs no exchange, and holds no customer funds. It is a news desk. Stories about proof-of-reserves reports, and frozen withdrawals are detected by the Risk Radar counterparty lens.
FAQs
Does proof of reserves mean an exchange is solvent? No. It shows assets the exchange controlled on one date. Solvency also needs a full count of what it owes, including debts outside customer balances.
What is a Merkle tree in proof of reserves? A structure that combines every customer balance into one total and one fingerprint. Each customer can check their own balance was included without seeing anyone else's.
How often should an exchange publish proof of reserves? No universal rule applies. More frequent reports leave shorter gaps between snapshots, and a long silence since the last one is worth noticing.
Can I check a proof-of-reserves report myself? Partly. You can check your own balance in the tree and read the published wallet balances on a block explorer. You cannot see the exchange's debts outside the platform.
None of this is investment advice. It describes how proof of reserves works, not whether any exchange, custodian, or report is a safe basis for holding funds.
