Everyone reads 100.25% as "fully collateralized." Comforting number. Binance publishes its Proof of Reserves snapshot in the weeks after FTX collapses. BTC and ETH backed at 100.25%. Crypto Briefing runs the story: strong reserves enhance exchange trust, stabilize markets, encourage broader adoption. The narrative closes. Headline published. Market breathes.
The arithmetic says something else. 100.25% means 0.25% of cushion between gross assets and reported customer liabilities. That isn't a shield. It's a paint job. A single liquidation cascade that moves BTC by more than a quarter of one percent puts the paper ratio below 100. Code doesn't lie. People do. And numbers engineered for optics usually tell you more about the marketing department than the balance sheet.
I learned this the expensive way. In 2020, as a junior at UT Austin, I spent twelve hours manually auditing the Uniswap V2 factory contract and found an integer overflow in the liquidity-minting logic that every automated scanner missed. Two thousand dollars in bounty. One permanent lesson: official verification documents are surface-level. Dig into the mechanism yourself, or you don't understand the risk.
Let's dig.
Context: A Balance Sheet With One Side Missing
December 2022. FTX is a crater. The timeline reads like an engineering failure postmortem. November 6: CZ announces Binance will liquidate its FTT holdings. November 7: FTX faces $5 billion in withdrawals in a single day. November 8: Binance signs a non-binding letter of intent to acquire FTX — then walks away. November 11: FTX files Chapter 11. The centralized exchange model just proved it can vaporize customer assets while posting cheerful tweets. Billions missing. No audit caught it. No reserve report existed.
Bitcoin hovers near $16,000 to $17,000. The fear index sits in terror territory. Trust in every CEX is zero. Into this vacuum, Binance releases a Proof of Reserves report.
Two months earlier, at the height of the bull market, a report like this would have been a footnote. In December 2022, it was a lifeline. Depositors wanted to know one thing: if I click withdraw, does the exchange actually have my coins? PoR attempts to answer that question. That is why the announcement traveled through every crypto media outlet within hours, and why Crypto Briefing framed it as a market-stabilizing event.
PoR is not new technology. Kraken pioneered the approach in 2014. BitMEX ran a version in 2020. The mechanism is well understood: the exchange collects every user's balance, hashes it, builds a Merkle tree, publishes the root. Anyone can verify their own branch is included in the aggregate. The cryptographic claim is narrow: "these addresses under exchange control hold at least this many tokens."
That's it. That's the entire claim.
The phrase "under exchange control" does all the heavy lifting. It says nothing about whether those assets are pledged as collateral, lent to a market maker, sitting in a derivatives book as margin, or backing a stablecoin the exchange issues. It says nothing about liabilities. The balance sheet has one side missing.
This distinction — asset existence versus solvency — is the entire ballgame. The market keeps confusing the two.
The demand for proof was existential. After FTX, depositors understood, correctly, that their balances were unsecured claims against a private company. The only reasons to hold assets on an exchange were convenience: cheaper fees, faster execution, access to derivatives. FTX demonstrated the counterparty risk embedded in that convenience. A PoR report is the exchange's answer to "we can show you the assets." It is not the answer to the harder question: "can you prove you don't owe more than you own?" That question requires a liability ledger. None of the major exchanges published one.
Look at the competitive landscape and the timing sharpens. FTX made "audited" the most valuable word in crypto. Every exchange scrambled for a badge. OKX published its PoR. Crypto.com released one. Coinbase pointed at its public-company filings. Binance, the largest of all, needed the loudest statement. A Merkle-tree PoR was the cheapest rigorous-looking option available. That makes the document a crisis-communication product. Understanding that changes how you read the number.
Core: What Proof of Reserves Actually Proves
Walk the mechanism from first principles.
A Merkle tree answers one question: is this item a member of this dataset? For PoR, the dataset is "all audited user balances." The item is "my balance." The exchange aggregates customer balances off-chain, runs them through a hash tree, and publishes the root. I can traverse the tree from my leaf to the root and confirm: the exchange counted me. That is real cryptographic verification.
Then the exchange signs a message declaring that certain on-chain addresses constitute the reserve pool. The total value of those addresses is compared to the aggregate of user balances. If assets equal or exceed liabilities, the ratio reads 100% or higher.
Here's the step most commentary skips. The comparison is between two numbers, but only one is cryptographically verified. The asset side is on-chain, immutable, checkable. The liability side is whatever the exchange claims users hold. There is no Merkle proof for the liability ledger. There is a database query, an export, maybe a CSV. The exchange's internal database is the only source of truth for its liabilities, and no external party verifies it. That asymmetry is the entire problem. PoR proves a lower bound on gross assets. It cannot prove the liability number is complete.
The 100.25% margin does the opposite of what it appears to do.
What does that fraction actually buy? Do the arithmetic. If the reported liability ledger is $100, and gross assets are $100.25, the buffer against adverse price moves is 25 cents per hundred dollars. A 1% drop in BTC against the book leaves the ratio at roughly 99.25%, unless other assets absorb the loss. A 5% drop leaves it at 95.25%. In a market where BTC routinely swings 5% intraday, 100.25% is not a solvency buffer. It is a rounding error wearing a solvency badge.
Consider the mechanics under stress. In a liquidation cascade, exchange systems sell collateral into falling markets, pushing prices lower, triggering more liquidations. If the disclosed buffer sits at 0.25%, the first wave of adverse moves consumes the entire cushion. The ratio reads below 100 — precisely the moment depositors check the document. The report built to reassure them becomes the trigger for the run.
The ratio also forces the question: what is in the pool? A reserve of volatile BTC collateralized at 100.25% is meaningfully weaker than a reserve of stablecoins at 100.25%. The ratio alone cannot tell you which one you are looking at. Composition, encumbrance status, maturity profile of lending — none of it appears in the headline number.
I audit the logic, not the hope. The logic chain:
- PoR proves selected addresses hold X tokens.
- PoR does not prove those tokens are unencumbered.
- PoR does not prove reported liabilities are complete.
- PoR does not prove there are no off-balance-sheet liabilities.
- Conclusion: PoR proves gross assets exceed a reported subset of liabilities. It does not prove net equity is positive.
Steps two through four are where every CEX catastrophe historically lives. FTX had billions in gross assets. The problem was never the existence of assets. It was undisclosed liabilities against those assets, customer funds used as collateral for Alameda positions, rehypothecation without disclosure, a house token minted to fill the gap.
What real solvency verification looks like — and why nobody runs it.
The technology exists to do this properly. A proof of solvency covers both sides of the balance sheet: liabilities (customer deposits) and assets (reserve addresses), plus a signed commitment from the custodian that no unrecorded borrowings exist.
The cryptography is harder but solved. To keep individual balances private, the exchange can use homomorphic commitments to sum liabilities without revealing them, then prove in zero knowledge that the committed sum is less than the public asset total. Range proofs ensure no negative balances. The result is a compact proof that net equity is positive, revealing nothing about any individual user. This has been written about for a decade. No major exchange runs it.
Why not? That is the smartest question an investor can ask. If the assets are genuinely there and truly unencumbered, a full solvency proof is engineering work — months of implementation, not a research breakthrough. The fact that the industry settles for half the proof reveals where the friction lives. It is not cryptography. It is commercial reality. Exchanges lend customer assets. They rehypothecate. They run proprietary desks with offsetting positions. A full liability snapshot would expose those practices to customer scrutiny. PoR is the PR solution that lets them avoid that exposure. There is also a convenient technical excuse: the zero-knowledge machinery is computationally heavy, think ZK rollup proving costs without the trading volume to amortize them. It is a real cost. It is not the reason.
Consider the incentives from the exchange's side. The opaqueness is the product. A CEX earns yield on customer assets, lends them to market makers, and uses them for its own treasury operations. Every disclosed liability reduces the room to run that business. Full transparency would force a choice between returning to a pure custody model, which carries lower margins, and admitting the fractional-reserve nature of the exchange model. No CEO is volunteering for that meeting.
The auditor problem compounds the asymmetry.
Mazars, the accounting firm Binance originally worked with on its PoR engagement, paused all work with crypto clients in early 2023 and scrubbed its website of the reports. Read that signal carefully. Either the auditor was never doing audit-depth work, or the auditor realized crypto attestations carry reputation risk they were not compensated for. Both explanations undermine the product.
And here is the deeper issue. A PoR "audit" is typically not an audit at all. It is an attestation, a review, or in the worst cases a screenshot of a balances API. No GAAP. No IFRS. No transaction sampling. No reconciliation of the liability ledger. The PoR standard is whatever the exchange's engineering team decides it is. Coinbase publishes audited financial statements as a public company — a legal obligation. Kraken has run PoR since 2014 with third-party verification. Binance, the largest exchange on earth, published a self-authored report, solicited community verification, and offered bug bounties. The bounties are a nice touch. They do not replace GAAP audits.
What my own audits taught me about claims like this.
My history keeps bringing me back to one conclusion: verify the mechanism, not the claim.
In 2021, I ran a flash-loan arbitrage script between SushiSwap and Uniswap for three weeks. I extracted $14,500 from a slippage-tolerance mismatch in smaller pools. No narrative. No community. No token. Pure mechanism exploitation. The lesson: where there is a gap between story and structure, the person who reads the structure first gets paid. Arbitrage is just patience wearing a speed suit. PoR creates exactly this kind of gap. The story is "we are solvent." The structure is "we disclosed gross assets and concealed the liability calculation." Incomplete information flatters the exchange. The public is systematically overoptimistic about what the report means.
The May 2022 Terra collapse taught me the same lesson from the disaster side. When the depeg hit, I lost 40% of my portfolio because I had chased yield without stress-testing correlation risk. Yield is deferred risk premium became my operating rule. PoR runs the same way in reverse: reassurance is deferred distrust. The trust this document purchases is an asset on the exchange's balance sheet, not on yours.
Late 2023, I allocated $25,000 into EigenLayer restaking, targeting AVS like EigenDA. I manually monitored smart contract interactions to understand slashing conditions. The complexity was higher than advertised. I exited half the position when the incentives got unclear. New tech frequently outpaces its own security model. PoR is not new tech — it is old tech deployed as a trust signal. The same principle applies: if the security model is incomplete, the position size should be small.
In 2025, I audited an AI-driven trading bot claiming 30% monthly returns. Reviewing its API keys and transaction logs, I found it was executing high-frequency, low-margin trades on DEXs while bleeding gas fees. There was no edge. I shorted the associated token. If you cannot verify the mechanism, do not buy the narrative. A PoR report that claims "we hold 100.25% of customer balances" is a claim about a mechanism. The mechanism is partially verifiable. But what most people verify is the headline, not the mechanism. They read "100.25%" and see "safe." They should read "100.25%" and see: asset-side disclosure only, liabilities unaudited, auditor rotating.
The derivatives blind spot.
There is another structural detail hiding in the PoR framing: what counts as a liability. Traditional exchanges count customer deposits as liabilities. Simple. But derivatives exchanges present a subtler picture. When a user holds a perpetual position, the exchange records margin, unrealized PnL, and position risk. A PoR that only covers spot BTC and ETH balances silently ignores the derivatives book. The same addresses that back spot deposits might also back margin requirements. No exchange clarifies whether reserve assets are segregated from proprietary trading collateral.
The regulatory angle makes this more than an academic question. The SEC has repeatedly targeted exchanges over customer asset custody and commingling. The CFTC oversees much of the derivatives market. MiCA in Europe and MAS in Singapore both push for customer asset segregation. PoR as currently practiced does not satisfy any of those standards. It is a voluntary disclosure with no standardized methodology and no regulator-mandated format.
The later history is instructive. Binance ultimately settled with the DOJ, CFTC, and FinCEN in late 2023, paying $4.3 billion over anti-money-laundering and sanctions violations. That settlement turned regulatory licenses into the deepest moat in the industry — a moat no new entrant can afford. The PoR report, initially a transparency experiment, became the first payment on a license to operate. The market treated it as a risk signal. Regulators treated it as a compliance gesture. Both interpretations were generous. The report was never designed to satisfy either.
Contrarian: The False-Security Trade
The real danger of Binance's PoR is not that it is fake. It is that it manufactures false confidence.
Read the market's reaction. The headline "100.25% collateralized" is internalized as "Binance is fully backed." Risk premiums compress. Withdrawal anxiety fades. Depositors behave as if they hold an insured claim. Regulators see a tick-box and move on. Every actor in the chain behaves rationally, and the aggregate result is collective mispricing of tail risk.
Here is the contrarian read: PoR is a risk-repricing tool, not a risk-reduction tool. It does not reduce the probability that a CEX holds undisclosed liabilities. It reduces the market's perceived probability. The gap between those two is where bad things happen.
Algorithms don't panic. They just reprice. When repricing happens — triggered by a withdrawal-delay rumor, a regulatory action, a counterparty default — the market discovers the depth of the information gap simultaneously. That is the anatomy of a bank run. The PoR document provides zero liquidity at that moment. 100.25% on paper does not stop a $5 billion withdrawal day. Binance faced withdrawal pressure in June and December 2022 and survived. The buffer was operational liquidity, not the attestation.
There is also a moral hazard component. The more convincing the PoR appears, the less depositors monitor the exchange's actual health. They skip the withdrawal dry run. They keep balances on the platform for convenience. The exchange's cost of funds stays low. In a system where the true solvency signal is unverifiable, the appearance of verification is the dangerous element. The FTX era did not end because users demanded better proofs. It ended because a house of cards collapsed under its own weight.
Retail reads "solvent." Smart money reads "liability side unverified." Those two populations are processing different documents. The information asymmetry favors the exchange and the sophisticated depositor — the ones who quietly withdraw when governance terms change or the auditor quits.
What actually moves my risk assessment? Five signals.
First, net exchange flows. If BTC and ETH net flows into Binance stay positive, the trust narrative holds. If they turn negative and stay negative, the report is cosmetic.
Second, withdrawal latency under stress. A PoR that precedes a frozen-withdrawal event is not a solvency proof; it is a speed bump in the collapse timeline. FTX had Twitter attestations. Same mechanism: claim, no verification.
Third, funding rates and basis. If the market genuinely believes the report, perpetual funding normalizes. If it stays suppressed, the market is not buying.
Fourth, audit continuity. Binance rotated auditors and switched methodologies. Each rotation is a signal that assurance is cosmetic. Auditors do not churn when the work is solid.
Fifth, and most important: liability disclosure. If any major exchange ever publishes a true zero-knowledge proof of solvency covering both sides of the balance sheet, that is the structural event worth pricing. Everything else is noise.
The FTX lesson was never "exchanges should publish reserve ratios." The FTX lesson was: customer assets should not be commingled with proprietary trading collateral, and no private entity should hold unilateral control over both. PoR does not address either structural flaw. It screens them behind a Merkle root.
Takeaway
The cold read is this. The 100.25% PoR report is a product of crisis communication, not financial engineering. It tells you Binance holds gross assets roughly matching reported customer deposits in BTC and ETH. It does not tell you whether Binance is solvent. It does not tell you whether the same assets back derivatives positions. It does not disclose the liability ledger at all.
Trust the stack, verify the exit. The Merkle tree is sound. The exit is the withdrawal transaction that lands in your wallet. Everything before that is a claim in a system where claims have historically been cheap.
The industry knows what real assurance looks like: a registry of unencumbered assets, a signed liability ledger, a third-party audit with teeth, a zero-knowledge proof of solvency. Until an exchange ships that stack, the rational posture toward every PoR headline — including Binance's — is the same: interesting, verified on the asset side, unresolved on the liability side.
So here is the question worth holding. If the world's largest exchange measures its solvency in a 0.25% buffer, what does "transparent" actually mean? Code doesn't lie. But the people who choose the numbers do. Verify the ones you can't see.