Nobody Knows How Much Bitcoin Is on Exchanges. Here's Why — and How We Solved It for DigiByte
Ask three on-chain analytics firms how much Bitcoin is sitting on exchanges today, and you will get three different answers. Not slightly different — materially different, often diverging by hundreds of thousands of BTC for the exact same date. None of them is lying. The uncomfortable truth is simpler: nobody actually knows. They are all estimating, and estimation is the only tool the Bitcoin ledger leaves them.
This isn't a knock on the analysts. It's a structural fact about how exchange reserves are measured on a chain with no native labels. Understanding why the estimates disagree is the fastest way to understand what a genuinely auditable reserve figure would look like — and why DigiByte can publish one when Bitcoin can't.
The discrepancy nobody advertises
Head to any two major analytics dashboards and compare their "exchange balance" or "exchange reserve" charts for Bitcoin. The shapes rhyme, but the absolute numbers rarely match. One provider's total exchange holdings can sit well above or below another's for the same block height. Revisions are common: a figure published for last month quietly changes as the model is retrained.
If exchange reserves were a simple on-chain fact, this couldn't happen — a fact is a fact regardless of who reads the ledger. The divergence is the tell. What's being published isn't a measurement. It's a model output.
Why exchange reserve estimates disagree
Bitcoin addresses carry no identity. The blockchain records that 3.5 BTC moved from one string of characters to another; it says nothing about whether either string belongs to Binance, a custodian, or a teenager's cold wallet. To turn raw addresses into "Binance holds X," an analyst has to infer which addresses an exchange controls. That inference is called clustering, and it rests on heuristics:
- Common-input-ownership: if several addresses are spent together in one transaction, they're probably controlled by the same entity. Probably — CoinJoin and other constructions deliberately break this.
- Deposit-address patterns: exchanges hand each user a fresh deposit address, then sweep balances to hot wallets. The sweep pattern leaks the cluster — until the exchange changes its wallet software.
- Behavioral fingerprinting: fee policy, change-address handling, batching cadence. Signal, not proof.
Every one of these is probabilistic. Two analysts applying reasonable-but-different heuristics will attribute different address sets to the same exchange, and therefore report different totals. The clustering models are also proprietary — the exact rules are the product, so you can't reproduce the number yourself even if you wanted to. And at Bitcoin's scale, with hundreds of millions of addresses, exhaustively enumerating every exchange-owned address is impractical, so providers sample and extrapolate.
Probabilistic, proprietary, and sampled. Three good reasons the estimates diverge — and three reasons you can't audit any of them. The best BTC exchange-reserve figure in the world is a defensible guess you're asked to take on trust.
How we solved it for DigiByte: enumerate, don't estimate
DigiByte is a smaller, fully public UTXO chain, and that changes what's possible. Instead of inferring exchange holdings from a clustering model, DigiScope enumerates them: we identify and verify individual exchange-owned addresses one at a time, publish the entire list, and sum the on-chain balances directly.
As of the latest block, that verified set covers more than 7.6 billion DGB — over 40% of the entire circulating supply — held across every exchange address we have been able to confirm. The largest reserve-holders (Binance, Crypto.com, Bitvavo, Uphold, OKX, and more) each get their own page with a live balance, address list, and flow history.
The difference from a BTC reserve estimate is not that our number is bigger or prettier. It's that you can check it.
- The full address list is downloadable — every verified exchange address, its balance, and when its label was last confirmed, as CSV or JSON. No proprietary model sits between you and the data.
- Every address is independently verifiable on-chain. Paste any of them into the block explorer and confirm the balance yourself against the DigiByte ledger. Our total is just the sum of numbers you can re-add by hand.
- Every address has a narrated page. Each exchange wallet links to a plain-language explanation of what it is and how it behaves, so the label isn't a black box either.
This is why we describe the coverage figure as a verifiable floor, not an estimate. It is the portion of supply we can prove is exchange-held, address by address. There may be more DGB on unlabeled exchange wallets we haven't confirmed yet — and if you find one, our methodology page shows exactly how the number is built so you can challenge it. The claim is falsifiable by design. A proprietary BTC estimate is not.
"Don't trust, verify" — applied to the number itself
The crypto ethos is "don't trust, verify," but exchange-reserve estimates quietly ask you to do the opposite: trust a model you can't inspect, reproduce, or falsify. That's not a criticism of the firms doing the hard work of clustering Bitcoin — it's the best anyone can do on a chain that scale and privacy have made un-enumerable.
DigiByte's advantage is that it doesn't force the trade. Full public ledger, tractable scale, open methodology. The result is the one thing a Bitcoin exchange-reserve chart can't offer: a reserve figure you can audit down to the last address, without asking anyone to take our word for it.
See the full method and download the dataset at the DigiScope exchange-reserve methodology page, or browse the live exchange analytics.