Most people think a circulating supply number is a disclosure. It is not. It is a definition wearing the costume of a fact. The line "63 billion XRP in circulation" is the output of one specific accounting choice, made by one specific data provider, on one specific day — and it tells you almost nothing about how many tokens are actually available to buy, at what price, on any given afternoon.
I found the shape of this problem the way I usually find these things. Not by reading a headline. By trying to reconcile three numbers that should have matched and did not.
Here is the anomaly. CoinGecko reports roughly 63.09 billion XRP as circulating. XRP Insights, using a broader definition, reports 68.59 billion in "non-custodial supply." That is a gap of 5.5 billion tokens — larger than the entire market cap of most assets in the top 100 — manufactured purely by a disagreement over what the word "circulating" means. Ripple's own disclosure adds a third figure. And a fourth source, an ETF prospectus, adds a fifth. None of these numbers can be added to or subtracted from the others. They are not measurements of the same object. They are measurements of different objects that happen to share a name.
That is the entire article. Everything below is the forensic work of showing why.
Context: What the XRP Ledger Actually Locks
To see why the numbers disagree, you have to understand what the XRP Ledger does with supply at the protocol level, because that is the only layer in this whole stack that offers a hard, verifiable answer.
XRP Ledger is an L1 settlement layer. It has been running for over a decade. Supply is fixed at 100 billion tokens. There is a fee burn — every transaction destroys a small amount of XRP — so the model is technically deflationary, though the burn rate is so small relative to supply that calling it "deflationary" is more of a category than a quantity. Nobody is pricing XRP on its burn.
The interesting mechanism is native escrow. Ripple locked a large portion of its holdings into protocol-level escrows that release on a monthly schedule, with any unused portion returned to escrow rather than dumped. This matters because of a distinction that most retail readers collapse into one bucket: protocol escrow is condition-locked — the tokens physically cannot move until the release condition is met — and this is categorically different from a team promising not to sell.
One of those is enforced by code. The other is enforced by a press release.
That distinction is the spine of this entire analysis. The escrow is the only part of the XRP supply picture that a buyer can verify without trusting a third party, a founder, or an analyst's spreadsheet. Everything else — exchange balances, founder holdings, dormant accounts, wrapped-token reserves — rests on estimates, labels, or definitions that change depending on who is doing the labeling.
Now layer on the data providers. CoinGecko maintains one circulating-supply methodology. XRP Insights maintains another. XRP Radar tracks on-chain wallets and produces estimates. Bitwise and Franklin Templeton publish ETF holdings in regulatory filings. Ripple publishes quarterly disclosures. Each of these uses a different definition and, critically, a different date.
And there is a data-quality flag I want to raise immediately, before any of the numbers get used: the snapshot underlying this analysis is dated October 8, yet it references disclosure documents dated January 2026 and June 2026. That is a timeline that does not close. Either the snapshot is stale, or the references are misattributed, or the compilation mixes vintages without saying so. Any supply figure that mixes dates is not a measurement — it is a collage. I will return to this at the end, because it is the kind of error that propagates silently through every downstream conclusion.
Core: Decomposing the Supply Stack, Layer by Layer
The clean way to do this is to treat each supply claim as a separate account, audit what it actually contains, and then refuse — absolutely refuse — to sum them until we have checked for overlap. That refusal is not pedantry. It is the entire point. The number that gets reported to you is a sum. The sum is where the error hides.
Layer 1: The Escrow — 31.4 Billion (or 32.6 Billion)
XRP Insights reports 31.4 billion XRP in Ripple-controlled escrow. Ripple's own 2026 disclosure reports 32.6 billion in escrow, out of a stated 37.656 billion total holding, leaving roughly 5.056 billion outside escrow. The two escrow figures differ by 1.2 billion tokens.
This is not a rounding error. It is a definitional gap. The two sources are counting the same escrow contracts on different dates, or counting returned-versus-released tokens differently, or one is netting the monthly return flow and the other is not.
Here is the one thing that survives the discrepancy: the escrowed tokens are the most auditable part of the XRP supply, because the lock is a protocol primitive, not a promise. When I audited zkSNARK circuit constraints for Zcash's Sapling upgrade, I learned to separate the constraints that the circuit actually enforces from the constraints that the documentation claims it enforces. Protocol escrow is the former. Founder lockups are the latter. Buyers who treat them as equivalent are pricing a promise as if it were a proof.
Layer 2: Exchange-Attributed Wallets — 2.197 Billion
Roughly 2.197 billion XRP sit in wallets attributed to exchanges. The instinct is to read this as sell-side pressure: coins on exchanges are coins that can be sold.
That instinct is wrong, and the reason is structural. Exchange-attributed wallets include cold storage and customer assets. A cold wallet holding customer deposits is not inventory the exchange can dump on the order book. It is custody. The label "on exchange" describes where the keys live, not whether the coins are for sale.
So when a reader sees 2.197 billion on exchanges and infers 2.197 billion of potential sell pressure, they have made a category error. The sellable subset is a fraction of that, and the fraction is not disclosed. This is the first place where the circulating-supply number starts to diverge from anything a trader can actually act on.
Layer 3: Wrapped Tokens and the cbXRP Double-Count
This is where the accounting breaks in a way that I find genuinely instructive, because it is a pure composability artifact.
About 477.4 million XRP sit in wrapped-token reserves, of which 140.4 million back cbXRP specifically. cbXRP is a cross-chain wrapped representation of XRP — an XRP-backed token living on another chain.
Now watch the overlap. The 140.4 million cbXRP backing appears simultaneously in the exchange-balance accounting and in the DeFi category. The same XRP is counted once as a reserve backing a wrapped token and once as part of an exchange or DeFi pool. This is the classic double-counting trap, and it is structural, not accidental.
Composability isn't free. It creates shadow supply. One XRP can now exist as a native ledger entry, a wrapped token on a second chain, and a liquidity-pool balance — three statistical appearances, one underlying token. When you sum across categories that do not know about each other, you double-count. When you subtract across categories that do not know about each other, you double-deduct. Both errors are live in the XRP supply debate.
I ran into the same class of problem during the 2020 DeFi Summer, when I wrote a Python simulation of flash-loan attack vectors across Uniswap V2 and Compound. The simulation kept producing arbitrage windows that did not exist, and the reason was that my model was double-counting the same liquidity in two pools that were, in fact, one pool seen through two interfaces. The bug was never in the math. It was in the belief that two labels meant two objects. The cbXRP overlap is the same bug, at market scale, running live in the data that feeds your exchange.
Layer 4: Founder Holdings — ~4.6 Billion (Estimated)
XRP Radar estimates founder holdings at roughly 4.6 billion XRP, with Chris Larsen's position being the large component. I want to be precise about what that number is: an estimate derived from on-chain clustering and attribution heuristics.
The source material is blunt about this, and the bluntness is correct: these figures cannot prove the existence of a current locked balance. An estimated cluster is a probabilistic statement about wallet ownership. It is not a covenant. It is not an escrow. It cannot be verified the way a protocol lock can be verified.
This is a governance-transparency gap, and it is measurable. The gap between "protocol-verified" and "analyst-estimated" is the gap between a number you can bet on and a number you can only hope about. When a founder cluster of that size moves, it is a market event — but the market cannot see the lock because there is no lock. There is only an estimate and a habit of trust.
Layer 5: Lost and Dormant — ~531 Million
About 531 million XRP are estimated to be lost or inaccessible, including dormant accounts. These are permanently removed from effective circulation in practice, though not from any ledger.
The asymmetry here is worth naming. Lost coins reduce real float but do not reduce reported supply. So the reported circulating number overstates the tradeable float by an amount nobody can precisely measure — because "lost" is itself an estimate. A supply number that cannot subtract its own dead weight is not a liquidity number. It is an upper bound.
Layer 6: ETF Holdings — 715.79 Million, Locked
Spot XRP ETFs introduce an entirely new holder class. Bitwise holds roughly 420 million XRP. Franklin Templeton holds roughly 296 million. Together, about 715.79 million XRP sit inside two funds.
Here is the mechanism that matters, and it is where the ETF structure quietly reshapes the float. Only authorized participants can create or redeem ETF shares, and they do so by delivering XRP or cash in basket-sized units. A retail holder cannot redeem a single share for its underlying XRP.
The consequence is a hard split: the XRP inside the ETF is locked at the fund level for as long as the fund holds it, while retail exposure to that XRP is confined to the secondary market. The underlying tokens are not free to trade. They are collateral for a share that trades somewhere else.
So when you add ETF holdings to your mental model of "supply," you are adding a category that behaves nothing like free-floating tokens. It is a supply sink, not a supply source. And note the coverage gap: the 715.79 million figure covers only two funds. Any ETF not captured by this count is invisible to the analysis, which means the ETF line item is simultaneously a lockup and an undercount.
Layer 7: The Number That Actually Matters — Order-Book Depth
Now we arrive at the part of the stack that the circulating-supply number is actively hiding from you.
Binance's XRP/USDT pair shows a +2% depth of roughly $2.4 million. Coinbase's XRP/USD pair shows roughly $4 million. Combined, about $6.4 million of resting liquidity within a 2% band of mid-price, across the two most important venues for the asset.
Sit with that. XRP is a top-five market-cap asset. Its aggregate +2% depth on the two largest relevant exchanges is in the single-digit millions of dollars.
For comparison, large-cap crypto assets routinely show tens of millions of dollars of +2% depth on a single major pair. XRP shows $6.4 million across two. This is the real liquidity figure, and it is orders of magnitude smaller than the circulating-supply number implies.
And even this figure is fragile. +2% depth records the orders resting within a price band, and it moves in real time as orders are placed and cancelled. Any snapshot of it is instantaneous and perishable. It cannot represent sustained available supply, because the book it measures is being rewritten every second.
I forked OpenZeppelin's ERC-721 library once to prototype a gas-optimized batch-transfer variant, cutting mint cost by roughly 40% through calldata compression. The lesson I carried out of that exercise was that the headline cost — the one everyone quotes — is almost never the binding constraint. The binding constraint is the second-order one nobody benchmarks. Here, the headline number is 63 billion circulating. The binding constraint is $6.4 million of depth. Those two numbers describe the same asset, and only one of them will determine what happens when a large order arrives.
The Reconciliation: Why You Cannot Sum the Stack
Put the layers side by side and the trap becomes explicit.
Escrow: 31.4 billion (or 32.6 billion, depending on source). Exchange-attributed: 2.197 billion. Wrapped reserves: 477.4 million, of which 140.4 million is cbXRP backing. Founder estimate: ~4.6 billion. Lost/dormant: ~531 million. ETF: 715.79 million.
These categories overlap. They cannot be summed, and they cannot be subtracted in sequence.
The double-count runs in both directions. CoinGecko's circulating figure already excludes escrow — so if you then subtract team and foundation holdings from it, you risk deducting the same tokens twice. Meanwhile the cbXRP backing is counted in exchange balances and again in DeFi reserves — so if you sum the categories, you double-count. One operation double-deducts. The other double-adds. Both are errors, and both are being committed by people who think they are doing careful analysis.
The 5.5-billion-token gap between CoinGecko's 63.09 billion and XRP Insights' 68.59 billion is the visible scar of this. It is not a data error. It is two methodologies measuring two different things and printing them under one label. The disagreement is the data. The number is the artifact.
There is one more reconciliation wrinkle worth flagging. XRP Insights reportedly excludes unlabeled U.S. spot-ETF custody wallets from its count. That means the broadest "non-custodial" methodology is simultaneously the most inclusive on some axes and the most exclusionary on others. Broad is not the same as complete. A definition can be wide in one direction and blind in another.
And underneath all of it sits the timeline anomaly: a snapshot dated October 8 citing disclosures from January and June 2026. A supply figure that cannot state a single consistent date cannot support a single consistent conclusion. When the timestamps do not reconcile, the totals do not either — they just look like they do.
Contrarian: The Market Is Debating the Wrong Variable
The consensus framing treats XRP's supply as a directional signal. More tokens unlocking is bearish. ETF lockups are bullish. Exchange outflows are bullish. Everyone is arguing about the sign of a number.
That argument is downstream of a mistake. Supply is not the variable that sets the price impact of a trade. Depth is. And the two have almost no relationship in this asset.
Look at the exchange-balance move everyone is citing: exchange wallets declined by roughly 44 million XRP in seven days. The reflexive read is bullish — coins leaving exchanges means holders are self-custodying, which means less sell pressure. But there is a second read that is at least as plausible: custody migration. Coins moving from an exchange hot wallet to a cold wallet or an institutional custodian is not a sentiment signal. It is an operational one. The same on-chain flow supports two opposite narratives, and the flow alone cannot distinguish them. That ambiguity is not a flaw in the data. It is a limit of what the data can express.
Here is the blind spot that I think is genuinely under-priced. Everyone treats the ETF lockup as neutral demand — money comes in, XRP gets locked, fine. But the lockup is a structural supply contraction that operates independently of price. It reduces free float mechanically, every day the funds hold. Meanwhile the data infrastructure that would let you see that contraction clearly does not exist in a consistent form. CoinGecko, XRP Insights, and XRP Radar disagree. Ripple discloses quarterly. The ETFs disclose in prospectuses. The asset's financialization is outrunning its data transparency, and that gap is the risk nobody is pricing.
I have seen this pattern before. When a system gets more complex faster than its observability tools improve, the failure mode is not a crash from a known risk. It is a surprise from a risk that was structurally present but statistically invisible. We don't get blindsided by things we measured. We get blindsided by things we defined incorrectly and then stopped questioning.
Takeaway: Forecast the Depth Shock, Not the Supply Shock
The next stress event in XRP will not be a supply shock. The escrow releases monthly, and the market has priced the schedule for years. It will be a depth shock. Somewhere, a large order — a fund rebalance, a liquidation cascade, an OTC desk crossing a size it should have crossed off-book — will meet a $6.4 million +2% book and discover that the 63 billion circulating number was never a liquidity guarantee. It was a definition.
Watch the depth-to-market-cap ratio, not the supply schedule. Watch for a second wrapped representation of XRP that re-collateralizes the first — because rehypothecation of cbXRP would turn a double-count in a spreadsheet into a double-claim on the same token. And watch the timestamps, because a number that cannot name its own date cannot name its own risk.
The tokens exist. That was never the question. The question is how many of them would actually sell to you at the price on the screen — and the honest answer is: far fewer than the headline claims, on a date nobody can pin down.
