Two weeks ago I opened a spreadsheet to answer one question: how much XRP can actually be bought right now? Not "circulating." Not "non-custodial." Bought. The answer, after reconciling five data sources, was a number roughly three orders of magnitude smaller than the headline figure every aggregator publishes. CoinGecko says 63.09 billion tokens circulate. The +2% order book depth on Binance's XRP/USDT pair says $2.4 million. Coinbase's XRP/USD says $4 million. Cold logic cuts through the noise of FOMO — and what it cuts to is a 63-billion-token asset that cannot absorb a single mid-sized institutional order without visible slippage.
That gap is not a rounding error. It is the entire story.
Let me set the baseline. XRP Ledger has run for over a decade. Fixed supply: 100 billion. The chain has no staking inflation, no governance token, no yield farm. Its native escrow is a protocol-level primitive — funds are conditionally locked and cannot move until release conditions are met. That is the one supply number in this entire ecosystem I trust, because it does not depend on a company's honesty or an analyst's estimate.
Everything else does.
When the SEC approved spot XRP ETFs — Bitwise holding 420 million tokens, Franklin holding 296 million — a new layer entered the supply chain. Authorized Participants. Creation and redemption baskets. Retail holders cannot redeem individual shares for XRP; only APs can, and only in 25,000-share blocks. This changes the float structure in ways no circulating-supply figure captures.
So when a methodology piece landed arguing that "63 billion circulating tokens don't tell buyers what's for sale," I read it twice. Then I rebuilt the numbers. The article was right about the diagnosis and sloppy about the evidence. I want to fix both.
The confusion is not accidental. It is the natural output of a market that never agreed on what to measure. Start with the definitional war. CoinGecko reports 63.09 billion circulating. XRP Insights reports 68.59 billion "non-custodial supply." That is a 5.5 billion token spread — roughly $3 billion at current prices — produced by different definitions of what counts as circulating. CoinGecko excludes escrow. XRP Insights applies a wider net. Neither is wrong. Both are published. Investors treat them as interchangeable.
They are not additive, and they are not subtractive. Every one of these figures uses a different snapshot date and a different inclusion rule. The article states this plainly. What it does not state is how badly the confusion compounds when you stack the layers.
Ripple's own disclosure — filed June 30, 2026 — reports 37.656 billion tokens held by the company: 32.6 billion in escrow, roughly 5.056 billion outside it. XRP Insights, using a snapshot from October 8, reports 31.4 billion escrowed. Two numbers for the same pot, differing by over a billion tokens, published months apart. I flagged the timeline inconsistency immediately: a snapshot dated October 8 cannot cite a June 2026 filing. Either the snapshot date is wrong, the filing is misattributed, or the piece was assembled from data captured at different times and never reconciled. That is a data-quality defect, not a formatting quirk.
Now the double-count. Wrapped XRP — cbXRP — accounts for 140.4 million tokens backing a cross-chain token. That backing appears twice in the ledger of ledgers: once under exchange balances, once under DeFi reserves. Same 140.4 million tokens. Two line items. Bridge-wrapped assets destroy single-ledger verifiability, and anyone summing these categories is counting the same coins twice.
Lay out the layers and the confusion becomes structural. Escrow: 31.4 billion. Exchange-attributed wallets: 2.197 billion. Wrapped-token reserves: 477.4 million, of which 140.4 million backs cbXRP. Founder holdings: an estimated 4.6 billion. Lost or inaccessible: roughly 531 million. Spot ETF holdings: 716 million. Add them and you get a number larger than the circulating supply itself — which is the tell that the categories overlap.
Here is where my own audit experience matters. In 2021 I wrote a Python script to parse 10,000 NFT mint transactions and proved the metadata was pre-determined rather than generative. The lesson was not about NFTs. It was that a claimed distribution and an actual distribution diverge the moment you stop trusting the interface and start reading the transactions. XRP's supply data has the same disease. The interface says 63 billion. The order book says otherwise.
Which brings me to the number that matters. Binance XRP/USDT: +2% depth of approximately $2.4 million. Coinbase XRP/USD: approximately $4 million. Combined, roughly $6.4 million of orders within 2% of mid-price. For an asset in the top five by market cap. Compare that to major large-cap pairs that routinely show tens of millions in +2% depth. XRP's book is thin relative to its size — and thin books convert large market orders into large price moves.
The article notes that +2% depth measures orders within a price band and shifts in real time. True. A snapshot is transient and perishable. But the perishability cuts against the bull case, not for it. If depth is fleeting and modest at rest, it is almost certainly absent under stress.
Then there is the exchange-wallet figure: 2.197 billion XRP held in exchange-attributed wallets, down 44 million over seven days. Investors read exchange balances as sell pressure. That reading is broken here, because those wallets include cold storage and customer assets — coins that are not on any order book. A 44-million-token decline could be net withdrawal, or it could be internal custody reshuffling. The data cannot distinguish. Anyone trading the headline is trading noise.
Consider the lost-coin estimate: roughly 531 million XRP in dormant or inaccessible accounts, per trackers. That figure reduces the effective float if accurate — but it is an estimate produced by heuristics, not a protocol fact. The same is true of the 4.6 billion founder-holdings figure. Neither is enforced by code. Both are asserted by analysts. When I traced the TerraUSD collapse in 2022, the lesson was identical: the architecture failed at the point where an assumption was never encoded as a constraint. Unverified numbers are assumptions wearing lab coats.
There is a deeper problem with cbXRP that the methodology piece waves at without naming. When wrapped XRP backs DeFi positions while its reserve still counts toward exchange balances, the same coins are doing two jobs in two venues. That is rehypothecation by another name. In a stress event, both claims cannot be satisfied simultaneously. The ledger does not show the double obligation. It shows two balances that each look independent.
The bulls have one thing right, and it is worth stating. XRP's escrow is genuine. Protocol-level conditional locking is more trustworthy than a team's promise not to sell — the code enforces what a blog post merely asserts. Ripple's monthly escrow releases have historically returned most of the unlocked tokens to escrow, meaning net supply pressure is lower than the gross 31.4 billion figure implies. And the ETF float — 716 million tokens held by Bitwise and Franklin — is locked inside fund structures, reducing freely tradable supply. On the margin, that is supply contraction.
They built on sand; I built on skepticism. But even a skeptic must concede the escrow mechanism is the most auditable transparency Ripple offers. My problem is not with the escrow. It is with everything surrounding it — the founder-holdings estimate of 4.6 billion tokens that XRP Radar produces and nobody can verify, the dormant-account estimates, the lost-coin figures. These are guesses wearing the costume of data. I can audit escrow. I cannot audit a guess.
The 63 billion number is not a lie. It is a category error — a size metric mistaken for a liquidity metric. The code doesn't care what the aggregator prints; it only knows what the order book will fill. So here is the question every XRP holder should ask before their next position: if 63 billion tokens circulate, why does $6.4 million of depth stand between the market and a double-digit percentage move?