Exchange Reserves & Net-Flow

The Bitcoin exchange-reserves chart plots total BTC held on centralised exchange wallets across the long tail of tracked venues, paired with a daily net-flow panel beneath. The upper region of the figure is a filled-area series of total reserves on the right linear scale, in millions of BTC; a muted Bitcoin price line overlays for cycle context. The lower third is a per-day reserves-change bar — sage when reserves fell (net outflow, coins leaving exchange wallets), rust when reserves rose (net inflow, coins arriving at exchange wallets).

Today's aggregate reads 2.48M BTC across 10 tracked venues. The trailing 30-day net flow is +1,523 BTC — Neutral. The top-three venues hold roughly 76% of the visible reserves; concentration matters for the interpretation in § 05. The figure refreshes overnight; spot in the reading row above auto-refreshes a few times a day in the browser.

Each venue's on-chain wallet cluster is identified through a combination of public deposit-address disclosure, exchange-published proof-of-reserves audits, and on-chain forensic clustering (common-input heuristics, change-output reattribution). The cluster is then summed at the daily UTC close, and the per-venue totals are aggregated into a single reserves figure. The pipeline preserves the per-venue split for the tooltip; the chart shows the aggregate plus the day-over-day change.

Net flow on this page is Δreserves, not a sum of individual deposits and withdrawals. That is a deliberate choice: what matters for the sellable-supply narrative is the net change in what's parked on exchange. The trade-off is that internal venue reshuffles (a venue rotating coins between hot and cold wallets within its own cluster) net to zero and stay invisible, while genuine customer flow shows through.

Two upstream quirks shape what the chart shows. The dominant US venue used to operate Coinbase and Coinbase Pro as separately-labelled wallet clusters; the June 2022 announcement and November 2023 completion of the Pro → Advanced consolidation collapsed the two labels into one. Most upstream feeds (including ours) report the merged label; pre-2024 cross-venue charts that show a sudden Coinbase reserve step-up around 2022–2023 should be read as a label consolidation rather than a flow event. Wallet reattribution — when a previously unidentified cold wallet is added to a known cluster — is the other source of one-day discontinuities; the methodology page walks both in detail.

The series begins 1 May 2024, the first daily snapshot from the upstream cross-venue feed we use. Long-horizon context comes from HODL Waves and Reserve Risk, both extending back to 2010. Provenance for the cross-venue feed lives on the data sources page.

The 30-day net-flow regime sorts into three bands. Below −50,000 BTC, reserves are draining at multi-month-historic pace — the print that pre-2024 read bluntly accumulative. Above +20,000 BTC, reserves are growing meaningfully — the print that historically tagged distribution. The middle band is noise. Runs of ten or twenty consecutive same-direction days carry signal; singleton bars do not.

Six anchors trace the regime rotation across the chart's life. Reserves and 30-day net-flow values are pulled from the same daily snapshot powering the chart above. The cross-cycle context that pre-dates this series — the November 2022 FTX cliff, the longer major-venue duopoly era — lives in § 05 as prose rather than a table row.

For most of Bitcoin's history the chart's interpretation was uncomplicated. Coins on exchange were sellable; coins off exchange were not; therefore a falling reserve reading meant supply was being pulled into self-custody and the marginal seller was disappearing. Two named windows anchor the pre-2024 record cleanly. November 2022, the week of the FTX collapse, produced what the lead on-chain analyst Checkmate (James Check) called one of the largest net declines in aggregate BTC balance in history in his The Fall of FTX weekly report — 72,900 BTC off exchange wallets in seven days, one of only four comparable weeks on the record at that point (the others were April 2020, November 2020, and June–July 2022). A week later the Week 47 follow-up framed the running total — outflows running at 172,700 BTC per month — as holders seeking safety in self-custody. That was the clean version of the indicator.

The clean version stopped applying on 11 January 2024. The US spot Bitcoin ETFs that launched that day need a regulated custodian for the coins underlying their share creations; for most of the issuers, the named custodian is Coinbase Custody Trust Company, LLC. The IBIT prospectus states it directly: Coinbase Custody Trust Company, LLC (the ‘Bitcoin Custodian’) is the custodian for the Trust's bitcoin holdings, with Anchorage Digital Bank N.A. named as the additional alternate custodian. Equivalent designations sit in the ARKB, BITB and GBTC filings; the lone significant exception is FBTC, where Fidelity self-custodies through Fidelity Digital Asset Services. By early 2026, Coinbase Custody held roughly 84% of the US spot-ETF bitcoin under custody.

The mechanical consequence: a coin moving from a Coinbase trading wallet to a Coinbase Custody wallet shows up on this chart as an exchange outflow, but the coin has not left the institutional book at all. It has changed wrapper. The August 2025 on-chain analyst quicktake that framed the regime called it cleanly: same coins, different wrapper. The pre-2024 read — outflow as bullish accumulation tag — still applies on the share of flow that is not AP-routed (retail self-custody, OTC desk withdrawal to private custodians). The portion that is AP-routed needs the ETF-flows page cross-read to interpret. The signal is not broken; the interpretation is now regime-dependent.

For a dollar-cost-averaging investor. The chart is coarse cycle texture, not a buy/sell trigger. A multi-month run of sustained outflows is the kind of structural backdrop that has historically preceded multi-quarter price advances; a multi-month run of sustained inflows is the kind that has tagged distribution windows. The regime carries the signal; the daily print rarely does.

For a cycle-timing trader. The pre-2024 mental model — outflow accumulative, inflow distributive — is now a starting point, not a conclusion. Read against ETF flows: falling reserves with positive ETF flows means the flow is doing what it looks like; falling reserves with flat or negative ETF flows means a custody-wrapper shift, not a supply shock. The Coinbase Premium adds the lit-orderbook leg.

For a researcher. The per-venue tooltip breakdown does more work than its placement suggests. Coinbase's share of the visible book is the load-bearing cross-section for the ETF-custody story; offshore venues' shares carry the retail story. Read the share-of-total split as structural, the daily delta as noise.

Wallet reattribution produces one-day step changes with no flow. When a previously-unidentified cold wallet is added to a known venue's cluster — or removed, after a forensic re-classification — the reserve total can step up or down on a single day with no underlying coin movement. The signal is most likely to fire on small venues; the top-three concentration on this chart limits the damage on the aggregate line, but the per-venue breakdown can swing visibly. Treat single-day cluster discontinuities sceptically.

The Coinbase / Coinbase Pro merge produced a structural step. Pre-2023 cross-venue feeds tracked Coinbase and Coinbase Pro as separate wallet clusters; the June 2022 announcement and November 2023 completion of the Pro→Advanced consolidation collapsed the two labels into one. Charts that span 2022–2024 with a sudden Coinbase reserve step-up are showing the consolidation, not a flow event.

The post-spot-ETF custody redirection is the dominant caveat. After January 2024, US institutional flow moved off lit order books and into AP-mediated ETF creations at NAV. The independent market-structure analysis from June 2024 documented the corresponding rise in US-exchange depth share to 45% since the start of the ETF-led rally in October, up from 35% a year ago. The on-chain side of the same shift is what this chart records as Coinbase outflows; the ETF-flows page records the wrapper side.

Canonical questions from Google's “People also ask” block for bitcoin exchange reserves and btc leaving exchanges. The first answer restates the definition; the rest map the regime-shift framing.