Funds accounting

The principal BTC figures measure four different things: value consumed from source addresses, value received by first collectors, value retained after consolidation, and transaction fees. They also depend on whether the set is limited to 500 transactions or includes 695 more transactions that Block preliminarily associated by a shared fingerprint and Galaxy Research attribute to the same operator. Keeping the stage and the set attached to each number removes most of the apparent disagreement.

Transaction-set figures are fixed to the pinned chain-derived dataset at commit 47d8f554. The four-address balance check is fixed to Bitcoin block 960,525 at 06:48 UTC on 1 August 2026. Neither figure tracks live balances. The cards measure the 30 and 31 July transaction sets this page reconciles and nothing wider. Galaxy Research has since published a larger estimated total covering a third wave that is not reconciled here; it is kept separate below rather than folded into these figures.

This page is the accounting rather than the advice. If your own addresses may be affected, start with the risk assessment, which works from the firmware a seed was generated on rather than from whether an address appears in the set below.

What the principal figures measure

The narrower set contains 500 transactions spending 1,324 UTXOs from 500 distinct source addresses. The transactions confirmed in four consecutive blocks, 960,188 to 960,191. Their inputs total 594.51379184 BTC, of which 594.47722484 BTC reached one first collector and 0.03656700 BTC was paid in transaction fees. TFTC publishes the collector receipt, not the gross input value.

Rob Hamilton's preliminary post reported the same 1,324 UTXOs, 500 transactions and approximately 594.48 BTC, followed by a consolidation of about 562 BTC. It described a "3 block window", while the transactions are confirmed across four blocks. The block timestamps run from 01:36:08 to 01:51:26 UTC, a span of 15 minutes 18 seconds, consistent with Hamilton's separate description of an approximately 15-minute period.

The first collector later spent 341 outputs into the largest final holding while retaining a remainder. The 695-transaction set followed a separate path through three first collectors and two batch consolidations. A single first collector and four final holdings are therefore successive stages of the flows, not competing descriptions of one stage.

Galaxy's wider attribution

Galaxy Research reported 1,196 addresses and 1,082.65 BTC from 01:10:20 to 01:51:26 UTC on 30 July, across blocks 960,183 to 960,191. It attributed the wider set to one operator using the transaction pattern identified by Block. Galaxy also published a derivation-path count of 1,183 BIP-84, 7 BIP-49 and 6 BIP-44 addresses.

Galaxy described the repeated target as a "hardcoded 30.0 sat/vB" fee rate with no change output. Galaxy Research, X, 31 Jul 2026

Realised fee rates cluster near that target but vary slightly because fees are whole satoshis. Across the source and consolidation transactions, the fee total in this site's reconciliation of the pinned chain-derived dataset is 0.08399331 BTC. That is the total paid to miners, not a measure of overpayment. Galaxy's comparison with prevailing fee rates is a separate reported estimate.

In this site's check, the source transactions from the two sets spend from 1,195 original addresses: 1,182 native SegWit P2WPKH, 7 P2SH and 6 legacy P2PKH. Adding the native SegWit first collector when it later spends produces Galaxy's 1,196 total and numerically matches its reported 1,183 BIP-84, 7 BIP-49 and 6 BIP-44 counts. This is a plausible reconciliation, not a verified description of Galaxy's method. On-chain output types do not independently reveal wallet derivation paths, and the post does not define whether an internal collector is included in its address count.

flowchart LR
  A["500 source addresses
594.51379184 BTC gross"] --> B["first collector
594.47722484 BTC"]
  B --> C["largest final holding
plus retained remainder"]
  D["695 attributed source transactions
488.13939738 BTC gross"] --> E["three first collectors
488.10957948 BTC"]
  D -. "same-operator attribution" .-> A
  E --> F["two batch consolidations"]
  C --> G["four final holdings
1,082.56919591 BTC theft-linked"]
  F --> G
    

Solid edges show the on-chain movements used in this site's reconciliation. The dotted edge marks the reported attribution of the 695-transaction set to the same operator. Both sets fall within Galaxy's 41-minute window.

The four accountings, side by side

The rows below retain each publisher's measurement stage. The site-calculated row then reconciles those stages without changing the publishers' claims.

Galaxy's 1 August revision covers a wider set than this page

At 18:38 UTC on 1 August 2026, Galaxy Research reported a third sweep wave of 207.7294 BTC and revised its estimated observed size of the incident to 1,367.05 BTC, stated as approximately US$88.6m, across 4,585 addresses. That supersedes the 1,082.65 BTC total Galaxy published on 31 July, which is the figure the accounting above reconciles.

Galaxy states its own method limit in the same thread. It bounds every Galaxy figure quoted on this page and is the first thing to read beside the revised total, not a footnote to it.

The data "is derived solely from analyzing Bitcoin block data and the unspent-output set". Galaxy has "not utilized compute to test whether the addresses we have identified as possible victims were indeed generated with low entropy", and calls the work its best effort on short notice, which "should not be considered to be complete or definitive". Galaxy Research, X, 1 Aug 2026

Galaxy separately reported 1,366.3865 BTC under attacker control, with all endpoint attacker addresses fully unspent on chain. Those endpoints cover all three waves in Galaxy's attributed set. They are not the same set as the four consolidation addresses this page checks at block 960,525, so the two holdings figures are not comparable and neither confirms the other. Galaxy does not publish the endpoint addresses, so the unspent finding is not rechecked here.

The same thread also states that the vulnerable firmware shipped on 17 March 2021 around block 674,951, and that no coin identified in waves 1 to 3 was created before that block. That bears on the affected firmware range rather than on the accounting here, and is taken up on the firmware page.

The estimate register

This register preserves the estimate as each source stated it. "Copy held" means the archive has a captured version that can be checked for later changes. The press rows mostly restate the primary accountings rather than add independent analysis.

The US$38 million and US$70 million figures primarily reflect different BTC sets, 594.48 BTC versus about 1,082.6 BTC, as well as each publisher's BTC/USD conversion time. Galaxy's approximately US$88.6m belongs to a third set again, the three-wave scope described above. The later 45.9 BTC cluster is not included in the main accounting on this page because its relationship to the principal set has not been established. Tracker rows carry the figure held at the stated capture time; those sites publish live totals that have continued to move since.

Published windows use different sets and clocks

The timing statements are not all measurements of the same transaction set. Where a source gives a date without a timezone, it cannot be converted to a UTC date without an additional assumption.

What the 500-address set shows

Within the 500-source-address set, gross amounts consumed range from 0.14999770 BTC to 29.89252501 BTC, with a median of 0.409059365 BTC. Because the set has an even 500 values, that median is the average of its two middle values and can fall on a half-satoshi; it is a statistical summary, not an on-chain amount. Of those addresses, 110 held more than 1 BTC and another 16 held exactly 1 BTC. The observed minimum is approximately 0.15 BTC; the data do not establish that it was an operator-selected threshold.

Hamilton reported that all addresses in the 500-transaction set used single-signature spending conditions, none used Taproot, and the spent UTXOs dated from 2021 to 2026. Those are properties of the observed outputs and spending scripts, not proof of each owner's wallet configuration or seed creation date. Independent output classification for the original source addresses gives 1,182 native SegWit P2WPKH, 7 P2SH and 6 legacy P2PKH across both attributed sets. These counts numerically reconcile with Galaxy's published BIP-84, BIP-49 and BIP-44 figures as described above, without independently establishing those derivation paths.

What the wider attributed set shows

A privacy-preserving aggregate derived from the first three waves in Kelbie's pinned independent chain dataset covers 1,195 original source addresses and 1,082.65318922 BTC gross. The median source-address value is 0.27001931 BTC, 231 addresses contain at least 1 BTC, 17 contain at least 10 BTC, and the range runs from 0.000045 BTC to 51.07343852 BTC.

Holdings at the 1 August 06:48 UTC check

Four addresses held the post-consolidation amount attributed to the two principal sets at block 960,525. The table reports the theft-linked component, not the address balance observed at that check. It therefore excludes later unsolicited deposits.

The original first collector for the 500-transaction set sent 562.01962301 BTC to the first row and retained 32.45055543 BTC in the fourth row. No funds left the four-address cluster after that final consolidation by 06:48 UTC on 1 August 2026. This is narrower than saying that the addresses had never sent funds. Later unsolicited deposits added 0.00046535 BTC to the first address and 0.00000777 BTC to the fourth, bringing the observed combined balance to approximately 1,082.56966903 BTC at that check time.

No captured source reports a mixer deposit, exchange deposit, seizure or recovery for this four-address cluster as of that time, and none appears in captures held to 2 August 2026. This is an absence of reported evidence, not a prediction about future movement. Two community trackers, coldcard-watch and coldcard-hack-tracker, monitor attributed addresses and publish changing totals. Their captured figures use differing treatments of the separately reported later cluster and later inbound deposits, so they should not be used as a fixed incident total without those qualifications.

Galaxy cautioned that future transactions involving affected keys would not need to share this transaction pattern. The observed fee and output pattern characterises the set attributed by Block and Galaxy; it is not a necessary property of exploiting the firmware flaw.

Reported handling, and what is not reported

Two threads about how the incident is being handled appear in the captured record and belong beside the accounting. Block said it was sharing the relevant information with the appropriate authorities, and that it would provide further updates when doing so would not interfere with the investigation. Coinkite, through NVK, said it would provide a written incident summary specific to a user's loss along with any transaction data it could share, for people pursuing a police report, an insurance claim or their own investigation, and said it was cooperating with the on-chain investigators and with any law-enforcement agency that opens a case.

Nothing in the held captures records an exchange freeze, a flagging commitment or any other custodial action against the four consolidation addresses as of 2 August 2026. No captured source names an exchange or service that has undertaken to stop or mark deposits from them. That is an absence in what has been published and collected here, not a finding that no such step has been taken.

One collection route is open on the analysis side. Galaxy's own disclaimer records that some of the addresses behind its on-chain topographies were received from victim reports on X, and that its analysis "should not be considered to be complete or definitive". Its identified set is therefore open to further reported attacker and victim addresses rather than closed, which is one reason its published total has already moved once. This archive does not collect or publish victim addresses, and nothing here should be read as advice to send an address to a third party.

Causation remains unverified

The timing and transaction pattern are consistent with the public incident, but the blockchain does not identify the hardware or software that generated a private key. TFTC explicitly separates its reproduced chain event from the unproven claim that every transaction resulted from this firmware flaw.

Where this page or the site's navigation uses the word theft, it describes the public incident attribution rather than an independently established finding about authorisation. The transactions, amounts, timing and consolidation are on the chain and reproduce from the pinned dataset. The chain does not show whether each transfer was authorised, who operated the sweep, or what share of the attributed value the firmware flaw caused.

Not yet known

• How many people were hit. Source-address counts and people are not equivalent. One person or entity can control multiple addresses, and an address can have multiple controllers. The number of distinct affected owners is unknown.
• Whether the 695 earlier transactions are the same operator. Block's fingerprint match is reported but not independently confirmed. These transactions occur in blocks 960,183 and 960,185, minutes before the 500-transaction set and within Galaxy's published window. They are not evidence by themselves of activity on an earlier date.
• Whether the reported 31 July cluster is related. Kevin Kelbie reported 1,216 transactions moving 45.9 BTC after Block's scan end. Seven of eight stated markers match, while the replace-by-fee behaviour differs. It is tracked separately and is not included in this page's principal total.
• What is in Galaxy's third wave. Galaxy's revised total spans three waves and 4,585 addresses. The third wave is not in the pinned dataset behind this page and has not been checked transaction by transaction here, so the difference between the two totals is not itemised. Galaxy states it has not tested whether the addresses it identifies were generated with low entropy.
• Whether the consolidated funds will move, and where. No forward statement is possible. The finding above is fixed to its stated 06:48 UTC check time.
• Any recovery, seizure or insurance payout. None has been reported in any captured source as of 2 August 2026.

Checking this page

Material claims and claim groups carry a scoped evidence label. Verified means it was checked against the cited artefact. Reported records an attributed statement without endorsing it. Derived identifies arithmetic from stated inputs, and unverified identifies a claim that available public evidence does not establish. Contested appears separately when relevant sources disagree. The full definitions are on the About page. Captured sources are available through the evidence pages where publication rights allow.

The selected transaction accounting is fixed to the pinned source dataset, while the four-address balance check is fixed to Bitcoin block 960,525 at 06:48 UTC on 1 August 2026. Live balances and published attributions may change; later revisions belong in a dated update rather than being folded silently into these figures.