Galaxy Research has confirmed the number: 1,367 BTC drained from addresses associated with Coldcard hardware wallets. At any 2025 price point, that is a nine-figure loss — roughly $100 million depending on exit timing. The details stop there. No attack window. No methodology disclosed. No official response from Coinkite. Just a data point from a research firm that rarely publishes noise. The absence of detail is itself a detail. Galaxy does not publish numbers like this without holding the supporting chain data. The question is why they published only the aggregate — and what they are still investigating.
This is the uncomfortable pattern of 2025. The largest losses are no longer occurring at exchanges or in smart contract failures. They are happening at the endpoint. The user. The device. The last mile of self-custody. Coldcard occupies a peculiar position in the Bitcoin stack. It is not the Ledger of the mainstream crowd. It is the wallet of the paranoid elite — open-source firmware, air-gapped signing, deliberately offline design, the "not your keys, not your coins" ethos taken to its logical extreme. Its users are the people who mocked exchange custodians, mocked wrapped Bitcoin, insisted on running their own validation. The marketing reads like a manifesto for operational paranoia. Coinkite built its brand on refusing the compromises that made other vendors vulnerable. The company famously ships devices without cameras, without batteries, without wireless connectivity. Every attack surface is a design decision they declined to make. That record makes this event harder to dismiss — and the silence more conspicuous. If Coldcard users can be systematically drained, the self-custody thesis itself is under pressure.
I spent my 2018 audit years verifying smart contracts, not hardware. The Bancor audit taught me that the most dangerous flaws hide in the assumption layer — the things developers never thought to validate. A universal rule emerged: when a system fails at scale, the flaw is almost never where the marketing says it is. The private keys did not leak through BIP39 or BIP32 — those derivations have survived a decade of global adversarial scrutiny. The math has no mercy, and it also has no agenda. It does not compromise itself.
What fails is the surrounding stack. Three vectors deserve attention.
First, address fingerprinting. Galaxy's phrasing — "Coldcard addresses" — suggests that attackers can identify which on-chain addresses belong to Coldcard users. This is not magic. Hardware wallets generate addresses through specific derivation paths, and their UTXO management patterns form subtle fingerprints. A Coldcard user consolidating UTXOs, using specific change-address behavior, or transacting in distinctive patterns becomes identifiable at scale. If an attacker can cluster and label hardware wallet addresses, they no longer need a universal exploit. They can build a target list. Knowing where the gold is stored is half the heist.
Second, the user workflow. Coldcard's security model assumes a clean room: a secure computer, verified firmware, a trusted supply chain. That assumption breaks in three places. Firmware updates require a verification ritual most users perform incorrectly. Companion tools like SeedOR or Iris introduce additional software into the trusted stack. And the seed phrase — the actual crown jewels — must be handled, written, and stored somewhere. Social engineering against hardware wallet users is significantly cheaper than breaking secp256k1.
Third, supply chain compromise. The lowest-probability vector but the highest-impact one. A device intercepted during shipping, fitted with a malicious component, and re-sealed is indistinguishable to the end user. Coldcard's tamper-evident seals and signed firmware provide meaningful protection. But against a well-resourced operation — state actors, or groups with logistics connections — the chain is the weakest link. I would not discount this until Coinkite publishes its investigation.
The aggregate statistics tell a deeper story. The 1,367 BTC figure, if confirmed by independent on-chain tracing, implies multiple victims over a prolonged period. A single hack is an incident. A sustained drain pattern is an industry-level problem. This was not one overwhelmed user. This was a systematic extraction.
Market surveillance offers another angle. If the stolen coins move to exchanges, the chain will show it. Large deposits to known exchange wallets typically precede sell pressure. Traders would be wise to watch for BTC inflows to major trading platforms in the coming weeks. The absence of movement is itself informative — it suggests the attacker is waiting, or the funds are already controlled by an entity that does not need to liquidate.
Here is where the contrarian view demands equal time. This event does not prove Coldcard's core cryptography is broken. The evidence so far suggests the opposite: the devices held. The mathematics held. The attack succeeded because of poor operational security, a compromised workflow, or an unknown supply chain failure — not because the signature algorithms cracked. Trust, verify the stack. And when you verify, you do so at the protocol level first.
The bulls are also right about the market. One hundred million dollars sounds catastrophic until you place it beside Bitcoin's daily spot volume. This is not a systemic market event. It is a systemic trust event — and trust events have a slower fuse.
The event may even accelerate the shift toward multisig, quorum-based custody, and institutional-grade self-custody that has been discussed for years but adopted slowly. Bad news has a way of concentrating minds. The industry was already tepid after FTX. A hardware wallet drain in 2025 forces the last true believers to diversify their security assumptions.
But the sobering truth remains: the highest-conviction holders are now the highest-value targets. High yield, high graveyard — in self-custody, the yield is sovereignty itself. The more you hold, the more you are hunted.
Three changes are overdue. Multisig is no longer optional. The address fingerprinting problem requires a coordinated response from wallet developers. And Coinkite owes the ecosystem a transparent post-mortem — not platitudes, but a forensic report with timeline, attribution evidence, and remediation steps.
Until then, the data point stands: 1,367 BTC, a fingerprint, and a reminder that hardware wallets do not make you anonymous. They make you a target. Rug pulls are just bad code. So are most hacks. The worst ones are bad processes wearing the costume of good security.