Coldcard Heists May Near $114 Million as Suspected Fourth Sweep Begins

The Coldcard liquidation cycle is still active

This is an active liquidation cycle, not a one-off breach. Recent chain activity shows 448.7 BTC moved from 709 suspected addresses at 13.8 sweeps per block, about 45 times the pre-incident baseline. That kind of throughput suggests vulnerable wallets are still being targeted in real time.

The pace matters more than the drama. On July 30, attackers swept 1,082.65 BTC from 1,196 addresses in 41 minutes, worth about $70.2 million at the time. The reported waves have so far involved 1,196 Bitcoin addresses in 41 minutes on July 30 and 709 potential victim addresses, moving 448.7 Bitcoin, on top of earlier victims. That keeps the pressure on holders of legacy wallets, even after the patch was released.

The practical debate: patch versus migration

The key point is simple: fixed firmware protects newly generated seeds, but it does not repair a seed created by the flawed code path. Affected users still need to migrate. Galaxy also said similar transactions remained in the mempool, so the window for additional sweeps was still open at the time of reporting.

If you treat the patch as the full fix, you may be underestimating what is still happening to older wallets.

Why the attack keeps finding victims

Emergency firmware was shipped on July 31, but Coinkite is clear that installing it does not repair an existing seed. Migration remains the real fix.

Why the weak-seed flaw matters

A March 2021 firmware integration error routed seed generation to a software pseudorandom number generator instead of the STM32 hardware RNG. In practical terms, the wallet stopped drawing from the strong randomness path and fell back to a more predictable software path.

That changes the problem from "guess my seed" to "guess a much smaller set of likely seeds." Coinkite estimates roughly 40 bits of effective entropy on the Mk3 and about 72 bits on the Mk4, Mk5 and Q, versus 128 bits for a 12-word BIP-39 seed. That gap is what makes offline candidate generation feasible.

How the attacker filters targets

Block's analysis says an adversary can reproduce candidate output streams offline if they can narrow down device UID, timer state, and prior RNG-call history. The attacker then derives addresses and compares them with public blockchain data. In other words, public onchain activity does much of the filtering.

That also helps explain why the attack can continue across many wallets. The exposure depends on the firmware state when the seed was created, not on firmware installed later. If a seed was generated on vulnerable firmware, updating the device afterward does not make that seed safe.

Why migration is the only real fix

The distinction matters. Updating firmware is useful for future wallets, but it does not undo the weakness in an already-generated seed. Coinkite's guidance is to generate a new seed on patched firmware and move funds. Restoring the old seed elsewhere carries the weakness with it.

So the real question is not whether the bug has been patched in newer releases. It is whether a given wallet was created on vulnerable firmware. If it was, the capital should be treated as exposed.

What the market is watching now

Migration is still the core fix. What changes the market read now is the behavior of the flow.

Smaller balances may widen the fear set

This latest wave is leaning into smaller balances. That matters because attacks focused on large holders can leave smaller users thinking they are not at risk. When the net widens, more people may start moving or selling coins before they fully understand their exposure.

Cleaner transfer patterns complicate tracking

Galaxy says most transfers now use a fresh destination created for each victim, rather than converging on one obvious collection wallet. That does not make the attack less dangerous. It can make classification harder and slow efforts to track the full flow of stolen funds.

The labels still come from blockchain analysis

Galaxy's Thorn said the latest cohort is "LIKELY Coldcard victims" based on UTXO patterns and sweep behavior. That is strong enough to warrant caution, but it is still blockchain inference, not formal attribution.

The broader market risk is uncertainty. Some funds have already moved beyond the first set of destination addresses, which can amplify fear even while investigators and researchers continue mapping the full extent of the thefts.

What matters most going forward

The closest risk is still slow migration from old vulnerable seeds. As long as coins remain tied to weak seeds, they remain exposed to additional automated sweeps.