Bitcoin’s security model rests on a simple assumption: a private key drawn from enough true randomness cannot be guessed. In late July and early August 2026, that assumption failed for thousands of holders who trusted a Coldcard hardware wallet. Reporting from Fortune, TechCrunch and CoinDesk describes an attack that reconstructed wallet seeds without ever touching the device, draining an amount that estimates place between roughly $70 million and more than $130 million depending on the source and the moment measured.
This article explains the mechanism, why the losses kept climbing in waves, and what the episode does and does not prove about self-custody.
What actually broke
A Bitcoin wallet seed is generated from a random number. If that number is drawn from a large enough space, no one can search all the possibilities. The intended standard for a seed of this type is about 128 bits of entropy, a range so large that brute-forcing it is not feasible.
According to The Hacker News, affected Coldcard firmware did not use the full strength of the device’s hardware random-number generator. Instead, older builds fell back to a predictable, software-derived value seeded from publicly readable chip data. The result was seeds carrying roughly 72 bits of entropy rather than 128. That gap sounds small when written as two numbers. In practice it is the difference between a key that cannot be searched and one that can.
Forbes and TechCrunch trace the root cause to a firmware build dating to March 2021, when the device stopped relying on strong, unpredictable randomness and took a shortcut that followed a pattern. Once researchers and attackers understood that pattern, they could reproduce the same process on their own computers, generate the likely candidate seeds, and check which ones unlocked wallets holding real bitcoin.
Why the theft came in waves
One striking feature of this incident is that it did not happen all at once. Fortune reported that since the disclosure, holders experienced several distinct waves of sweeps affecting more than 5,200 addresses. CoinDesk described a fourth wave still running as losses neared $120 million.
The wave pattern is consistent with how a weak-entropy exploit scales. Attackers do not instantly find every vulnerable key. They work through the reduced search space over time, and each pass surfaces more affected addresses. Because the flaw was tied to how seeds were generated years earlier, any wallet created on the affected firmware remained exposed until its owner moved funds, regardless of how carefully that owner otherwise behaved.
CoinDesk also noted that the latest wave lit up Bitcoin’s memory pool, the holding area for unconfirmed transactions. When many parties race to move the same threatened coins, fee competition rises and the mempool fills. In this case, some transactions could be overridden while still unconfirmed, adding a second layer of competition between attackers and defenders.
What this does not mean
It is easy to over-read a hardware-wallet failure. A few clarifications matter.
- The Bitcoin protocol was not broken. Its cryptography and consensus rules are untouched. The failure was in one vendor’s key-generation code.
- Not every hardware wallet is affected. This was a specific implementation defect on specific firmware, not proof that all cold storage is unsafe.
- Updating firmware is necessary but not sufficient. A patched device generates strong seeds going forward, but a seed already created with weak entropy stays weak. The only remedy is a new seed and a migration of funds.
The distinction between “the device is fixed” and “my existing key is safe” is the single most important point for affected users, and it is the one most easily missed under pressure.
The lesson hiding in the details
The uncomfortable part of this story is that many victims did everything they were told to do. They bought a reputable hardware wallet, kept it offline, and guarded their seed phrase. The weakness was upstream of all of that, in code they could not inspect and a randomness source they had no way to independently verify.
That points to a structural gap rather than a user error. Consumers are asked to trust that a manufacturer’s most critical function, the generation of unpredictable keys, works exactly as described, with no standard mechanism to confirm the approved randomness path is the one actually executing in shipped firmware. When that trust is misplaced, careful custody habits do not help.
For holders, the practical response is defense in depth: seeds generated with additional verifiable randomness, passphrases layered on top of a device seed, and multisignature setups that combine independently designed devices so that no single implementation failure can drain funds on its own.
Bottom line
The Coldcard incident was not a break in Bitcoin’s cryptography. It was a failure to feed that cryptography enough randomness, and it shows how a single upstream shortcut can undo years of otherwise sound self-custody. Affected users should treat a patched device as a fresh start, not a repair: generate a new seed and migrate funds. Everyone else should treat the episode as a reminder that “offline” and “hardware” are necessary conditions for safety, not sufficient ones.
Advertisement
Sources and review
This article was checked against the primary or authoritative sources below .
- The $120 million Coldcard wallet hack lights up Bitcoin's memory pool — CoinDesk
- Hackers steal over $130M by exploiting bug in offline hardware wallets — TechCrunch
- Coldcard Hardware Wallet Flaw Linked to $70 Million Bitcoin Theft in 41 Minutes — The Hacker News
- Bitcoin owners rocked by $116 million hack: What we know about the Coldcard exploit — Fortune
Frequently asked questions
According to reporting from TechCrunch and The Hacker News, older Coldcard firmware fell back to a predictable software value when generating wallet seeds instead of using the device's hardware randomness. That produced seeds with far less entropy than intended, which attackers could reproduce and brute-force.
Loss estimates vary by source and by the time of measurement, ranging from roughly $70 million to over $130 million. CoinDesk reported total losses approaching $120 million as a fourth wave of thefts ran, after Coinkite disclosed the flaw in late July 2026.
No. A firmware update does not repair a seed that was already generated with weak entropy. Affected users need to generate a new seed on fixed firmware and move their funds to it.
No. The flaw was in how one hardware wallet generated private keys, not in Bitcoin's cryptography or network. Wallets that used strong, unpredictable entropy are not affected by this specific issue.
Advertisement