Coldcard wallet flaw allows theft of $116 million in Bitcoin

Hardware wallets are supposed to be the safest place to store bitcoin, kept offline and out of reach of remote attackers. A newly disclosed flaw in one of the most trusted devices on the market shows that “cold storage” is only as safe as the code that generated the keys in the first place.

A five-year-old bug surfaces

Beginning July 30, 2026, an attacker started exploiting a firmware flaw in Coinkite’s Coldcard hardware wallet that traces back to a March 2021 firmware release. According to research from blockchain intelligence firm TRM Labs, a build configuration error in that release caused affected devices to generate seed phrases using a weak software random number generator instead of the device’s dedicated hardware-based entropy source. In practical terms, some Coldcard wallets set up under the flawed firmware never had truly random private keys, making them theoretically guessable years after the fact.

The theft unfolded in four distinct waves rather than a single event. Bloomberg reported that hackers ultimately moved approximately 1,816 bitcoin, worth close to $116 million at the time, out of more than 5,200 individual addresses. TechCrunch put the total haul even higher, at over $130 million, once additional affected wallets were accounted for.

Third-largest hack of the year

TRM Labs’ research places this among the largest crypto thefts of 2026, ranking it the third-biggest hack of the year and pushing 2026’s cumulative crypto hack losses past $1.2 billion across 276 separate incidents. Unlike an exchange breach or a smart contract exploit, this attack targeted a device category, hardware wallets, that the industry has long marketed as close to unhackable precisely because private keys never touch an internet-connected computer.

What went wrong

The root cause was not a remote exploit against a live device but a defect baked into how certain units generated their keys at setup time. Any wallet initialized under the affected firmware carried a seed phrase with far less randomness than intended, meaning an attacker who reconstructed the flawed generation process could, in effect, work out which private keys were possible and check them against known addresses. Fortune reported that Coinkite has since patched the firmware and urged owners of potentially affected devices to move funds to newly generated wallets.

A hardware wallet’s core promise is that the device itself, rather than the operating system of a phone or computer, generates and stores the private key, using a dedicated source of hardware entropy that never leaves the chip. That design is meant to remove the internet-connected attack surface entirely: even a computer riddled with malware cannot extract a key that was never transmitted to it. What this incident shows is that the design assumption only holds if the entropy source is actually used correctly in every code path. A build configuration error that silently substitutes a weaker software-based generator for the intended hardware one defeats the entire model without the device ever being connected to a network or physically tampered with. The keys were wrong from the moment they were created, years before any attacker touched them.

Why the theft took years to surface

The gap between the March 2021 firmware release and the July 2026 exploitation is itself instructive. A flawed seed generated in 2021 does not announce itself; the wallet functions normally, addresses receive and send funds as expected, and nothing about using the device day to day reveals that its keys came from a narrower, more predictable set than intended. Discovering the flaw required either an external security researcher identifying the weakness through code review, or an attacker independently reverse-engineering the generation pattern and beginning to search it, a process that can take considerable time even once a target is identified. That lag is a recurring feature of cryptographic implementation bugs generally: the vulnerability exists from the moment of the flawed release, but exploitation, and disclosure, can trail years behind, leaving a long window in which affected users have no way of knowing their holdings are at risk.

Why hardware wallet incidents ripple across the industry

An incident affecting one hardware wallet vendor tends to raise questions about the entire product category, even when the specific defect is not shared by other manufacturers, because the core value proposition of a hardware wallet, that it is safer than software wallets and safer than leaving funds on an exchange, rests on a level of trust in code that most users have no ability to independently verify. Other hardware wallet makers, whose devices were not affected by this particular flaw, still face the reputational consequence of a high-profile theft in the category, since the news that reaches most users is “hardware wallet hack” rather than the more precise “one vendor’s build configuration error in a specific firmware range.” That dynamic creates pressure across the industry toward more rigorous, third-party-audited entropy generation and build processes, and toward clearer, more prominent firmware version disclosures that let users check their own exposure without needing to parse a technical security bulletin.

The broader 2026 hack tally in context

Placing this incident within 2026’s cumulative hack total of more than $1.2 billion across 276 separate incidents is useful for calibrating how attackers are currently allocating their effort. A large share of crypto theft historically has targeted centralized exchanges and DeFi smart contracts, where a single successful exploit can drain a shared pool of user funds in one transaction. An attack that instead targets a hardware wallet’s key generation process is a different category of theft entirely: it requires identifying a defect specific to a device’s firmware, reconstructing the flawed logic, and then searching across a large number of individually generated seeds rather than breaching one shared vault. That this approach nonetheless ranks among the year’s largest thefts by dollar value signals that self-custody devices, long treated as outside the threat model that concerns exchanges and DeFi protocols, are now squarely within it once a defect of this kind is found.

What it means for Canadian holders

Canadians who self-custody bitcoin on hardware wallets, Coldcard or otherwise, should treat this as a reminder that self-custody security depends on firmware history, not just on keeping a device offline. Anyone using a Coldcard set up years ago should check Coinkite’s official advisories, confirm which firmware version generated their existing seed, and move funds to a fresh wallet generated under patched firmware if there is any doubt. As with most crypto theft, funds moved off compromised addresses are extremely unlikely to be recovered once transferred. That irreversibility is precisely why prevention, checking firmware history before there is any reason to suspect a problem, matters more in this asset class than in traditional banking, where a fraudulent transfer can sometimes still be clawed back through the receiving institution.