A hardware wallet is supposed to be the boring, reliable part of holding Bitcoin. Plug it in, generate a seed, keep the seed offline, sleep well. That premise took a direct hit this summer when Coinkite, maker of the popular Coldcard wallet, confirmed that a firmware bug going back to March 2021 had quietly made thousands of private keys guessable. Attackers drained roughly 1,816 BTC, worth about $116 million at the time, from more than 5,200 addresses in four separate waves between July 30 and August 4, 2026. By mid-September, the incident still ranked as the largest hardware wallet exploit on record, according to blockchain analytics firm TRM Labs.
The Coldcard hack lands in a year already crowded with nine-figure crypto losses, but it stands apart from the usual DeFi bridge exploit or exchange breach. Nobody stole a password. Nobody phished a private key. The wallets did exactly what buyers were told they would do: generate a seed phrase offline, on a dedicated device, away from any network. The seed was just never as random as Coinkite claimed. That distinction is why security researchers, self-custody advocates, and now competing wallet makers are treating this as a different kind of warning than a typical hack headline.
What Happened: Inside the $116 Million Coldcard Theft
The first sign of trouble surfaced on July 30, 2026, when on-chain analysts spotted roughly 594 BTC, about $38 million, leaving close to 500 dormant wallets in a single burst. Coindesk covered the initial theft the next day, framing it as an exploit that let attackers “recreate wallet recovery phrases and steal bitcoin from what users believed were securely self-custodied wallets,” according to the outlet’s July 31 report. Coinkite responded the same day with a security advisory to Coldcard Mk3 owners, warning that funds tied to certain firmware versions might be at risk.
That was only the opening move. A second wave pushed total losses to about 1,082 BTC by August 1. A third wave, tracked by blockchain intelligence firm Galaxy Research, brought the running total to roughly 1,367 BTC, or about $89 million, drained from 4,585 addresses by August 2 and 3. A fourth wave detected that Monday morning pushed the confirmed total to approximately 1,816 BTC, close to $116 million, from more than 5,200 addresses, TRM Labs reported. Security writer Boaz Sobrado, covering the incident for Forbes, put the individual toll in stark terms: one victim alone lost the equivalent of C$1.6 million from a single Coldcard-generated wallet, according to his August 4 account.
Not every forensics shop landed on the same number. Galaxy Research’s separate tally put confirmed losses at 1,596 BTC, roughly $103 million, spread across a wider set of about 7,300 addresses, with a warning that a fourth suspected wave could push the eventual total as high as 2,055 BTC, or around $130 million, once fully verified. The spread between estimates says less about sloppy accounting and more about how hard it is to cleanly separate stolen-because-of-the-bug from moved-for-unrelated-reasons when thousands of addresses generated by the same flawed process sit on-chain for years before anyone notices the pattern.
Timeline: Four Waves of Theft in Five Days
Strip away the competing dollar totals and the sequence itself tells a clean story about how an attacker who has cracked a randomness flaw actually cashes in. Wave one hit the oldest, most dormant wallets first, the ones an attacker could brute-force or precompute against with the least urgency. Each subsequent wave moved faster and touched more addresses, suggesting the attacker, or attackers, scaled up tooling once the first sweep confirmed the bug was exploitable at scale rather than a one-off fluke.
- July 30, 2026: Wave one drains roughly 594 BTC (~$38 million) from about 500 addresses.
- July 31, 2026: Coinkite issues a security advisory to Coldcard Mk3 owners, and Coindesk and Bitcoin.com report the incident publicly.
- August 1, 2026: Running total climbs to roughly 1,082 BTC as a second wave is detected.
- August 2–3, 2026: Third wave brings the total to about 1,367 BTC (~$89 million) from 4,585 addresses.
- August 3–4, 2026: A fourth wave pushes confirmed losses to approximately 1,816 BTC (~$116 million) from more than 5,200 addresses.
- Mid-September 2026: Forensic firms still disagree on the final total, with estimates ranging from 1,596 BTC to a possible 2,055 BTC once outstanding waves are verified.
Five days from first detection to the fourth confirmed wave is fast by the standards of most crypto incidents, where bridge exploits or exchange breaches often unfold in a single afternoon and then spend weeks in recovery negotiations. Here, the slow bleed happened because the vulnerability wasn’t a smart contract bug that gets patched and forgotten. It was baked into how the affected devices had generated keys for roughly five years, meaning every wallet created in that window stayed a live target until its owner moved funds to a freshly generated seed.
The Root Cause: A Five-Year-Old Entropy Bug in Coldcard Firmware
Every hardware wallet’s entire value proposition rests on one number: how much true randomness it can pack into the private key it generates. Coldcard devices ship with a dedicated hardware random number generator chip specifically so that entropy doesn’t depend on a predictable software routine. According to Sobrado’s Forbes analysis, the vulnerability, present since firmware version 4.0.0 in March 2021, “caused the device to bypass its dedicated hardware randomness chip during key generation, instead using a predictable software substitute.”
The practical effect was a collapse in effective entropy. Payments and custody platform Bitso summarized the incident bluntly on its blog, stressing that “this was not a hack of the device,” but rather “a flaw in how Coldcard generated private keys, dating back to a 2021 firmware update, which left those keys predictable for anyone who could reconstruct them without ever touching the wallet,” per the company’s August coverage. Self-custody education site SelfCustody Labs framed the mechanics even more directly: “The attackers simply guessed the seeds, because for over five years, a firmware bug made those seeds guessable.”
The severity varied by device generation. Older Mk2 and Mk3 hardware running firmware 4.0.1 through 4.1.9 saw effective entropy fall to roughly 40 bits, several orders of magnitude weaker than the 128-bit standard the industry treats as secure. Newer Mk4, Mk5, and Q models running any firmware released before July 31, 2026 were still affected, though less severely, at around 72 bits. Forty bits of entropy is small enough that a well-resourced attacker with GPU or ASIC-accelerated search can brute-force the keyspace in a realistic timeframe, especially once they know roughly when and on what firmware a wallet was generated.
Which Coldcard Models and Firmware Versions Are Affected
Coinkite’s July 31 advisory specifically named the Mk3 line, covering firmware 4.0.1 through 5.0.3, described as the device’s final supported release. Bitcoin.com’s news desk quoted the advisory’s scope directly: “Anyone who generated a seed on a Mk3 running firmware version 4.0.1, released in March 2021, through version 5.0.3, the final release supporting the Mk3, may be affected.” Coinkite’s early analysis, cited in that same advisory, indicated the newer Mk4, Q, and Mk5 models were “not affected” in the same way, though outside technical write-ups later complicated that reassurance by pointing to the weaker 72-bit entropy window on those devices for firmware released before the fix.
That distinction matters for anyone trying to figure out if they’re exposed. A Coldcard Mk3 owner who generated a seed any time between March 2021 and July 2026 falls squarely in the highest-risk category. Owners of newer hardware face a smaller, but not zero, exposure window tied to firmware released before the July 31 cutoff. Coinkite’s guidance, echoed by outside researchers, is unambiguous on the fix: patching firmware alone does not retroactively secure a seed generated under the flawed randomness path. The private key itself needs to be abandoned, not just the software that produced it.
Coinkite’s Response, and Why a Firmware Update Isn’t Enough
Coinkite’s playbook followed a familiar shape for hardware security failures: acknowledge the affected models, ship a fix, and tell customers to move funds. What made this response harder to execute cleanly is that step three isn’t optional the way it sometimes is with a software patch. Coindesk’s reporting was explicit on this point, noting that affected users must generate entirely new wallets and move their funds because updating the firmware alone doesn’t eliminate the risk. A corrected firmware version stops new keys from inheriting weak entropy, but it does nothing to protect a seed that already exists on the blockchain with a guessable private key behind it.
That creates an uncomfortable race condition for anyone who hasn’t yet acted. Once the flaw became public knowledge on July 31, every remaining at-risk wallet became a target for any attacker capable of reproducing the guessing technique, meaning delay directly increases risk rather than being a neutral wait-and-see option. Coinkite’s messaging drew a sharp line between a flaw and a hack, and technically that’s accurate, no device was physically compromised and no server was breached. For the roughly 5,200-plus addresses already drained, the distinction offered little comfort.
2026’s Biggest Crypto Hacks, Ranked
The Coldcard incident doesn’t top this year’s loss leaderboard, but it’s the only entry in the top tier that stems from a hardware wallet rather than a bridge, exchange, or protocol. TRM Labs’ running tally for 2026, cross-referenced with additional incident tracking from Galaxy Research and CertiK, puts the year’s largest thefts in this order.
| Incident | Date | Amount Stolen | Category |
|---|---|---|---|
| Liquid Network (Blockstream) | Sept. 6, 2026 | ~$319 million | Bitcoin sidechain reserve drain |
| KelpDAO / LayerZero | April 18, 2026 | ~$292 million | Cross-chain bridge exploit |
| Drift Protocol | April 1, 2026 | ~$285 million | Privileged-access compromise |
| Cetus | May 2026 | ~$223 million | Protocol-level exploit |
| Coldcard (Coinkite) | July 30–Aug. 4, 2026 | ~$116 million | Hardware wallet entropy flaw |
| Tectonic (Cronos) | Aug. 2026 | ~$74 million | Lending protocol exploit |
Context helps here. TRM Labs’ half-year report counted 207 separate crypto hacking incidents in the first six months of 2026 alone, totaling close to $972 million in losses, with the two April incidents, KelpDAO and Drift, accounting for nearly 60 percent of that figure on their own. Separately, PeckShield’s August tracker recorded 50 major exploits that month totaling $136.3 million, a 49 percent drop from July, a trend the industry had been citing as evidence that DeFi security was finally improving before the Coldcard and Liquid Network incidents reset the conversation in the opposite direction.
Coldcard vs Trezor: Comparing 2026’s Hardware Wallet Security Incidents
Coldcard wasn’t the only hardware wallet brand dealing with a security crisis this year, but the nature of the failures couldn’t be more different. Trezor, one of Coinkite’s main rivals, spent most of August and September managing the fallout from a third-party breach rather than a cryptographic flaw in its own devices. The comparison is instructive because it shows two entirely separate ways a self-custody vendor can put customer funds and data at risk.
| Incident | Vendor | Disclosed | Scope | Direct Fund Loss | Root Cause |
|---|---|---|---|---|---|
| Entropy flaw | Coinkite (Coldcard) | July 31, 2026 | 5,200+ BTC addresses | ~$116 million | Firmware bypassed hardware RNG chip since 2021 |
| ShipMonk vendor breach | Trezor / SafePal | Aug. 17, 2026 | 53,487 customers exposed | None reported | Shipping partner hit via Metabase SQL injection |
| Expanded breach disclosure | Trezor | Sept. 4–5, 2026 | 67,000 additional US customers | None reported | Same ShipMonk incident, wider scope confirmed |
| STM32 phishing campaign | Trezor | Sept. 9–10, 2026 | Targeted customer emails | None confirmed | Attackers impersonated Trezor using leaked contact data |
| Sidechain reserve drain | Blockstream (Liquid Network) | Sept. 6, 2026 | Reserve wallet, multiple exchanges | ~$319 million (85% later returned) | Range-proof cache bug in Elements software |
The Trezor case traces back to ShipMonk, a third-party logistics provider that handles order fulfillment, not to any weakness in Trezor’s own hardware or firmware. Attackers tied to the ShinyHunters group exploited a Metabase SQL injection flaw, tracked as CVE-2026-72898 with a maximum CVSS score of 10.0, to pull customer records. The Hacker News reported that ShipMonk notified Trezor of the breach on August 10, 2026, and that the logistics company has since secured the affected systems and improved its security following the incident, per the outlet’s September coverage. Coldcard’s problem, by contrast, sat entirely inside Coinkite’s own code for half a decade.
Entropy 101: Why Random Number Generation Is the Whole Ballgame
Bitcoin private keys are, at bottom, a 256-bit number picked at random. Security doesn’t come from the algorithm being secret, it comes from the number being unpredictable enough that guessing it is computationally impossible. A hardware random number generator chip exists specifically to source that unpredictability from physical noise, thermal variance, or similar unclonable physical processes, rather than from a software algorithm that a determined attacker could reverse-engineer or predict given enough information about the device’s internal state.
When Coldcard’s firmware quietly swapped that hardware source for a software substitute, it didn’t just weaken the keys a little. Dropping from 128 bits of entropy to 40 bits isn’t a linear degradation, it’s roughly the difference between a search space no computer on Earth could exhaust before the heat death of the universe and one a determined attacker with modern GPU clusters could plausibly work through in a targeted campaign. That gap is exactly why researchers describe this as one of the more severe cryptographic failures to hit consumer hardware wallets, distinct from data breaches or phishing campaigns that expose personal information but leave the underlying key math intact.
The Self-Custody Debate Reignites
Bitcoin’s self-custody pitch has always rested on a simple trade: hold your own keys, and you’re no longer trusting an exchange, a bank, or a custodian not to lose your money. The Coldcard incident complicates that pitch because it shows the trust doesn’t disappear, it just moves. Instead of trusting an exchange’s security team, self-custody users are trusting a hardware manufacturer’s firmware engineering, and in this case that trust was misplaced for roughly five years before anyone noticed.
Coindesk’s initial coverage flagged this tension directly, suggesting the exploit could push some investors toward spot Bitcoin ETFs as an alternative to direct custody. That framing understates how the industry actually tends to respond to incidents like this. Rather than abandoning self-custody outright, the more common reaction among serious Bitcoin holders is to diversify hardware wallet brands, adopt multisignature setups that require compromise of multiple independent devices rather than one, and treat firmware update history as a genuine due-diligence item rather than an afterthought. Still, for newer or less technical holders, an incident this large and this well covered gives custodial alternatives a real talking point they didn’t have before.
Market and Industry Reaction
The dollar figures involved, while large in absolute terms, didn’t move Bitcoin’s price in any measurable way, a sign of how routine nine-figure crypto losses have become in 2026. What did move was scrutiny. Coingabbar’s forensic write-up ranked the Coldcard incident as the third-largest crypto attack of the year by the time all waves were tallied, a notable distinction given that most of the incidents ranked above it involved DeFi protocols with far larger total value locked than any single hardware wallet maker’s customer base.
Competing wallet vendors used the moment to reinforce messaging around audited, open-source firmware and hardware-attested randomness testing, though the coverage gathered here doesn’t name a specific rival making direct public comparisons. What’s clearer is the effect on the broader narrative: an industry that spent the first half of 2026 pointing to falling DeFi hack totals as proof of maturing security practices now has to reconcile that story with a hardware wallet flaw that sat undetected since 2021 and a sidechain bug that cost Blockstream $319 million just weeks later.
A Decade of Hardware Wallet Security Failures: Historical Context
Hardware wallets have earned their reputation as the safest mainstream option for holding crypto precisely because catastrophic, wallet-wide failures are rare. Most previous incidents in the category involved supply chain tampering, where a malicious reseller intercepted a device in transit and pre-loaded a compromised seed, or phishing campaigns that tricked users into typing their recovery phrase into a fake app or website. Those attack types share a common thread: the device itself, running unmodified firmware, was never the weak link. The user’s handling of it, or the supply chain before it reached them, was.
The Coldcard entropy bug breaks that pattern entirely. It’s a case where a legitimate device, purchased through legitimate channels, running firmware the manufacturer itself shipped and supported for five years, generated cryptographically weak keys by design flaw rather than tampering. That places it in a smaller and more alarming category: a manufacturer-side cryptographic defect that persisted through multiple firmware release cycles without being caught by internal testing, external audits, or the broader open-source security research community that regularly examines popular wallet firmware. It’s a reminder that audited and open-source reduce risk but don’t eliminate it, especially for a subtle entropy-path bug that doesn’t announce itself through crashes or obvious errors.
Regulatory and Law Enforcement Response
As of mid-September, none of the reporting gathered on this incident names a specific law enforcement agency or financial regulator that has opened a formal case tied to the Coldcard theft. That’s a notable contrast with several other high-profile 2026 crypto incidents, where agencies moved relatively quickly to announce investigations or, in the case of some exchange hacks, attribute the activity to specific state-linked hacking groups. The absence of a public regulatory response here likely reflects the diffuse nature of the victim pool: thousands of individual wallet holders scattered globally rather than a single centralized entity with the resources and incentive to coordinate a law enforcement referral.
Blockchain intelligence firms including TRM Labs and Galaxy Research continue to track the flow of stolen funds, which is often the precursor to eventual law enforcement action even when it takes months or years to materialize publicly. Given the scale involved and the involvement of firms whose core business is supporting exactly this kind of investigation, additional developments on the enforcement side remain plausible even though nothing concrete has surfaced yet.
What Coldcard Owners Should Do Right Now
The guidance from Coinkite and outside researchers converges on the same core steps, and the sequencing matters because doing them out of order leaves funds exposed longer than necessary.
- Update firmware first. Install the latest Coldcard firmware for your specific model before generating any new seed, so the corrected randomness path is active from the start.
- Generate a brand-new seed. Never reuse or refresh an existing seed generated on affected firmware. The old private key remains mathematically guessable regardless of what firmware now sits on the device.
- Move funds in a single, deliberate transaction. Sweep the full balance from the old, potentially compromised address into the new wallet rather than leaving a partial balance behind.
- Check your firmware history. If you can’t confirm your Coldcard has run firmware released after July 31, 2026, treat every seed generated on it as compromised until proven otherwise.
- Consider a second hardware brand for large balances. A multisignature setup spanning two different manufacturers means a single-vendor flaw like this one can’t compromise the entire balance on its own.
Five Predictions for Hardware Wallet Security After Coldcard
The incident’s ripple effects are still working through the self-custody ecosystem, but a few directions look likely based on how the industry has responded to comparable failures in the past.
- Third-party entropy audits become a selling point. Expect competing manufacturers to commission and publicize independent randomness testing of their hardware RNG implementations, turning what used to be an assumed baseline into an explicit marketing claim.
- Multisig adoption rises among larger holders. Users holding significant balances are likely to spread custody across multiple hardware brands specifically to hedge against a single-vendor firmware defect wiping out an entire position.
- Open-source scrutiny intensifies, but doesn’t guarantee safety. Coldcard’s firmware is open-source, and the bug still went undetected for five years, which will fuel calls for funded, systematic entropy-path audits rather than relying on incidental community review.
- Bitcoin ETF and custodial products gain a talking point. Expect custodial providers to reference this incident in marketing aimed at less technical holders who want Bitcoin exposure without managing hardware wallet firmware themselves.
- Compensation pressure builds on Coinkite. Given the scale of confirmed losses and a five-year window during which the flaw existed, affected users are likely to pursue compensation claims, following a pattern seen after comparable disclosures in the wallet and exchange space.
Frequently Asked Questions
What exactly caused the Coldcard hack?
A firmware bug present since version 4.0.0 in March 2021 caused affected Coldcard devices to bypass their dedicated hardware random number generator chip during key generation, substituting a predictable software-based process instead. That weakened the effective randomness of generated seeds, in some cases down to roughly 40 bits from an expected 128 bits.
How much money was stolen in the Coldcard exploit?
Estimates vary by forensic firm. TRM Labs and SelfCustody Labs put confirmed losses at roughly 1,816 BTC, about $116 million, from over 5,200 addresses. Galaxy Research’s separate analysis found 1,596 BTC, about $103 million, across roughly 7,300 addresses, warning the eventual total could reach 2,055 BTC if additional waves are verified.
Which Coldcard models are affected?
Coinkite’s advisory names Mk3 devices running firmware 4.0.1 through 5.0.3 as the primary risk group. Mk4, Mk5, and Q models running firmware released before July 31, 2026 also carry weakened entropy, though at a less severe level than the Mk3 line.
Is updating my Coldcard firmware enough to fix the problem?
No. Updating firmware prevents new seeds from being generated with weak entropy, but it does nothing to secure a private key that was already generated under the flawed process. Anyone who generated a seed on affected firmware needs to create an entirely new wallet and move funds to it.
Is this the same issue as the Trezor breaches this year?
No. The Coldcard incident was a cryptographic entropy flaw inside Coinkite’s own firmware. Trezor’s 2026 incidents stemmed from a data breach at ShipMonk, a third-party shipping and fulfillment partner, that exposed customer contact information but did not involve any weakness in Trezor’s own device firmware or key generation.
Was Bitcoin’s core protocol affected by this hack?
No. The vulnerability was confined to Coldcard’s firmware implementation. Bitcoin’s underlying protocol and blockchain were never compromised. The theft occurred because private keys generated on affected devices were predictable, not because of any flaw in Bitcoin itself.
How does this compare to the Liquid Network hack from September 2026?
The Liquid Network incident, which cost Blockstream roughly $319 million before 85 percent of funds were returned, involved a bug in the Elements software underpinning that Bitcoin sidechain rather than a consumer hardware wallet. It’s a larger dollar figure but affected a shared infrastructure reserve wallet rather than individual holders’ private keys.
What should I do if I’m not sure whether my Coldcard is affected?
Treat any seed generated before July 31, 2026 as potentially compromised. Update to the latest firmware, generate a completely new seed, and move your full balance to the new wallet in one transaction rather than leaving any funds on the old address.
