$130 Million Gone Because the Randomness Was Predictable: Which Hardware Wallet You Can Still Buy

Anyone following the relevant forums this week will find a type of post that has been rare for years: people opening their address in a block explorer and discovering the balance is gone. No phishing email, no lost backup, no identifiable mistake of their own.

The cause is a device that spent years being regarded as the safest thing you could buy for bitcoin: the Coldcard, made by Canadian company Coinkite. Since 31 July 2026, more than $116 million has left over 5,200 addresses. Some assessments now put the figure at around $130 million.

What makes it worse is who it hit. Not the careless, but the careful: offline storage, cold wallets, seed stamped into steel, never photographed.

This article covers three things. What happened, which manufacturer you can still buy in good conscience, and why switching off your savings plan in a week like this tends to be expensive.

What happened at Coldcard

The Coldcard has been the enthusiast's wallet for years: bitcoin only, no concessions in the interface, a very committed community. In firmware version 4.0.0, shipped since March 2021, the device bypassed its own randomness chip when generating a seed phrase. A software substitute took over whose output can be predicted. For five years, nobody noticed.

Anyone who generated a seed on the device in that period therefore holds a computable sequence rather than a random one. Whoever knows the flaw can recalculate the keys. The rest is a question of compute.

On 31 July, around 594 BTC left roughly 500 wallets within 25 minutes. By 2 August the figure stood at 1,367 BTC; a fourth sweep on 3 August took a further 449 BTC. The Mk2, Mk3, Mk4, Mk5 and Q models may all be affected, depending on which firmware was running when the seed was created.

The distinction matters: what was compromised is not the device but the randomness it used to create the key. A firmware update does not fix existing seeds. Anyone affected has to generate a new seed and move every holding.

Timeline and technical detail are here: 594 BTC gone in 25 minutes — the Coldcard flaw explained and the fourth sweep.

Track record beats spec sheet

Hardware wallets all advertise the same terms. Secure element, open source, air gap, PIN, passphrase — that sits on practically every spec sheet and compares about as usefully as horsepower figures on cars.

What is not on the sheet: how a manufacturer has behaved in recent years when something went wrong. After this week, we consider that the more important criterion. Here are the three vendors we are asked about most often.

Coinkite (Coldcard)

An excellent technical reputation, consistently bitcoin-focused, one of the most loyal user bases in the market. And now a key-generation flaw that went undetected for five years. That is the most serious category, because the damage cannot be repaired after the fact. Coinkite has published corrected firmware and is telling users to move funds to newly generated wallets. Anyone looking to buy a device right now should wait for the independent review.

Ledger

The French market leader has the longest incident list in the sector. What tends to get lost in the argument is that none of those incidents affected the hardware itself.

• July 2020: a breach of the online shop exposed more than a million customer records. They remain the basis for phishing letters and counterfeit replacement devices to this day.
• May 2023: the announcement of Ledger Recover, the option to split a seed into encrypted shards, contradicted the promise many customers thought they had bought.
• December 2023: malicious code reached the Connect Kit, and with it a range of dApps, through a former employee's phished npm account. Around $484,000 was drained; the victims were dApp users, not wallet holders.
• January 2026: customer data again, this time via shop partner Global-e.

Ledger documents the incidents in its own security incident report. The risk for users here is not the firmware but the letterbox: the 2020 delivery addresses have circulated for years, and the scams built on them keep improving.

Tangem

The counter-design: Tangem uses cards instead of a device with a screen, a cable and a battery. The key is generated inside a secure element certified to Common Criteria EAL6+ and does not leave the card, and backup runs through a set of two or three paired cards. There is no USB port, but there is no screen of its own either, and the firmware is not open source.

The record is mixed. Independent reviews by Kudelski Security (2018), Riscure (2023) and Cure53 (2026) reported no critical findings. In December 2024, however, the app wrote the private keys of seed-phrase wallets into logs that could be sent to support; Tangem fixed the bug and said fewer than 0.1 percent of users were affected. In 2025 and 2026, Donjon, the security lab of manufacturer Ledger, showed two attacks on the access code that require physical access to the card and cannot be fixed by an update on cards already shipped; Tangem considers them impractical. We therefore do not recommend Tangem: on top of the findings come the missing screen, the closed code and our disappointing experience with the company's communication, and there is no partnership. If you already use a card, choose a long access code made of letters, numbers and symbols.

Disclosure: CryptoTicker runs partner programmes with some of the providers named here. The assessment above follows the publicly documented incident record, not the commercial relationship. Where our view differs from it, we say so — as with Coldcard and Ledger.

Prices, supported coins and ratings in detail are in our hardware wallet comparison.

Three rules for choosing

1. Buy on history, not on features

Before ordering, it is worth searching the manufacturer's name alongside "incident", "breach" or "vulnerability". Zero results is not a good sign but usually just an indication of poor documentation. What matters is the response: how quickly the disclosure came, whether it was complete, whether an independent audit followed. Manufacturers who dissect their own failures in public are generally the safer bet.

2. Do not leave entropy to a single device

The Coldcard case was a randomness problem. Two established methods address it. Either you generate the seed from dice rolls, which several devices support, or you build a multisig using devices from two different manufacturers. In a 2-of-3 setup spanning two brands, a vendor-wide firmware defect no longer means total loss.

3. Check in regularly

Cold storage tempts you not to look for years. A quarterly appointment is enough: check balances through a block explorer — the public address is sufficient, and the seed goes nowhere — skim the vendor's firmware changelog, verify where the backup is kept. That costs fifteen minutes, four times a year.

Why the savings plan should keep running

Bitcoin trades around $64,100 on 5 August 2026. The news flow is unpleasant: the wallet disaster, two perp DEXs drained in July, and the first US spot ETF closing its doors. Many investors respond by suspending their savings plan until things calm down.

That is understandable and usually a mistake. A savings plan is not a market instrument but an instrument against your own mood. Its value is created in the months when you want to switch it off. Anyone who paused in 2022 missed the cheapest stretch of the cycle and restarted in summer 2024 at markedly higher prices.

Our pick: Coinbase — but not via the savings-plan button

For European investors, Coinbase is currently the obvious choice. Since June 2026 its European business has run under a MiCA licence based in Luxembourg, consolidating the previous national authorisations, including the German one. Since the last transition period in the EU expired on 1 July 2026 (in Germany it already ended on 31 December 2025), that authorisation determines who is allowed to stay in the European market. Add clean SEPA rails, local-language support and a tax export the common tools can handle.

Correction (September 30, 2026): An earlier version gave July 1, 2026 as the end of the transition period without the German rule. Under Article 143(3) MiCA (Regulation (EU) 2023/1114), the transitional regime applied EU-wide until July 1, 2026 at the latest, and member states could shorten it. Germany did: under section 50(2) no. 3 of the German Crypto Markets Supervision Act (KMAG), the continued authorisation expired on December 31, 2025 at the latest.

Less well known is that the convenient recurring buy in the app is the most expensive way to buy there.

On €200 a month that adds up to roughly €60 a year, for an identical result in the portfolio. If you want both, set up a standing SEPA transfer into your Coinbase account and place one limit order a month in Coinbase Advanced, then withdraw to your own wallet. The automation then sits in the standing order rather than in the buy button.

Providers side by side — fees, minimum instalment, execution frequency, withdrawal costs — are in our guide to buying bitcoin. For which exchanges still operate under regulation in Europe after MiCA, see regulated crypto exchanges compared.

What to take away

The Coldcard incident is not an argument against self-custody. It shows that choosing the manufacturer is the real decision, and that the choice should rest on documented history rather than on standing within the scene.

Three steps worth taking this week:

1. Establish your exposure. If you use a Coldcard and generated the seed on the device after March 2021, create a new wallet today — on corrected firmware or on another manufacturer's device — and move everything. Waiting behind a test transaction achieves nothing here: the flaw is public and so are the affected addresses.
2. Review your custody setup. Not on the feature list or the price, but on incident history and response behaviour. The hardware wallet comparison does the research for you.
3. Keep the savings plan running and, while you are at it, move it from the app to a monthly limit order. Five minutes of work, roughly €60 saved a year.

Beyond this week, one conclusion holds for wallets as much as for exchanges: in practice, security is less often a technical question than a question of selection. Choose providers on verifiable history rather than on promises and you have avoided the larger part of the risk.

(As of 5 August 2026. This article is not investment advice. Prices and fee models change; check current terms with the provider before every purchase.)