Why Paper Isn’t Enough
A seed phrase written on paper is a perfectly good backup — until the house fire, the flood, or the pipe that leaks into the cupboard where you kept it. If the paper goes and your hardware wallet fails in the same event, your crypto goes with them.
Metal backups exist for exactly that scenario. But “metal” is not a guarantee. Some designs fail in fire or crushing, and some fail in ways that only show up when you need them.
This guide covers what independent testing shows, the design features that separate good backups from bad ones, and how to store a metal backup without turning it into a target.
What the Independent Tests Show
The most thorough public testing of metal seed backups comes from Jameson Lopp, a Bitcoin security engineer who has stress-tested dozens of devices over several rounds. In his sixth round, published in August 2022 and the most recent we could find, he tested 22 devices three ways:
- Heat: about 2,000°F (around 1,090°C) for 10 minutes, then dropped into water, as a firefighter’s hose would do
- Corrosion: submerged in muriatic acid for more than 12 hours
- Crushing: mostly in a 20-ton hydraulic press
Lopp is clear that these are “not highly scientific experiments” — but they are the most rigorous public comparison available.
The best performers
- Tinyseed — Lopp’s favourite. It withstood more than 15 tons before deforming, and the data stayed legible.
- qreg-hr — a close second, barely changed by heat, acid or crushing.
- CryptoTag Odin — no data loss in any test, and fully legible after the acid bath.
No data loss, with minor issues
- CryptoTag Zeus — no data loss and some template fading after acid. Lopp praised its slide-on fasteners. Ledger sells the Zeus in its accessories store.
- Codl Punchplate, Coinplate Punch and Secret Seed Enigma — no data loss in any test.
- Bravo Seed Card and Bravo Seed Coin — no data loss; the acid caused pitting.
- Attenuo — no data loss; the acid actually made the stamps more legible.
- Willi’s Recovery Seed Plate — no data loss; surface pitting from the acid.
- DIYSeed — the paper template burned off, but the punched marks could still be read with a new template.
- CryptoNumeris Board S and Plate S — no data loss, though the printed templates became hard to read or disappeared after acid.
- CipherSeks Single and Double Plate — the data survived, but the acid erased the template, and heat stripped the screws on the double plate.
- Coinplate Alpha and Grid — the data survived, but heat seized the screws.
- CRYO — no data loss from heat or acid; crushing seized the screws.
- ImpressSeed Titanium — some lightly stamped letters were lost after acid, but the data was still recoverable.
The failures
- CipherTag — catastrophic failure. The acid dissolved much of the device, and it broke under pliers.
- Copper Seed Safe — the worst result. Heat melted an edge and caused slight data loss, acid caused partial loss, and crushing was catastrophic.
Two caveats. Products change after testing, so check that the model you buy matches what was tested. And these were extreme conditions — the point is to see how designs fail, not to predict exactly what a house fire does.
Five Design Lessons That Matter More Than Brand
Lopp’s testing points to design features you can check on any product, including ones he never tested.
1. The template must be deep, not printed
The letter grid or guide that tells you where each character goes should be deeply etched or cut into the metal. Printed or lightly engraved templates were the first thing to degrade in testing. If the template disappears, recovery becomes slow, error-prone guesswork.
2. Screws are a weak point
Small screws often stripped or seized after heat or crushing, leaving a device that is hard to open precisely when you need it. Slide-on fasteners proved easier to open after deformation.
3. Punching is easier than stamping
Punching marks into a pre-drilled or marked plate was the fastest and most user-friendly method Lopp found. Stamping letters onto small plates was tedious, and thin plates warped. Easier recording means fewer mistakes, and a mistake in your backup is as bad as losing it.
4. Compact encodings demand precision
Some devices encode words as dots or binary patterns to save space. They work, but recording and reading them requires care and good eyesight. That’s fine for some people and a real risk for others — including the family member who may one day need to read it.
5. Thickness and material matter
Thin material was a weakness. Copper performed poorly. Titanium pitted on the surface but generally kept its data, and stainless steel performed solidly.
The Main Types
| Type | How it works | Strengths | Watch out for |
|---|---|---|---|
| Punch plate | Punch a mark through a grid of letters or numbers | Fast, durable, hard to get wrong | Make sure the plate is sized for your backup format |
| Stamped plate or tag | Stamp letters with a hammer and letter punches | Very durable, compact | Slow; thin plates warp; letters can be shallow |
| Letter tiles in a frame | Slot metal letter tiles into a holder | Quick, reusable | More moving parts; check independent test results for the specific model |
| DIY | Stainless steel washers and a letter punch set | Cheap | Quality depends entirely on you |
Two products worth knowing
Trezor Keep Metal — Trezor’s own backup, made of AISI 304 stainless steel with a black surface treatment, described as resistant to water, acid, alkali and corrosion. It uses a punching method: a pre-marking pen to mark each letter first, a holding box to keep the plate steady, and an included punch. It comes in a single-share version and a 20-word multi-share version, which has fields to record the total number of shares and the recovery threshold. Related: Hidden Wallets and Split Backups: How to Set Up a Passphrase and Multi-share Backup
CryptoTag Zeus — a stamped titanium design. It showed no data loss in Lopp’s testing, has slide-on fasteners, and is sold through Ledger’s accessories store.
Trezor Keep Metal · CryptoTag Zeus
Disclosure: TrueToCrypto earns a commission on purchases through these links, at no extra cost to you. We have not run our own destruction tests; the results above are Jameson Lopp’s.
Choosing the Right One
Match the format of your backup. A standard 12- or 24-word recovery phrase needs a device with room for 24 words. A Trezor multi-share backup uses 20-word shares, and you need one plate per share.
You usually only need the first four letters. In the standard English word list used by most wallets (BIP39), the first four letters of each word are unique. Many metal devices are designed around this, which saves space and effort. Check the manufacturer’s instructions for your format before relying on it.
Budget for it. Most quality metal backups cost roughly €50 to €150, plus tools if they aren’t included. For anyone holding meaningfully more than €10,000, it is one of the best-value security purchases available.
The Part Nobody Talks About: Metal Is Easy to Find
A metal backup survives fire. It also survives being found.
Burglars and home invaders now know what seed backups look like. Physical attacks on crypto holders rose sharply in 2026, especially in Europe, and home invasions became the fastest-growing type. Related: Crypto Has Gone Physical: Wrench Attacks, Home Invasions, and What European Holders Should Do A steel plate in a desk drawer is a complete key to your wallet for anyone who finds it.
How to store it
- Not next to your hardware wallet. Someone who finds both has everything they need.
- Not labelled. No “seed”, “Bitcoin” or “Ledger” on the case or the plate.
- Somewhere a quick search won’t find it — or away from home, such as a safe deposit box.
- Never photographed, not even to check that you stamped it correctly.
Make the plate insufficient on its own
- Add a passphrase. With a passphrase-protected wallet, the plate alone opens only your standard wallet. Keep the passphrase somewhere else entirely — never stamp it on the same plate.
- Use multi-share backup. Split the backup into shares, for example any 2 of 3 or any 3 of 5, and store each plate in a different place. A burglar who finds one plate finds nothing usable.
Related: Hidden Wallets and Split Backups: How to Set Up a Passphrase and Multi-share Backup
After You Stamp It: Test the Backup
A backup you haven’t tested is a backup you’re hoping works.
- Use your wallet’s backup check. Trezor Suite offers a backup check, and Ledger offers a recovery-check app. These confirm the words on your plate match your wallet without exposing them to a computer.
- Enter the words on the device itself, never on a computer, phone or website.
- Recheck once a year, as part of a quarterly security audit. Related: Revoke Day: The 10-Minute Quarterly Crypto Security Audit
Frequently Asked Questions
Is paper good enough?
For small amounts, paper in a fire-resistant place is reasonable. For anything you couldn’t afford to lose, add metal.
Does a fireproof safe make paper safe?
It helps, but many safes are rated to keep paper below its ignition point for a limited time only. Metal plus a safe is better than either alone.
Titanium or stainless steel?
Both did well in testing. Titanium pitted on the surface but generally kept its data. What matters more is thickness, how deep the marks are, and how the device is fastened.
Can I just engrave it myself?
Yes — a stainless steel plate and a letter punch set work. Make the marks deep, and test the result.
Should I make two metal backups?
More copies mean more places it can be found. If you want redundancy, multi-share backup gives it without creating extra complete copies.
Can I photograph it to check I stamped it correctly?
No. A photo turns your fireproof backup into a digital copy that can be leaked, synced or stolen. Check it on the device instead.
The Bottom Line
Metal backups protect against the disasters paper can’t survive, but design matters more than the word “metal”. Look for deeply cut templates, solid material, and fasteners that won’t seize, and prefer punching over stamping if you value accuracy.
Then store it like the key it is: hidden, unlabelled, away from your hardware wallet, and ideally made insufficient on its own with a passphrase or a multi-share split.
Transparency: TrueToCrypto may earn affiliate commissions from links in this article. We have not performed our own destruction tests. Test results are from Jameson Lopp’s published stress tests (Round VI, August 2022). This article is educational and is not financial advice.
Contact: info@truetocrypto.com
Test results from blog.lopp.net (Round VI, 27 August 2022). Product details from the manufacturers’ websites. Verified October 2026.
