"Not your keys, not your coins." Crypto's oldest mantra has become background noise. People nod, leave their funds on exchanges, and only scramble for hardware wallets after the next collapse or hack. It happens every time. Hardware wallet searches spiked 42% and sales climbed 36% in the weeks following every major incident in 2024 and 2025.
The market has noticed. Hardware wallet revenue hit $564 million in 2025 and is projected to reach $2.6 billion by 2034. Institutional cold storage adoption surged roughly 50% year-over-year as regulators pushed firms to demonstrate actual key control. The demand is real. What is shifting is what a hardware wallet needs to become.
At EthCC in Cannes, we sat down with Jan Setina from Trezor and Guy-Louis Grau from Keycard. They are building different things and making the same underlying argument.
Trezor has been fully open source since launching the industry's first hardware wallet in 2014. Jan was blunt about what that costs: some closed drivers simply cannot be used, forcing the team to rewrite components from scratch. It slows everything down. The latest device, the Safe 7, was built with components ready for post-quantum upgrades before the software even exists to need them.

Jan was careful about it:
We are looking into some post-quantum signatures. We are doing some benchmarks on the device level. But even in the post-quantum world, users can verify that the device they got from Trezor is legit and comes from us.
Keycard came at the problem from a different angle. Their product is a smart card with an EAL6+ secure element, NFC-compatible with 15-plus wallets, and designed so the user picks the interface while the key stays on the card. Guy-Louis made a point that rarely comes up in hardware wallet discussions: physical threat modeling. If someone knows you are holding crypto and you are carrying a device that advertises it, you have a problem. Keycard's answer is a plain black card that looks like a gym pass, a duress PIN that opens a decoy wallet, and lockout after three wrong attempts.
The most forward-looking part of either conversation was about AI agents. Jan's framing was precise: the hardware wallet becomes the device through which a human signs the constraints on what automated systems can do on their behalf. The agent can spend up to $200 a week. You sign that policy yourself. The hardware is the last human checkpoint.
On quantum, the threat is moving faster than expected but remains distant from Bitcoin's 256-bit encryption. The harder problem is coordination. No central authority can mandate migration across every wallet, exchange, and dormant holder. That process, everyone agrees, needs to start well before urgency arrives.