The chip generates private keys within itself and does not allow them to be exported (in the clear - encrypted backups could be allowed). It stores and signs transactions as any wallet would. The user interface might be provided by a special purpose device, or the chip can be integrated with smartphones.

However it works, when the chip signs a Bitcoin transaction using wallet keys, it also signs the transaction with a private key that was issued to the chip at manufacturing time. The purchaser can then provide the signed Bitcoin transaction, chip-specific signature over that transaction and the certificate chain proving that it's a genuine chip, all sent to the seller who then verifies all the signatures. If they match, you know the Bitcoin transaction was created by a chip that won't allow double spends to be created.

Given that modern smartcard security is extremely strong, this gives you a good level of confidence in the Bitcoin transaction. You don't have to broadcast it to the network immediately. You can broadcast it whenever is convenient, secure in the knowledge that the buyers don't have any technical way to create double spends.

In reality, custom chips+devices for this is probably too hard and not the way to go. Trusted computing (ARM TrustZone et al) can provide similar assurances, without the need to design custom chips. It can all be done in software because the boot processes are strictly controlled and keys can be encrypted under the fingerprints of specific execution environments.



It may be true but those smart cards were already licensed around the world, and the US also uses smartcards in many places (GSM network SIM cards, DirecTV cards, etc). I don't think smart card patents explains it.