It can be a lot more complicated than sending a "text message" to disable the chip.
The chip itself may need to determine if it has been moved to an embargoed or sanctioned country. For example, a straw-buyer in a Germany purchases 100 GPUs then re-exports them to North Korea.
When each chip powers up and is unable to authenticate location, it must not function. This is not a "backdoor" in the traditional sense, but rather an export control.
Depending on the various architectures and use cases, there are dozens of strategies. Everything from simple solutions like "phone home" to more complex seismic activity distance detection to determine geo-location. For chips that must function in clusters, there can be a single authenticator system in the network with a low latency requirement and cryptographic authentication. Only the authenticator may need to "phone home" to enable all the chips in the cluster. Any chip that cannot securely communicate with the authenticator over low latency (e.g. in the same building) will not complete booting and move into lock-down mode. You do not typically want to trash the chip on a failure, because a intermittent error could accidentally destroy billions of hardware that was operating legitimately in an approved location.
I've written proposals for these solutions in order to secure export approval from Commerce and other agencies.
