Confidential compute today is mostly used in high-risk B2B contexts.

For example, defense contracting, or health contexts.

In those cases, the end-user implicitly trusts the service provider to do what they say.

Confidential compute is more about the service provider not having to trust the cloud host.

The end-user is likely satisfied by an infrequent and manual audit of the service provider by a trusted auditor.

But it's possible to use confidential compute primitives for new use cases.

For example assembling a fabric of heterogeneous nodes operated by different, unknown parties… all running the same code, so creating a trusted fabric.

In those cases, you might need to do remote attestation to a previously unknown, skeptical third party at any moment.

Not hugely dissimilar from normal uses of confidential compute, but definitely distinct.