The situation doesn't appreciably change as it exists now with TPM, unless Microsoft uses the opportunity to cut out the service provided for signing Linux and

The situation doesn't appreciably change as it exists now with TPM, unless Microsoft uses the opportunity to cut out the service provided for signing Linux and other third-party kernels. Presumably the only difference here, AFAICT, is that with the enclave on-chip you can implement disk encryption without exposing the secret key outside the enclave. Currently, if you're using the TPM for disk encryption the key is simply unwrapped by the TPM, exposing the key either in the kernel or (if loaded into another enclave, like SGX) during transmission across the bus. This is because the TPM simply doesn't have the bandwidth or processing power for bulk encryption. Apple's T2 doesn't have this issue with disk encryption as the T2 chip itself actually can service disk block requests directly; the host OS needn't ever know the disk encryption key, even temporarily. The question for me is how Microsoft's new chip integrates with authentication, especially biometric authentication. AFAIU, on Apple hardware the T2 chips controls the fingerprint scanner, rather than relying on the host OS to present the scan, which means you can't bypass biometric authentication (i.e. presence) by cracking the host kernel. The T2 actually hosts an entire microkernel with far more rigorous and well analyzed process separation than the macOS or Windows kernels (or various ARM Secure Enclave host CPU microkernels, for that matter). Doing this "on chip" (i.e. on the main CPU) isn't necessarily better from a security perspective, especially if it's implemented in the main host kernel. Balancing software complexity and hardware isolation is a complex problem.