Intel’s next-generation Xeon platform, code-named Diamond Rapids, will be the last to include Software Guard Extensions. Intel states this in a post on its site by Mike Ferron-Jones, who leads go-to-market for platform security and integrity. From here the company is focusing its confidential computing roadmap on Trust Domain Extensions, TDX, which isolates whole virtual machines and containers instead of individual enclaves.
Adoption, not a flaw in SGX, is the reason Intel gives for the change
The post credits SGX with establishing confidential computing in the data center and says it still serves security-conscious customers. Its explanation for the move is that the industry has coalesced around virtual machine isolation because it offers the scalability, ease of deployment and operational model customers need for broad adoption. It mentions no security problem with SGX.
The deployments Intel lists show where TDX is already in use. Microsoft Azure, Alibaba Cloud, Google Cloud, IBM Cloud and Volcano Cloud already offer TDX-based services, and Red Hat, Canonical and VMware have added TDX support for enterprise and private clouds. Apple’s Private Cloud Compute, expanded together with Google Cloud, runs on TDX trust domains on Xeon, and OPPO has launched a comparable private computing cloud with Intel.
SGX platforms stay supported into the early 2030s, so the change is a migration window
Intel says it is announcing the direction now to give customers several years to plan, validate and migrate at their own pace, with support and security updates for SGX platforms continuing into the early 2030s. For anyone running SGX enclaves today, that makes Diamond Rapids the last platform to buy for that model, while existing hardware keeps its updates. Anything new that can run in a confidential VM would point to TDX.
On the TDX side Intel lists three additions it continues to invest in: trusted I/O through TDX Connect, post-quantum cryptography and enhanced attestation. It gives no dates for them. It also says it is working with Nvidia to coordinate confidential computing across CPU and GPU, so that data and model protection can extend through an AI pipeline.
The post is about Xeon. It gives no launch date for Diamond Rapids and does not mention Core processors, so it says nothing about SGX on Intel’s desktop and laptop chips.