Business news |
By Brian Tristam Williams
NVIDIA says its GPUs with Confidential Computing are now being used for confidential inference in Apple’s Private Cloud Compute, as Apple extends the system beyond its own data centres to Google Cloud. The move brings NVIDIA Blackwell GPUs into the hardware security architecture that supports some server-side Apple Intelligence workloads.
Apple built Private Cloud Compute, or PCC, to handle AI requests that are too demanding for on-device models while keeping user data out of reach of operators and administrators. In its security update, Apple says the Google Cloud implementation keeps the same PCC requirements: stateless computation, enforceable guarantees, no privileged runtime access, non-targetability, and verifiable transparency.
The new implementation combines NVIDIA Confidential Computing with NVIDIA GPUs, Intel CPUs with TDX, and Google’s Titan chip. Apple says devices will trust only PCC software that has been cryptographically approved by Apple, and that the company retains control over the PCC software stack regardless of where the infrastructure is hosted. It also plans to publish binaries for public inspection, extend research tooling, and give security researchers access to live PCC nodes in research mode through the Apple Security Bounty Program.
The expansion is tied to Apple’s third-generation foundation models. Apple says the new family includes two on-device models and three server-based models running on PCC. Its most capable server model, AFM 3 Cloud Pro, is intended for demanding tasks such as agentic tool use and complex reasoning, and was optimised for NVIDIA GPUs. Apple also says the model family was custom-built with Google, using technologies behind the Gemini family of models.
This does not turn Apple Intelligence into a conventional third-party cloud service. Apple is trying to keep the user-facing privacy claim intact while using outside accelerator capacity for larger inference workloads. For NVIDIA, the deployment gives its confidential GPU stack a high-profile consumer AI reference, not just another enterprise security use case.
As previously reported by eeNews Europe when NVIDIA and Google Cloud targeted physical AI factories, Google has already been previewing Gemini on Google Distributed Cloud running on NVIDIA Blackwell and Blackwell Ultra GPUs, along with Confidential G4 VMs for AI workloads that need hardware-backed protection.
The Apple PCC deployment takes the same broad direction into personal AI: more cloud inference, but with stronger claims around attestation, isolation, and operator access. Apple says PCC on Google Cloud will ramp towards the complete set of protections during the summer preview period. More technical detail is expected at the Confidential Computing Summit in San Francisco on June 23 and 24, 2026, and in an updated PCC Security Guide later this year.
If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。