On June 8, Apple announced it is extending Private Cloud Compute (PCC), its privacy-preserving cloud AI foundation, beyond its own data centers to run some Apple Intelligence workloads on NVIDIA GPUs hosted on Google Cloud.
June 9, 2026 · Apple Intelligence
Apple Extends Private Cloud Compute to NVIDIA GPUs on Google Cloud
The heaviest Apple Intelligence workloads—agentic tool-use and complex reasoning—will now run on NVIDIA Blackwell B200 GPUs inside Google Cloud, alongside Apple Silicon. Apple says its privacy guarantees are "ported" intact onto a third-party cloud for the first time.
B200
NVIDIA Blackwell GPUs powering the new Siri & high-load tasks
3
Confidential Computing layers: NVIDIA, Intel TDX, Google Titan
AFM 3
Gemini-based Cloud Pro foundation model
The compute base — from one platform to two
2024 · WWDC
Apple Silicon servers only
→
NVIDIA B200 / Google Cloud
Apple Silicon
2026 · Now
Two backbones, same guarantees
How privacy travels to a third-party cloud
Encrypted data + model loaded onto GPU
→
Confidential Computing shields it from hosts & admins
→
Stateless processing, no data retained
PCC principles: no privileged access · non-targetability · verifiable transparency · enforceable guarantees — designed so neither Google nor NVIDIA can view the data.
Praise from engineers
Combining Confidential Computing with Apple's verifiability means neither Google nor NVIDIA can see user data — some call cloud Gemini-based models a once-in-a-decade shift.
Skeptics caution
"Despite the name, it's Google Cloud plus NVIDIA B200 — privacy now rests on encryption," and the move "exposes the limits of Apple's vertical integration."
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article…
