Report: Apple is developing at least six cloud infrastructure projects, driven partly by the suspicion that servers from third parties can be compromised
Jordan Kahn / 9to5Mac :
Context & Ripple Effects
In March 2016, 9to5Mac reported Apple was developing at least six internal cloud infrastructure projects, motivated partly by suspicion that third-party servers could be compromised — at a time when iCloud data still ran on rented capacity. The suspicion proved well-founded as a description of the status quo: two years later, Apple's own iOS Security Guide confirmed Google Cloud Platform was storing iCloud data.
What makes this report worth revisiting is where the program landed. The internal build-out eventually surfaced publicly as Private Cloud Compute, running on servers with Apple silicon whose code independent experts can inspect, then opened further through a public research environment and vulnerability bounties. By 2024–2025 the same infrastructure ambition had grown into an exploration of renting Apple's own chips to developers — notable given Apple's roughly $7B annual spend renting Nvidia chips from AWS and Google.
First-order effects
- Apple's iCloud and Siri services gain a path off third-party hardware, directly addressing the compromise risk the report cites — though follow-up reporting on infighting between the iCloud and Siri teams showed the internal build-out itself was slowing development.
Second-order effects
- Hyperscale landlords AWS and Google face gradual attrition of one of their largest tenants' workloads, converting rental revenue into Apple-owned capacity — and giving Apple leverage in future pricing negotiations.
Third-order effects
- If the arc from six secret projects to inspectable Apple-silicon clouds holds, Apple ends up owning a vertically integrated compute stack — its own silicon in its own datacenters, potentially resold — collapsing the distinction between cloud customer and cloud provider.
The trend: Major device makers are in-sourcing cloud infrastructure onto proprietary silicon, trading hyperscaler dependence for self-controlled, independently auditable compute.