Apple's prized "privacy fence" is facing unprecedented infrastructure challenges. On March 2, 2026, according to leaked documents and industry intelligence disclosed by The Information, Apple is re-evaluating its cloud infrastructure strategy.

Due to the weak performance of its internal "Private Cloud Compute (PCC)" platform when running the new Gemini-powered Siri, Cook's team may seek deeper-level computing power support from its old rival, Google.

Core Issue: M2Ultra Faces an "Intellectual Ceiling"

Apple initially tried to balance AI performance and privacy with its self-developed "Apple Chip Data Center," but the reality has been harsh:

Insufficient Hardware Performance: The current PCC servers mainly rely on modified M2Ultra processors. While they perform well in consumer applications, they are far less efficient and have lower throughput than specialized AI acceleration chips (such as NVIDIA H200/B200 or Google TPUs) when handling large-scale models like Gemini.

Extremely Low Resource Utilization: Due to lower-than-expected usage of the first batch of Apple Intelligence features, many purchased servers are now "dust collecting" in warehouses. A fragmented internal technical structure has also caused serious waste of computing resources.

Software Update Traps: The PCC runs a highly customized and restricted operating system. Each software iteration is extremely complex and time-consuming, making it difficult to keep up with the rapid evolution of AI, which often updates weekly.

Strategic Shift: Using Google Servers for Siri?

Faced with complaints from the finance team about high maintenance costs, and the expected traffic surge brought by the new version of Siri launching at the end of 2026, Apple is considering alternative solutions:

Renting Google Servers: Apple may ask Google to set up dedicated server clusters within its data centers, following Apple's privacy protocols, to directly run the new Gemini-powered Siri.

Privacy Negotiation: Although Apple insists on "no compromise on privacy," if its own computing power cannot support Siri's logical reasoning and multi-step task execution, outsourcing traffic to more mature infrastructure providers like Google Cloud may become an unavoidable move in the face of AI competition.

Industry Observation: "Hard Power" Anxiety in the AI Era

Apple had long used Google Cloud and Amazon AWS for non-core services such as iCloud storage. But if even the "brain" of Siri has to rely on competitors' data centers, this is not just a financial decision, but also a major blow to Apple's "software and hardware integration" myth. This also explains why recent rumors suggest Apple is accelerating the development of a new AI server codenamed J226C, equipped with the M5 chip