Apple's Private Cloud Compute Infrastructure Sits Idle as Apple Intelligence Struggles with Low Adoption

Summary

Apple's ambitious Private Cloud Compute infrastructure is facing significant underutilization, with the company using only about 10% of its AI server capacity on average, according to a new report from The Information. Some already-manufactured Apple servers are reportedly sitting unused on warehouse shelves due to lower-than-expected adoption of Apple Intelligence features. The issue stems from multiple challenges: the system is underpowered for running frontier AI models, software updates are complex and time-consuming, and the current M2 Ultra-based chips lack the computational power needed for Apple's upcoming Gemini-based Siri chatbot.

The report reveals that Apple's cloud infrastructure suffers from severe fragmentation, with different teams running independent systems rather than sharing a centralized resource pool. This architectural inefficiency has frustrated Apple's finance team, which is concerned about the costs of duplicate infrastructure but reluctant to invest billions in a complete overhaul. Several attempts to unify the infrastructure over the past decade have reportedly stalled.

As a result, Apple is now in advanced talks with Google to host the new Siri models in Google's data centers while maintaining Apple's privacy standards. Google already provides some iCloud infrastructure for Apple and has extensive experience with large-scale LLM deployments through its Gemini product. While Apple may invest more heavily in in-house infrastructure going forward, implementing such changes represents a much longer-term project than the immediate needs of its AI ambitions.

• Apple is negotiating with Google to host new Siri models in Google data centers due to Private Cloud Compute's limitations
• Lower-than-expected Apple Intelligence adoption has cast the expensive Private Cloud Compute buildout in a negative light internally