Apple uses only 10% of its own AI servers: Equipment sitting idle in warehouses

Apple has run into serious difficulties in developing its own cloud platform, Private Cloud Compute, designed to handle Apple Intelligence requests. According to The Information, the company’s specialized servers are operating at an average of just 10% capacity, and some of the purchased hardware remains unused in warehouses. This raises questions about Apple’s strategy of building a fully closed and private AI infrastructure.

Why the Servers Are Sitting Idle

The main issue is fragmentation within Apple’s internal teams. Different divisions operate with isolated technology stacks, making it difficult to redistribute resources between projects. The company’s finance department is also reportedly dissatisfied with rising infrastructure costs and is unwilling to approve multi-billion-dollar investments to rebuild it.

Technical limitations are making matters worse. Private Cloud Compute servers are built on modified M2 Ultra processors, which, according to sources, do not deliver sufficient performance to efficiently run large language models. Software updates require significant time and effort, and user interest in Apple Intelligence features remains moderate — many customers simply are not using advanced AI capabilities.

Talks with Google and Searching for a Solution

Inside the company, there are discussions about potentially hosting new versions of Siri on Google’s infrastructure. This does not mean abandoning Apple’s own platform entirely, but rather taking a pragmatic step: Google already has proven experience in large-scale LLM infrastructure deployment through its Gemini project. Apple continues to search for the optimal configuration for its AI workloads, but so far it has been unable to utilize even a quarter of its existing capacity.

What This Means for Users and Apple’s Strategy

Private Cloud Compute was positioned as a key advantage — full privacy, with request processing that does not rely on external servers. However, if servers are sitting idle and Apple is considering partnerships with Google, it could undermine confidence in promises of “complete privacy.”

For everyday users, this is not yet critical — Apple Intelligence continues to function both on-device and in the cloud. But the low utilization rate suggests that Apple’s plans to build an independent AI empire are not going according to schedule. In the long term, this could lead to a strategic shift: more cloud partnerships and fewer proprietary data centers.

In Brief

Apple is using only 10% of its Private Cloud Compute cloud platform capacity, with some hardware sitting unused in warehouses. The reasons include internal team fragmentation, high costs, insufficient M2 Ultra performance for large models, and limited user interest in Apple Intelligence. The company is discussing hosting new versions of Siri on Google’s infrastructure. This calls into question Apple’s strategy of full AI independence and may result in a reassessment of plans for its own data centers.