Apple outsources Siri: AI servers remain unused

This article may contain affiliate links. Learn more on our Partner Program page.

Make Apfelpatient a preferred source One click – and you'll see us more often on Google

Apple has stood for vertical integration and maximum control over its own infrastructure for years. Especially in the field of artificial intelligence, Apple Intelligence was intended to mark the beginning of a new chapter: powerful models, operated in a privacy-friendly manner on Apple's own hardware.

However, recent reports paint a different picture. Parts of Apple's AI servers are apparently sitting unused on warehouse shelves. At the same time, Apple is in advanced talks with Google to run new Siri models in their data centers.

The underlying reasons reveal structural problems in Apple's cloud strategy, technical limitations of its own infrastructure, and an AI landscape that is changing faster than internal processes can keep up.

Apple Intelligence and the claim to its own infrastructure

With Apple Intelligence, Apple aims to deeply integrate AI capabilities into its own ecosystem. The technical foundation is intended to be Private Cloud Compute – a server infrastructure developed by Apple specifically designed for privacy-sensitive AI processing.

The goal was clear: to provide AI services not primarily via external cloud providers, but via in-house hardware.

However, according to several reports, including those from Bloomberg and The Information, this very approach is currently reaching its limits.

Low occupancy: Only about 10 percent of capacity is used

A key problem is the utilization of Apple's existing AI infrastructure. Reports indicate that the company is only using an average of around 10 percent of its available capacity.

This has concrete consequences: Some Apple servers that have already been produced are sitting unused on warehouse shelves.

The reason appears to be that the initial Apple intelligence features were used significantly less than expected. Demand fell short of forecasts. In this context, the expansion of private cloud computing seems excessive.

Technical limitations of private cloud computing

Besides the low utilization, there are technical problems. Private cloud computing is described as undersized.

The servers are currently said to be based on modified M2 Ultra processors. However, these chips are considered insufficiently powerful to run modern Frontier models like Gemini. It is precisely such models that are expected to form the basis for the new Siri.

Additionally, the system is described as requiring extensive maintenance. Software updates are considered complicated and time-consuming. The effort is apparently disproportionate to the benefit.

While Apple anticipates significantly higher demand for new Siri chatbot features in the future, the current private cloud compute stack does not appear capable of running these models.

Fragmented cloud infrastructure at Apple

Another structural problem lies in the organization of Apple's cloud infrastructure.

According to the report, the infrastructure is highly fragmented. Different teams operate their technologies independently. There is no central pool of resources that all departments can access. This results in inefficiencies.

Parts of the infrastructure remain unused.
Other departments could use additional server capacity, but do not have access to it.

Apple's finance team is reportedly growing increasingly frustrated with the costs of this duplicated infrastructure. Nevertheless, there appears to be no willingness to invest billions in a fundamental overhaul of the entire stack.

Several attempts to unify the systems have been launched in the past ten years. However, these projects have repeatedly stalled.

Outsourcing to Google: Siri is to run in Google data centers

Earlier this year, Bloomberg reported that Apple planned to host the heavily delayed new Siri models on Google servers – instead of on its own private cloud compute infrastructure.

The Information now confirms a similar development. According to the report, Google has been contracted to operate Siri servers in its data centers. Google is expected to adhere to Apple's data privacy standards.

Google, through its Gemini platform, has extensive experience in building and operating large LLM server infrastructures. This expertise makes the company a natural partner for scalable AI workloads.

This collaboration isn't entirely new. Apple already uses Google Cloud for certain iCloud functions, for example in the area of cloud storage.

Strategic adaptation in a rapidly changing AI landscape

The AI industry is currently evolving rapidly. Frontier models are becoming more powerful, requirements are increasing, and investments need to be flexibly adapted. For Apple, this apparently means a reassessment of its cloud strategy.

While outsourcing to Google is a possibility in the short term, a more significant internal expansion could follow in the long term. However, a comprehensive modernization of the infrastructure would be a long-term process and would require substantial investment.

The current move suggests that Apple is reacting pragmatically to remain competitive – even if that means resorting to external data centers for core AI functions.

Apple between its own cloud infrastructure and Google partnership

Apple is caught between the demands of its own infrastructure, cost control, and technological advancement. Low utilization of private cloud compute, unused AI servers on warehouse shelves, technical limitations of its M2 Ultra-based systems, and a fragmented cloud architecture highlight structural challenges.

At the same time, Apple is apparently planning to run new Siri models in Google data centers, as the existing infrastructure is insufficient to support modern Frontier models like Gemini. (Image: Shutterstock / Frame Stock Footage)