Apple’s AI Servers Gather Dust in Warehouses as Low Demand Stalls Apple Intelligence Use

Apple’s Private Cloud Compute (PCC) was intended to be the linchpin for the next generation of Siri, yet internal reports suggest the platform is far from delivering on that promise. According to 9to5Mac, most of the custom‑built AI servers—equipped with modified M2 Ultra chips—are now languishing on warehouse shelves, with average utilization hovering at roughly 10 percent. The under‑use is not merely a logistical hiccup; it reflects a deeper mismatch between Apple’s hardware rollout and the actual demand for its “Apple Intelligence” features, which have yet to achieve the adoption rates the company projected.

The capacity shortfall is compounded by a fragmented cloud architecture that forces each product team to operate in isolation. The Information notes that Apple’s various divisions run separate stacks rather than a unified pool, creating “idle” resources in one area while another struggles for compute. Finance executives have reportedly expressed frustration over the duplicated infrastructure costs, yet the company remains reluctant to pour billions into a comprehensive overhaul—a sentiment echoed by multiple internal attempts to consolidate the stack that have stalled repeatedly over the past decade.

Performance constraints further diminish PCC’s viability. Bloomberg earlier reported that Apple was already exploring the use of Google’s data centers to host the delayed new Siri models, a move confirmed by The Information, which says Google has been tasked with running Siri servers while adhering to Apple’s privacy standards. The same source describes PCC as “underpowered” for cutting‑edge models such as Gemini, the large language model that will underpin the upcoming Siri chatbot. Updating the proprietary software on Apple’s custom hardware is reportedly “trickier and takes time,” making the platform less agile than the public‑cloud alternatives that Google already operates at scale.

The low utilization figures have strategic implications for Apple’s AI roadmap. The company’s initial rollout of Apple Intelligence features—ranging from on‑device translation to context‑aware suggestions—has not generated the usage spikes Apple anticipated, leaving PCC’s buildout in a negative light. As a result, Apple is reportedly in “advanced talks” with Google to shift the new Siri workloads to Google Cloud, leveraging the latter’s experience with mass LLM deployments. This pivot underscores a broader shift in Apple’s approach: while the firm may still invest in in‑house AI infrastructure over the long term, immediate pressures are forcing it to rely on external partners to meet performance and capacity demands.

The situation also raises questions about Apple’s broader AI ambitions ahead of its March product event, where analysts expect announcements ranging from the iPhone 17 e to a potential new MacBook M5. If Apple cannot reliably scale its AI services internally, the company may need to lean more heavily on third‑party clouds for future features, potentially eroding the privacy narrative that has been a cornerstone of its brand. For now, the idle servers serve as a tangible reminder that Apple’s AI infrastructure is still a work in progress, and that the company’s ability to monetize its AI investments may hinge on how quickly it can reconcile its fragmented cloud strategy with the growing expectations of developers and consumers alike.