Apple Pivots to Google Cloud for Siri
Apple is reportedly shifting its AI strategy, asking Google to host future versions of Siri on its data centers using the Gemini model. The move comes as Apple's own AI servers are reported to be sitting idle due to low usage. Apple is also said to be in talks with OpenAI and Anthropic, signaling a major pivot from its in-house, on-device focus.
The pivot away from a strictly in-house, on-device AI strategy has been developing for some time. Under former AI head John Giannandrea, who retired in early 2026, Apple championed on-device processing for privacy and performance, viewing competitors' reliance on the cloud as "technically wrong." This philosophy is evident in the Neural Engine in Apple Silicon, designed for efficient local execution of machine learning tasks. However, the massive computational and infrastructure costs required to train and run cutting-edge large language models have posed a significant challenge. Reports indicate that Apple's own Private Cloud Compute infrastructure, which uses modified M2 Ultra processors, is both underpowered for frontier models and underutilized, with only about 10% of its capacity being used on average. This has led to manufactured servers sitting idle in warehouses. The reported deal with Google, potentially worth $1 billion annually, provides access to Google's powerful Gemini models and the massive-scale infrastructure needed to run them. This allows Apple to accelerate its AI feature roadmap for a revamped Siri without incurring the multi-billion dollar capital expenditures seen at rivals like Google, Microsoft, and Meta, whose AI-related spending is surging. This move signals a pragmatic shift, prioritizing speed to market and user experience over maintaining a purely internal AI stack. The strategy leverages a hybrid approach: continue developing on-device capabilities for privacy-sensitive tasks while outsourcing the most demanding, large-scale generative AI processing to a proven leader. It's a calculated decision to focus on Apple's core strength—hardware and software integration—while treating the underlying large language model as a component that can be sourced externally. This partnership follows a leadership shake-up, with Amar Subramanya, formerly of Google and Microsoft AI, taking over as VP of AI. The internal AI group, reportedly nicknamed "AIMLess" due to indecision and delays on the Siri upgrade, is now being restructured under software chief Craig Federighi, services head Eddy Cue, and COO Sabih Khan. Apple continues to invest in its own AI infrastructure, including a secret chip initiative, "Project ACDC" (Apple Chips in Data Centers), and a significant expansion of its Houston manufacturing campus to produce AI servers. The company has pledged a $600 billion investment in U.S. manufacturing and AI infrastructure over the next four years. This suggests a long-term goal of reducing reliance on partners, even as it currently leans on them to close the gap.
