Apple reportedly to pay Google ~$1B/year to power parts of the new Siri

Recent reporting indicates Apple plans to pay Google roughly $1 billion per year for a customized version of Google’s Gemini model to power key components of its redesigned Siri. That Google‑derived model would run on Apple’s Private Cloud Compute and handle Siri’s summarizer and planner functions — systems that synthesize information and sequence multi‑step tasks.

Apple reportedly intends the Gemini layer as a temporary boost: the company is said to be developing its own very large cloud model (around one trillion parameters) to replace Google’s tech once it reaches production readiness. The arrangement would let Apple accelerate Siri’s capabilities while it builds out its in‑house models.

What’s changing for Siri

• Role of Gemini: Summarization and planning duties — enabling Siri to compile information and plan multi‑step actions across apps and services.
• Hybrid architecture: Apple will combine Google’s cloud model with its own on‑device and server models to balance capability, latency and privacy.
• Cost & scale: The reported fee is about $1 billion per year, underscoring the strategic priority Apple places on upgrading its assistant.

Why Apple might do this

Using an established, high‑capability model from Google lets Apple ship more advanced assistant features sooner rather than waiting for an in‑house model to catch up. That pragmatism can improve user experience in the short term but creates a temporary dependency on a rival’s technology.

Risks and questions

• Privacy & data flow: Apple plans to run the custom model on its private cloud, reducing cross‑company data exposure, but observers will want detailed assurances about what data is logged, how prompts are retained, and what prevents leakage.
• Transparency: Reports say Apple won’t advertise the Google link; users may expect clarity on which models handle their requests.
• Vendor dependency: Short‑term reliance on a competitor poses strategic risk if terms change; Apple’s plan to replace the layer with its own trillion‑parameter model aims to address that over time.

Timeline & context

Apple has explored partnerships with multiple model providers and delayed Siri’s relaunch to refine safety and capabilities. Bloomberg’s reporting suggests Apple hopes to have an extremely large cloud model ready within the next year to phase out the Google‑based layer — though exact timelines could shift.

Implications

This deal — if accurate — illustrates how fierce competition and rapid advances in AI are reshaping even tightly integrated tech stacks. Companies are increasingly mixing internal and external models to accelerate product roadmaps while trying to preserve user privacy and control.

For further reading, see Bloomberg’s reporting and related coverage (opens in a new tab): Bloomberg.

Discussion: Would you prefer Apple ship a more capable Siri now using a Google‑derived model, or wait until Apple’s own model is ready? What privacy or transparency guarantees would make you comfortable with a hybrid approach?