Apple licensing Gemini
Apple is reportedly licensing Google’s Gemini model for on‑device features — a sign that intelligence is becoming infrastructure and companies may rent frontier models instead of building them all in‑house. This could compress AI economics and change platform strategy across Big Tech, including how Apple balances privacy with rented capabilities reported.
Bloomberg first reported that Apple is set to pay roughly $1 billion per year to license a custom Gemini model for Siri and Apple Intelligence. cnet.com Multiple outlets cite the custom model as roughly 1.2 trillion parameters, a scale that dwarfs the parameter counts Apple has publicly disclosed for its in‑house models. scmp.com Reporting describes a hybrid deployment that places core Gemini inference in Apple’s Private Cloud Compute while surfacing personalized inference on iPhone hardware for privacy‑sensitive queries. macrumors.com Bloomberg and others say Apple evaluated OpenAI’s and Anthropic’s models during selection, then settled on Google after comparative testing earlier in 2025. businesstimes.com.sg Apple and Google characterized the arrangement in a joint statement as a “multi‑year” collaboration with Gemini powering Apple Foundation Models and Siri’s next overhaul later in 2026. cnbc.com Public messaging around the deal used three explicit levers—an anchor timeline (“later this year”), a quantified commitment (~$1B/year), and partner naming (Google/Gemini)—seen across CNBC, CNET, and Bloomberg coverage. cnbc.com
