Apple licenses a custom 1.2-trillion-parameter Google Gemini model to power the rebuilt Siri at ~$1B per year — largest model-licensing deal on record reorganizes the Apple-Google-OpenAI triangle
WWDC 2026's Siri rebuild runs on a custom 1.2T-parameter Gemini variant licensed from Google for roughly $1 billion annually, ending years of speculation that Apple would ship its own frontier model. It's the largest model-licensing arrangement on record and effectively concedes the frontier-training contest to Google while keeping Apple at the inference, distribution, and on-device personalization layer.
The substantive piece is the strategic concession. Apple has spent four years publicly positioning around an internal frontier-model program; the WWDC 2026 announcement that the rebuilt Siri runs on a custom 1.2T-parameter Gemini variant terminates that narrative. The choice of Google over OpenAI is the secondary signal — Google's Gemini family had been the third option in most enterprise multi-vendor evaluations through Q1 2026; landing the Apple deal moves Gemini into category-leader position for distribution-anchored inference partnerships.
The competitive read against the SpaceX-xAI $1.25T merger is that two structurally-different model-deployment strategies are arriving in the same week: Apple bets on inference-layer ownership with licensed frontier capability, while Musk bets on vertical integration of compute + model + launch infrastructure. The H2 2026 frontier-AI distribution landscape will be shaped by which strategy generates the stronger commercial pipeline through Q3-Q4 evaluations.
Medium — Weekly Tech Talk June 15, 2026: Apple's Secret Strategy → · Build Fast With AI — AI News Today June 7 2026 → · CNBC — Big Tech earnings 2026: AI spend trajectory →