WWDC 2026 Preview: Apple’s AI Strategy Is the Opposite of Everyone Else’s

Apple’s WWDC keynote is Monday, June 8. It will be Tim Cook’s final WWDC as CEO — he hands the role to John Ternus on September 1. And the centerpiece of every announcement will be a completely rebuilt Siri powered by Google’s Gemini.

That last sentence contains the most important strategic decision Apple has made in a decade. Let’s unpack why.

Everyone Else Builds Models. Apple Builds Distribution.

OpenAI built GPT. Anthropic built Claude. Google built Gemini. Microsoft built MAI. Meta built Llama. Nvidia built Nemotron. Every major tech company decided that owning a frontier AI model was strategically essential.

Apple chose the opposite. Instead of building its own frontier model, Apple signed a ~$1 billion/year deal with Google for a custom Gemini model. Apple distills Gemini into a smaller version that runs on-device. Server-side processing runs on Google Cloud with Nvidia Confidential Compute. And iOS 27 will let users choose their default AI provider — ChatGPT, Claude, Gemini, Grok, Copilot, or Perplexity.

This is not a failure to build. It’s a deliberate strategic choice: Apple wants to be the AI distribution platform, not the AI model provider. Every AI company pays for the privilege of being on 2 billion Apple devices. Apple controls the interface. The model is a replaceable input.

What’s Being Announced

LLM Siri. Craig Federighi confirmed Apple scrapped the hybrid approach (legacy Siri + LLM) and rebuilt from scratch. New standalone Siri chatbot app with conversational UI. On-screen awareness — Siri sees what you’re looking at. Conversation memory across sessions. Multi-step task execution via App Intents. Camera integration — dedicated Siri mode alongside Photo, Video, Portrait.

iOS 27. Foldable iPhone software support (side-by-side apps, iPad-like multitasking on iPhone). Visual Intelligence expansion (scan nutrition labels, phone numbers, addresses). Third-party AI defaults for Writing Tools and Image Playground. Liquid Glass UI refinements after last year’s complaints.

Hardware. Mac Studio with M5 Max/Ultra. Mac mini with M5. Possibly iMac with M5. HomePad (7-inch smart home hub) preview. iPhone Fold software foundation — hardware in September at ~$2,500.

Developer tools. Xcode 17 with MCP (Model Context Protocol) support — Claude and Codex integrated natively for coding assistance. Apple embracing third-party AI agents in its own development environment.

The Strategic Comparison

Three approaches to AI, three philosophies:

Microsoft (Build 2026): Build your own models (MAI), own the orchestration layer, make OpenAI a replaceable input. Enterprise-first. Agent-first. The harness is the moat.

Google (I/O 2026): Build the best model (Gemini), put it everywhere — Search, Android, Workspace, glasses, cars. Model-first. The model is the moat.

Apple (WWDC 2026): Don’t build the model. Control the 2 billion devices where the model runs. Let users choose. The distribution is the moat.

Apple’s bet is that the model layer commoditizes — exactly what’s happening as GPT-5, Claude 4, and Gemini 3.5 converge in capability. If models become interchangeable, the company that controls the interface where users access them captures the value. That’s Apple’s historical playbook: don’t make the content, own the store.

Tim Cook’s Swan Song

This is Cook’s final WWDC as CEO after 15 years. His legacy: doubling Apple’s revenue, building the services business to $109 billion, and navigating the company through the most disruptive technology shift since the smartphone. Whether the Gemini bet and the distribution-over-model strategy is the right call will be determined over the next 3-5 years — but it’ll be John Ternus’s problem to solve, not Cook’s.

Monday 10 AM PT. The last keynote of the Tim Cook era. We’ll have full coverage within hours.

For the full structural map of the AI economy, read The Map of AI Redrawn on Business Engineer.