Siri Overhaul WWDC 2026
Published: June 7, 2026 | Updated: June 7, 2026
Apple will unveil a completely rebuilt Siri at WWDC 2026 tomorrow — powered by a custom 1.2-trillion-parameter Google Gemini model at a reported cost of ~$1 billion per year — alongside iOS 27, a standalone Siri app with conversation history, Dynamic Island integration, and agentic extensions that let it complete multi-step tasks across third-party apps. This is Tim Cook’s final WWDC keynote as CEO; John Ternus takes over September 1.
What Happened
Apple’s annual Worldwide Developers Conference opens Monday, June 8 at 10 AM PT at Apple Park. Bloomberg’s Mark Gurman, corroborated by TechCrunch, 9to5Mac, and MacRumors, reports the keynote will center on a rebuilt Siri powered by a custom Google Gemini model — a 1.2-trillion-parameter mixture-of-experts LLM Apple is paying roughly $1 billion per year to license. Alongside the Siri overhaul, Apple will announce iOS 27, iPadOS 27, macOS 27, watchOS 27, visionOS 27, and tvOS 27, with developer betas expected the same afternoon.
Why It Matters
The coverage of tomorrow’s Siri announcement has focused almost entirely on the Google Gemini deal — the partnership, the price tag, the model specs. That framing misses the more significant story.
Apple is the only major technology company among Google, Microsoft, Meta, Amazon, Anthropic, and OpenAI that does not own a frontier AI model. Every one of its competitors committed years ago to the thesis that owning the model is strategically essential — and spent tens of billions of dollars building that capability. Apple spent those years arguing that on-device processing, privacy, and hardware distribution were the better bet.
Tomorrow’s Gemini announcement is Apple conceding that thesis was insufficient. Not wrong, exactly — on-device AI is genuinely differentiated and the privacy angle remains real — but insufficient to close the capability gap Siri has accumulated since 2022. The ~$1 billion annual payment to Google is not simply a licensing fee; it is the cost of two years of strategic miscalculation on AI infrastructure, paid to a company that Apple has treated as a search competitor for the better part of a decade.
The architecture itself is notable. Bloomberg reports the model runs in a hybrid configuration: cloud inference on Google Cloud using Nvidia Confidential Computing (a hardware-level privacy isolation layer), while Apple’s own smaller on-device models handle privacy-sensitive tasks locally. The mixture-of-experts design activates only a relevant subset of the 1.2 trillion parameters per query, keeping inference costs manageable. This architecture — large cloud model plus small on-device model, with a privacy partition between them — is the closest Apple has come to matching what Google and Microsoft can do with Gemini and Copilot, without building the underlying infrastructure itself.
For the roughly 1.4 billion active iPhone users, the practical change is significant: the first version of Siri that can hold a multi-turn conversation without resetting context, access personal data (emails, photos, calendar, messages) to answer questions like “what’s the restaurant my colleague mentioned last Tuesday?”, understand what’s on screen and act on it without manual copying, and complete multi-step tasks across apps. These were all promised at WWDC 2024. None shipped. Tomorrow is the second attempt.
One framing detail no outlet is leading with: Apple’s WWDC 2026 developer site carries the tagline “All systems glow” — a reference to the new Dynamic Island-anchored Siri visual design, but also an implicit acknowledgment that the WWDC 2024 and 2025 promises did not land.
What Comes Next
Tomorrow, June 8: Keynote at 10 AM PT. Developer betas of all six OSes expected the same afternoon. The first real look at whether the rebuilt Siri performs as described in controlled demos — or whether it carries the same qualifications that muted the iOS 18 Apple Intelligence launch.
The two scenarios:
Scenario A — Siri 2.0 delivers: If Apple demos on-stage agentic tasks (booking a restaurant, composing and sending an email from a voice command, cross-app workflows) that work consistently and without visible latency, the Gemini deal will be validated immediately and Apple’s distribution-over-model strategy gets a serious second look from investors who have been skeptical. Watch for whether Cook demos these features himself — Apple executives historically demo only what they’re confident works.
Scenario B — Staged demo with September caveats: If the on-stage demo is carefully scoped — chatbot capabilities available now, agentic features “coming in a future update” — the pattern from iOS 18 repeats. The real test then shifts to September’s iPhone 17 launch, where Ternus will need Siri to perform at the level Cook promises tomorrow.
The John Ternus context matters most here. Ternus is a hardware engineer. His success as CEO depends on the iPhone 17 and, according to multiple reports, a foldable iPhone either in 2026 or early 2027. Both of those products need Siri 2.0 to work — the case for upgrading from an iPhone 15 or 16 gets materially weaker if the AI experience doesn’t match the hardware. Whatever Cook announces tomorrow, Ternus owns the delivery.
Axis Intelligence will update this article during and after the June 8 keynote.