WWDC 2026: Siri’s Reckoning and Tim Cook’s Final Keynote

Apple’s WWDC 2026 keynote opens Monday, June 8, at 10am Pacific time with a rebuilt Siri at center stage, iOS 27, and an AI infrastructure built around Google’s Gemini model. Developer betas of every operating system in the Apple family are expected the same day; the full public release targets September.

Tim Cook steps down as Apple’s chief executive on September 1 and hands the role to John Ternus, the company’s hardware engineering chief. The rebuilt Siri Cook is expected to demonstrate on Monday is the same upgrade Apple announced at WWDC 2024 and has been unable to deliver in the two years since.

Two Years of Broken Promises

Apple’s WWDC 2024 keynote showed a Siri that could understand what was on your screen, pull context from emails and photos, and chain commands across apps from a single prompt. The iPhone 16 launched that fall with Apple Intelligence as a marketing centerpiece; Apple ran TV commercials advertising those features for months before they were ready to ship. What arrived in iOS 18 was a glowing new interface, a ChatGPT integration via OpenAI, and a text-based option called Type to Siri. The conversational, context-aware assistant from the demo didn’t ship.

Bloomberg reported in March 2025 that Craig Federighi, Apple’s software chief, and other senior executives raised strong concerns during their own internal testing that the Siri features didn’t work as advertised. Some engineers in Apple’s AI division discussed scrapping the work and rebuilding from scratch. Apple brought in veteran software manager Kim Vorrath to help stabilize the effort. The company’s public statement acknowledged the problem: “It’s going to take us longer than we thought to deliver on these features and we anticipate rolling them out in the coming year.”

A second Bloomberg report, in June 2025, set a new internal target of spring 2026 via an iOS 26.4 update. That slipped. 9to5Mac reported in February 2026 that Apple had spread the Gemini-powered upgrades across multiple iOS 26 point releases rather than a single shipping moment. Apple told CNBC that all promised improvements would arrive before the end of 2026. Apple’s next fall operating system update is now the target.

Amazon launched its upgraded Alexa+ assistant in early 2025; Samsung spent the same period integrating AI deeply into the Galaxy line. Apple held the largest installed base in premium smartphones through this stretch and spent it unable to ship the software its own marketing had already sold.

The New Siri’s Design and Capabilities

A New App, a New Surface

The rebuilt Siri in iOS 27 looks more like a ChatGPT or Claude app than the assistant Apple shipped fifteen years ago. Bloomberg’s leaked mockup of the standalone Siri app shows an “Ask Siri” bar at the top, a paperclip icon for attaching images, PDFs, and files, and conversations displayed as iMessage-style chat bubbles. A history panel holds past sessions as cards with conversation summaries.

Siri reportedly moves into the Dynamic Island, where a swipe surfaces a “Search or Ask” prompt that could absorb Spotlight as the phone’s primary search interface. Apple is planning three activation paths: the power-button hold, the “Siri” wake word, and the new swipe-down gesture from the Dynamic Island.

Visual Intelligence, previously accessed through the Camera Control button on iPhone 15 Pro and 16 models, moves into the Camera app as a dedicated mode alongside Photo, Video, Portrait, and Panorama. MacRumors reports the expanded Visual Intelligence will identify nutrition labels, scan contact information from business cards, and recognize objects in a scene, filling a gap that Google Lens has covered on Android for years.

What the Assistant Can Do

Pre-keynote reporting from Bloomberg, MacRumors, and AppleInsider describes the rebuilt Siri as capable of:

• Personal context access: reading across emails, messages, photos, and files to answer questions about the user’s own data, such as a flight time pulled from a confirmation email.
• On-screen awareness: acting on whatever is currently displayed, including summarizing a visible article, adding a phone number to Contacts, or drafting a reply to an open email thread.
• Cross-app task chains: a single prompt can pull a flight confirmation from Mail, create a Calendar event, and send a message to a relevant contact.
• File and image uploads: the standalone app accepts attachments, matching the interface pattern of ChatGPT and Claude.
• Model choice: users can reportedly set a default AI for Writing Tools and Image Playground, choosing from the native model, ChatGPT, Anthropic’s Claude, Grok, and Perplexity.

Apple Intelligence device requirements remain unchanged: iPhone 15 Pro, 15 Pro Max, or any iPhone 16 and later. iOS 27 supports the iPhone 12 and newer, dropping the iPhone 11, but the headline Siri features stay gated to the newer chip generations. The new OS is also expected to include software support for the upcoming foldable iPhone, with a split-view multitasking mode designed for the device’s larger unfolded screen.

A $1 Billion Deal with Google

The January Partnership

On January 12, 2026, Apple and Google issued a joint statement confirming a multi-year AI collaboration. The statement said the next generation of Apple Foundation Models would be based on Google’s Gemini models and cloud technology, with Gemini identified as “the most capable foundation” available. Google Cloud chief Thomas Kurian confirmed the arrangement publicly at Google Cloud Next 2026 in April, saying the Gemini-powered Siri remained on track for later in the year. Tim Cook addressed the deal on Apple’s Q1 2026 earnings call:

You should think of it as a collaboration. And we’ll obviously independently continue to do some of our own stuff.

Cook did not discuss the financial terms. Bloomberg’s Mark Gurman had already reported them: roughly $1 billion per year for a custom 1.2 trillion-parameter Gemini model, about eight times larger than Apple’s own 150-billion-parameter cloud foundation model. Apple registered the genai.apple.com subdomain on May 23, about two weeks before WWDC, signaling the AI platform is ready to move publicly.

The arrangement also opens the system to outside AI providers. A new Siri Extensions system lets third-party developers build integrations on top of the rebuilt infrastructure, and users will be able to pick their default AI for specific features, choosing from the native model, ChatGPT, Claude, and others, in the same way they currently change a default browser.

How the Data Moves

The infrastructure detail that surfaced this week sharpens the picture. The Information reported June 3 that Siri’s most demanding reasoning will route to a cloud environment running on Nvidia Blackwell B200 graphics processors. The B200 packs 208 billion transistors and a Transformer Engine that supports four-bit floating-point math for large-model inference, making it one of the few chips capable of serving a 1.2 trillion-parameter model at query speed. Nvidia’s confidential computing extension, first available on the B200, encrypts all data in GPU memory during inference, covering model weights, user input, and the output itself. Apple’s Private Cloud Compute framework wraps the exchange: queries are anonymized and not stored after processing, and model weights run on Apple-controlled servers rather than the cloud provider’s infrastructure.

Siri’s most demanding requests now route through hardware owned by two of Apple’s biggest competitors. Apple says the anonymization and encryption layers mean no user data is stored or retained after processing; developers and regulators will have their own tests of that claim once the software ships.

The Snow Leopard Strategy

Bloomberg’s Mark Gurman has described Apple’s next release as a “Snow Leopard moment,” a reference to the 2009 Mac OS X update Apple shipped primarily to stabilize code after a feature-heavy cycle. The Liquid Glass design language from iOS 26 stays but gains a transparency and contrast slider, addressing the readability complaints that mounted through last year. Keyboard improvements, Apple Maps with satellite imagery as the default view, and a Safari feature called Organize Tabs that groups open tabs automatically are among the expected additions.

The Photos app gains three on-device generative AI tools. Extend uses generative fill to expand an image’s edges outward. Enhance adjusts quality, lighting, and color in a single tap. Reframe shifts the apparent camera angle on spatial photos in post-capture editing. Apple’s existing Clean Up tool, which removes objects from scenes, reportedly gets a significant improvement alongside the new three.

macOS 27 follows the same broad priorities: stability work, Liquid Glass contrast refinements, and Siri integration. Intel Mac support ends with this release; Apple’s macOS Tahoe is the last version compatible with Intel hardware. High component costs, specifically RAM and chip prices, pushed M5 Mac Studio and Mac mini refreshes past their original schedule according to Apple’s most recent public statements, so new Mac hardware is not expected at WWDC. The macOS 27 name remains unconfirmed; MacRumors spotted “Big Bear” as a potential candidate in Apple’s WWDC hashmoji filename on X, though Apple hasn’t confirmed it.

watchOS 27 brings the Modular Ultra-style watch face to standard Apple Watch models and adds refined heart-rate tracking. An internal Apple project known as Project Mulberry, described in Bloomberg reporting as an AI health coaching system, has been scaled back from a standalone product and folded into the Health app. visionOS 27 picks up the Siri and Apple Intelligence upgrades landing across the rest of the OS family, and Apple confirmed in May that the headset will also add a Wheelchair Control feature, using the Vision Pro’s eye-tracking to operate a powered wheelchair.

John Ternus Inherits Apple’s AI Problem

Apple confirmed in April that Cook hands over the CEO title on September 1 and stays with the company as executive chairman. Ternus, senior vice president of Hardware Engineering since 2021 and a presenter of last year’s iPhone Air at WWDC 2025, takes a role held for the past 15 years, bringing a background built on the Apple silicon transition, the A-series neural engines, and the chip architecture on which on-device inference now depends.

The outgoing CEO’s tenure included roughly tripling Apple’s annual revenue and building the services business to $109 billion by 2025, per Bloomberg. Siri has not kept pace with ChatGPT or the broader generative AI wave since 2022, a gap Apple’s own engineers and executives have documented through two years of internal delays and public admissions.

Ternus becomes CEO at almost exactly the moment the rebuilt Siri is due in users’ hands. Apple has publicly committed to shipping before 2026 ends; any further delay lands on the new chief executive from his first weeks. Dow Jones Market Data shows Apple’s stock has historically dipped slightly on WWDC day before recovering an average of roughly 14% over the following three months, a pattern built over a decade of conferences that mostly delivered on their promises.

Developers at WWDC 2026 this week have more than 100 sessions covering the new Siri Extensions APIs, the OS 27-family frameworks, and Xcode 17, which reports say will include Model Context Protocol support integrating AI coding assistants into Apple’s native development environment. The rebuilt Siri and the CEO transition are both scheduled for September, which means the company has given itself exactly one more cycle to deliver what it first promised two years ago.