The End of an Era at Apple Park

The opening moments of the keynote carried the weight of history. Cook, wearing his signature dark blue shirt and jeans, began with a personal reflection on his tenure. "When I became CEO in 2011, Apple was valued at $350 billion," he said, his voice steady against the backdrop of the Steve Jobs Theater's curved glass walls. "Today, we stand at $4.2 trillion. But numbers are not the legacy I care about. What matters is that we have put tools of creation into the hands of billions."

The confirmation of his departure, long rumored but never explicitly stated, sent Apple's stock up 2.3% in after-hours trading. Cook will remain chairman of the board, but day-to-day operations will transition to Johny Srouji, currently senior vice president of hardware technologies, who was formally named chief executive officer effective January 1, 2027. Srouji, an engineer by training who joined Apple in 2008 after a decade at IBM, has led the development of every Apple silicon chip since the A4.

"Srouji is the most important person at Apple that most people have never heard of," said Ben Bajarin, CEO of Creative Strategies. "He built the engine. Now he gets to drive the car."

Apple Intelligence 2.0: On-Device Reasoning

The centerpiece of the keynote was Apple Intelligence 2.0, a suite of AI capabilities that Apple claims will run entirely on-device for the first time. The system is built around a new family of models called Apple Reasoning Engine, or ARE, which can perform multi-step logical deduction without sending data to cloud servers. During a live demonstration, an iPhone 17 Pro solved a complex travel planning problem, comparing flight prices, hotel availability, and weather forecasts across 14 variables, in 1.2 seconds without any network connection.

"This is not pattern matching," said Craig Federighi, Apple's senior vice president of software engineering. "This is genuine reasoning, and it happens in your pocket with zero data leaving your device." The ARE models range from 3 billion to 30 billion parameters, with the smaller variants running on iPhones and the larger ones on Macs with M5 chips. Apple claims the 3-billion-parameter model outperforms GPT-4 on certain reasoning benchmarks while using 1/500th of the energy.

The privacy implications are profound. Because no data leaves the device, Apple can offer AI capabilities in regions with strict data localization laws, including China and the European Union, without building separate infrastructure. "Apple just solved the regulatory problem that has plagued every other AI company," said Carolina Milanesi, president of Creative Strategies.

Siri Reborn with Persistent Memory

Perhaps the most visible change for consumers is the complete rebuild of Siri. The voice assistant, long criticized as lagging behind Amazon's Alexa and Google Assistant, has been rearchitected around a persistent memory system that maintains context across conversations, devices, and even time. Siri now remembers that you prefer aisle seats on flights, that your daughter is allergic to peanuts, and that you typically call your mother on Sunday evenings.

During the demonstration, Federighi asked Siri to "book a restaurant for my anniversary." Without further prompting, Siri identified that his anniversary is June 22, cross-referenced his wife's dietary preferences from previous restaurant bookings, filtered for restaurants with availability on that date, and presented three options within a 15-minute drive of his home. The entire interaction took eight seconds.

"Siri has gone from a voice-activated search box to a genuine personal assistant," said Gene Munster, managing partner at Deepwater Asset Management. "The gap between Apple and its competitors in natural language understanding just collapsed."

The Google Partnership Nobody Predicted

In a move that stunned the technology industry, Cook revealed that Apple had signed a multi-year agreement with Google to integrate Gemini models as a fallback option when on-device processing is insufficient. The partnership, negotiated in secret over 14 months, means that iPhone users will be able to access Google's most advanced AI capabilities directly through Apple's interface, with Apple handling the billing and user experience layer.

"We compete fiercely with Google in many areas," Cook said. "But when it comes to serving our users with the best possible AI, we do not let corporate rivalry stand in the way." The deal reportedly involves a revenue-sharing arrangement where Apple keeps 70% of subscription revenue from Gemini-powered features on iOS, with Google receiving the remainder.

Developer Tools and Market Reaction

For the 6,000 developers attending in person and the 12 million watching online, Apple introduced Xcode AI, a coding assistant that understands entire codebases rather than just the current file. During a live coding session, an Apple engineer used natural language to instruct Xcode AI to "refactor this networking layer to use async/await," and the system modified 47 files across three modules in under 30 seconds, preserving all existing tests.

Wall Street responded cautiously. Apple's stock rose 1.8% during the keynote but gave back half those gains by the market close, as analysts debated whether the AI features would drive the upgrade cycle that Apple needs to reverse three consecutive quarters of declining iPhone revenue. "The technology is impressive," wrote Erik Woodring of Morgan Stanley in a note to clients. "But consumers have shown they will keep their phones for four or five years if the improvements are incremental. These need to feel revolutionary at the retail level."

Apple Intelligence 2.0 will ship with iOS 19, macOS 16, and iPadOS 19 this September, with a public beta beginning in July. The features require an iPhone 17 Pro or later, an iPad with M3 or later, or a Mac with M4 or later, meaning the full experience will initially be available to fewer than 15% of Apple's installed base.