Tim Cook's Final Curtain Call: Apple Bets Its Future on AI, a Hardware CEO, and the End of Intel Macs

Apple's WWDC 2026 keynote on June 9 was, by any measure, a singular event in the company's history. Tim Cook used his final keynote as CEO to unveil an overhauled Siri assistant built on Google's Gemini models, announce macOS 27 "Golden Gate" as an Apple Silicon-only release, and confirm that John Ternus, the company's SVP of Hardware Engineering, will succeed him on September 1 [1][2]. Cook signed off by calling it the "honour of a lifetime," but the real question is whether the company he built into a $4 trillion colossus can keep its footing in an AI race where rivals have been shipping for over a year [3].

The Cook Era: 2,000% Returns and a $4 Trillion Company

When Cook replaced Steve Jobs in August 2011, Apple's market capitalization stood at roughly $350 billion with $108 billion in annual revenue [4]. By April 2026, Apple's market cap had surged past $4 trillion, and fiscal year 2025 revenue exceeded $416 billion [4]. The stock returned nearly 2,000% over Cook's tenure—roughly four times the S&P 500's 503% gain over the same period [4].

Source: TheStreet / Yahoo Finance

Data as of Jun 9, 2026CSV

Cook's legacy rests on three pillars: building Apple Services into a $31 billion-per-quarter business, launching category-defining products (Apple Watch, AirPods), and executing the transition from Intel to Apple Silicon [4]. His critics will note that the company under his leadership was consistently late to AI—a gap that WWDC 2026 explicitly tried to close.

Apple's R&D spending has grown from $16.2 billion in fiscal 2019 to $34.6 billion in fiscal 2025 [5]. But that figure, while impressive in absolute terms, pales against what competitors are spending specifically on AI infrastructure.

Source: Apple SEC Filings / Motley Fool

Data as of Jun 9, 2026CSV

Enter John Ternus: A Hardware Engineer for an AI Era

Ternus, 51, is a 25-year Apple veteran who joined the company in 2001 after a brief stint at Virtual Research Systems, a virtual reality startup [6]. He studied mechanical engineering at the University of Pennsylvania, where he also won competitive swimming titles in the 50-meter freestyle and 200-meter individual medley [6]. He became a VP of hardware engineering in 2013 and was promoted to SVP in 2021, overseeing the engineering of the iPhone, iPad, and Mac product lines [7].

His age mirrors Cook's when he took over—Cook was 50 in 2011—positioning Ternus for potentially a decade or more of leadership [6]. Cook will remain as Executive Chairman [8].

The choice of a hardware engineer to lead Apple into the AI era may seem counterintuitive. But Apple's AI strategy is fundamentally a hardware distribution play: rather than building frontier AI models, Apple controls the roughly 2 billion active devices where those models run [9]. Ternus's deep familiarity with the silicon and hardware stack—particularly the Neural Engine architecture in Apple's chips—could prove more relevant than a software background if Apple's bet is that on-device AI processing is the decisive advantage.

Siri's Overhaul: A Rebuilt Assistant, 18 Months Late

The centerpiece announcement was Siri AI, a ground-up rebuild of Apple's assistant with a chatbot-style interface, multi-modal capabilities (text, audio, image, and video understanding), and on-screen awareness [10]. A new "Search Ask" feature integrates Siri into the Dynamic Island, and the assistant can now perform multi-step commands across apps [10]. The underlying architecture was co-developed with Google, running a custom Gemini model under a deal reportedly worth approximately $1 billion per year [9].

Critically, iOS 27 will also let users choose their default AI provider from among ChatGPT, Claude, Gemini, Grok, Copilot, or Perplexity [9]. This is a notable departure from Apple's walled-garden tradition and an acknowledgment that no single AI model dominates every use case.

What Siri Still Lacks

Siri AI will launch as a beta in September 2026, initially in English only [10]. Samsung's Galaxy AI, by contrast, has been shipping since early 2024 and runs on 400 million devices globally, with the company projecting 800 million Gemini-powered devices by end of 2026 [11]. Samsung's Galaxy S26, released in February 2026, already demonstrates Gemini's ability to take autonomous actions inside third-party apps—a capability Siri AI has announced but not yet shipped [11].

Google's own Gemini assistant on Android and Pixel devices has had roughly 18 months of real-world deployment, giving it a substantial lead in user feedback and model fine-tuning. The question is not whether Apple can eventually match these features, but whether the September 2026 beta will be polished enough to change the perception that Apple is playing catch-up.

The AI Spending Gap: Apple vs. the Field

The scale of Apple's relative underinvestment in AI infrastructure becomes stark when compared to its peers. Microsoft projects approximately $190 billion in capital expenditure for 2026, Google over $115 billion, and Meta roughly $110 billion [12]. Apple's total capex, estimated at around $25 billion, is a fraction of those figures [12].

Source: CNBC / Yahoo Finance

Data as of Jun 9, 2026CSV

Apple's counter-argument is strategic: it does not need to train frontier models because it can license them (Gemini from Google) while focusing its spending on the on-device silicon and privacy infrastructure where it has a durable advantage [9]. The company's new Apple Foundation Models include a 20-billion-parameter on-device model designed for the A19 Pro chip, requiring no cloud connection for many tasks [13].

But the steelman case that Apple is structurally late deserves serious consideration. Samsung already has hundreds of millions of devices running sophisticated AI features [11]. Google's Gemini is embedded in Search, Android, Workspace, and hardware products [9]. Microsoft is building its own MAI models and owns the orchestration layer through Azure and Copilot [9]. If developer ecosystem adoption and enterprise deployment lock in around these platforms before Apple's features exit beta, Apple's distribution advantage in consumer devices may not be enough to win in AI-assisted productivity, coding, and enterprise workflows.

macOS 27 "Golden Gate": The End of Intel Macs

macOS 27 will run exclusively on Apple Silicon Macs, formally ending support for every Intel-based Mac [14][15]. The last Intel Macs sold—the 2019 Mac Pro, 2020 27-inch iMac, and 2020 13-inch MacBook Pro—will be stuck on macOS Tahoe (macOS 26), which will continue to receive security patches for approximately three years, through 2028-2029 [14].

Apple has not disclosed the exact number of Intel Macs still in active use. The first M1 Macs shipped in November 2020, meaning every Mac purchased before that date runs Intel hardware. Industry estimates suggest millions of Intel Macs remain in circulation, particularly in enterprise, education, and creative professional environments where upgrade cycles are longer [15].

Historical Comparison: PowerPC vs. Intel Transitions

Apple's previous architecture transition—from PowerPC to Intel—began in 2005 and was completed across the product line by 2006, faster than the projected two years [16]. The original Rosetta translation layer was dropped with Mac OS X Lion in July 2011, roughly five and a half years after the first Intel Mac shipped [16].

The Intel-to-Apple Silicon transition has been more gradual. Six years after the first M1 Mac, Apple is now dropping Intel support in macOS 27, with Rosetta 2 remaining available through macOS 27 but scheduled for significant reduction in macOS 28 (expected fall 2027), which will cut off compatibility for more than 18,800 Intel-only apps [17]. This gives developers a longer runway—seven years versus roughly five and a half—reflecting the larger installed base of Intel Macs compared to the PowerPC era.

Comparison to Microsoft's Approach

Microsoft's Windows upgrade cutoffs have historically been more generous in pure calendar time. Windows 10, released in 2015, received its final security update in October 2025—a ten-year lifecycle [15]. However, Microsoft's approach to Windows 11 imposed strict hardware requirements (TPM 2.0, specific CPU generations) that also locked out millions of otherwise functional PCs. The key difference: Microsoft offered extended security updates for Windows 10 as a paid program, while Apple's three-year security patch commitment for Tahoe is free but time-limited.

Privacy as Strategy: Private Cloud Compute

Apple's privacy commitments remain its most differentiated feature. Private Cloud Compute (PCC) enforces stateless computation—user data is never stored on servers, used only for immediate inference, and the process is cryptographically verifiable [13][18]. Apple publishes measurements of all PCC code in a tamper-proof transparency log and makes software images available for independent security research [18].

The WWDC 2026 expansion includes running Apple Intelligence workloads on Google Cloud and NVIDIA infrastructure while maintaining Apple's exclusive control over PCC software [13]. Apple devices will only trust PCC software cryptographically signed by Apple, regardless of where the servers physically sit [13].

This contrasts with competitors' approaches. Google's Gemini processes queries on Google's servers under Google's privacy policy, which permits data use for model improvement unless users opt out. Microsoft's Copilot similarly routes through Azure with varying data retention policies depending on enterprise vs. consumer tiers. Samsung's Galaxy AI relies on Google's cloud infrastructure for complex tasks, inheriting Google's data practices.

The legal distinction matters. Apple's claim is architecturally enforced (cryptographic attestation, stateless compute) rather than policy-based (privacy policy promises). Whether this distinction is meaningful in practice depends on whether Apple's PCC implementation withstands independent audit—a process Apple has invited but which has not yet produced public results for the 2026 expansion to third-party data centers.

iOS 27: Developer and Enterprise Impact

iOS 27 introduces several API changes with material implications for developers. The new "Liquid Glass" UI framework requires apps to migrate from older UIVisualEffectView blur effects, and apps with custom glass implementations will need manual updates [19]. App Intents 2.0 expands the framework that connects apps to Siri, Spotlight, and Shortcuts with richer entity types [19]. The SDK also ships with foldable-device-aware APIs ahead of Apple's expected foldable hardware launch later in 2026 [19].

A new Foundation Models framework gives developers on-device access to Apple's language models without API tokens [19]. iOS 27 will be compatible with iPhone 11 and later [20].

No major third-party developers have publicly signaled compatibility crises with iOS 27's API changes, though the Liquid Glass migration represents real engineering work for apps with heavily customized interfaces. Enterprise customers running Intel Mac fleets face a more concrete deadline: macOS 27 will not run on their hardware, meaning they must either upgrade to Apple Silicon or maintain parallel macOS Tahoe environments with a finite security-patch window.

What Analysts Expect Next

The leadership transition introduces uncertainty that Wall Street is watching closely. Cook's tenure delivered a roughly 20x stock return, driven by Services revenue growth and disciplined capital allocation including more than $600 billion in share buybacks [4][5]. Ternus inherits a company that is the world's most valuable by market cap but faces its first significant platform-strategy risk in a decade.

The bull case: Apple's 2 billion active devices represent the largest AI distribution network in the world, and the company's privacy-first approach will attract users and enterprises increasingly wary of data-hungry competitors. Ternus's hardware expertise positions him to execute on the Apple Silicon roadmap and whatever comes after the Vision Pro.

The bear case: AI is a software and data game where scale of training infrastructure matters more than device distribution. Apple's ~$25 billion capex versus Microsoft's $190 billion is not a sign of strategic restraint—it is a structural disadvantage. If Siri AI launches as a mediocre beta while Samsung and Google are on their third or fourth generation of shipped AI features, the perception gap could become a reality gap that device sales alone cannot close.

The truth likely sits somewhere between these poles. Apple has been counted out before—most memorably when the company was 90 days from bankruptcy in 1997—and its ability to execute when it commits resources remains formidable. But the AI race is moving faster than any technology cycle Apple has previously navigated, and the September 2026 launch of Siri AI will be the most consequential product beta the company has shipped since the original iPhone.