After years of internal delays, strategic hesitation, and quiet leadership reshuffles, Apple has made a decision that marks a clear break from its historical playbook.
It has stopped trying to own the brain.
Apple has officially abandoned the effort to build a fully independent, end-to-end foundation model for Siri and Apple Intelligence. Instead, it has partnered with its oldest rival, Google, making Gemini the default intelligence layer across iOS.
The move is not cosmetic. It is structural. And it reshapes the balance of power in consumer technology.
The Deal, Clearly Explained
This is not an app integration or a temporary bridge.
Under the agreement, the next generation of Apple Foundation Models will be built directly on top of Google’s Gemini, running on Google’s cloud infrastructure. Siri, system search, writing tools, and cross-app actions will all rely on Gemini as the underlying reasoning engine.
Apple keeps the interface, the hardware, and the privacy messaging. Google supplies the intelligence.
Financial terms have not been officially disclosed, but Bloomberg estimates Apple is paying approximately $1 billion per year for Gemini access. That figure sits alongside the $20 billion annually Google already pays Apple to remain the default search engine on Safari.
Crucially, this AI integration does more than add features, it legally reinforces Google’s search dominance. By embedding Gemini directly into Siri and system-level queries, Google has effectively armored that $20 billion payment against future DOJ challenges. Search is no longer just a browser setting; it is an AI dependency.
The first public result, a fully redesigned “Gemini-powered Siri,” is expected with iOS 26.4 in March or April 2026, bringing long-promised features like on-screen awareness and autonomous cross-app actions.
Why Apple Blinked
Apple did not make this move from a position of strength.
Its internal AI program, overseen until recently by John Giannandrea, failed to close the reasoning and latency gap with Google’s Gemini 3. Continuing alone would have pushed a truly competitive Siri into 2027 or later, an eternity in the current AI cycle.
By partnering with Google, Apple achieved three things immediately:
- Compressed time-to-market by at least a year
- Capped AI R&D risk, avoiding further multi-billion-dollar internal spending
- Preserved ecosystem control while outsourcing intelligence
For a company built on vertical integration, this is a rare concession and a revealing one.
The Quiet Casualty: OpenAI
The biggest loser is not Apple’s pride. It is OpenAI’s position.
In 2024, ChatGPT was introduced as a headline Siri feature. In 2026, it has been demoted to an opt-in, pro-tier plugin.
Gemini now handles the default, invisible tasks: system search, summarization, navigation, and daily actions. ChatGPT is reserved for edge cases creative writing, complex reasoning, or niche queries only when users explicitly choose it.
That distinction is fatal at scale.
In the modern software economy, defaults are infrastructure. Plugins are cost centers. If the default works, users never opt in. OpenAI has effectively lost habitual access to nearly 2 billion devices, becoming a specialist tool rather than a daily companion.
Infrastructure Wins the AI War
This outcome reflects a deeper financial reality.
AI leadership is no longer decided by model quality alone. It is decided by cost of computation at global scale.
Google’s advantage is vertical:
- Gemini models
- Proprietary data centers
- Custom silicon
In November 2025, Alphabet launched its TPU v7 “Ironwood” chips. According to institutional estimates, Ironwood allows Google to run large-scale AI queries at a total cost of ownership roughly 40% lower than competitors relying on Nvidia’s Blackwell GPUs.
This cost advantage explains the economics of the Apple deal. Google can afford to pay Apple billions, absorb Siri’s massive query volume, and still expand margins, something no AI company paying the “Nvidia tax” can match.
Wall Street’s Judgment
Markets understood this immediately.
On January 12, 2026, Alphabet closed with a market capitalization of $4.03 trillion, overtaking Apple for the first time since 2019. But the more telling signal is not the absolute number, it is the multiple.
Even after the surge:
- Alphabet trades at ~24× forward earnings
- Apple trades near 29×, despite slower growth and rising dependency
This is classic re-rating behavior. Investors are no longer valuing Google as a mature search company, but as an AI infrastructure utility. Apple, by contrast, is being repriced as premium hardware with rented intelligence.
Why Alphabet Now Sits at the Center
Several forces reinforce Google’s dominance:
- Distribution lock-in: Gemini now powers both Android and iOS intelligence
- Margin control: TPUs eliminate reliance on Nvidia
- Cloud pull-through: Google Cloud revenue rose 34% last quarter as enterprises move closer to Gemini
This is no longer an “AI feature race.” It is a battle for who owns the default interface to computation.
What Google Is Building Next
Alphabet’s ambitions extend beyond assistants.
In 2026, Google plans to:
- Launch Aluminium OS, merging Android and desktop computing into a unified AI-first platform
- Roll out Gemini 4.0, focused on autonomous, agentic task execution
- Expand AI hardware, including smart glasses and third-party TPU deployments
- Turn Google TV into an interactive learning platform with narrated, AI-generated deep dives
The goal is clear: replace apps, menus, and workflows with intent-driven computation.
The Final Reality
Apple remains one of the most profitable companies in the world. But in AI, profitability now follows infrastructure, not craftsmanship.
By choosing Gemini, Apple acknowledged a new hierarchy.
And there is one final, uncomfortable detail that underscores how complete Google’s victory has become: Gemini now powers the intelligence layer of both the Galaxy S26 and the iPhone 17/18.
For the first time in technology history, a single company controls the “brain” of every premium smartphone on the planet, regardless of whose logo is on the hardware.
That is not a partnership.
That is platform ownership.
0 Comments