From Siri To Campos – Why Did Apple Choose Economics Over Sovereignty?
Apple is replacing Siri with Campos, a new AI chatbot powered by Google’s Gemini 3 models instead of building its own infrastructure.
The move trades sovereignty for economics: Google’s cost per token is several times lower than competitors, making it the only sustainable choice at Siri’s scale. Apple owns the interface. Google owns the intelligence layer underneath.
Video – Why Apple Surrenders to Google?
Why Apple Chose Google for Campos
Apple is replacing Siri with Campos, a ChatGPT-style chatbot running on Google’s Gemini 3 models.
Key reasons for choosing Google:
- Gemini 3 offers several times lower cost per token than OpenAI or Anthropic
- Apple spent $12.7 billion on infrastructure in 2025 vs. Google’s $90 billion
- Apple pays Google $1 billion annually, potentially $5 billion total
- Alternative models would be unsustainable at Siri’s scale (86.5-87 million U.S. users)
Not OpenAI. Not Anthropic. Not even their own infrastructure at scale.
This is not a product decision. This is an infrastructure capitulation.
What the Infrastructure Gap Reveals
Apple spent $12.7 billion on property, plant, and equipment in fiscal 2025. Google will spend $90 billion this year.
That gap is not a rounding error. That gap is a strategic choice.
Apple is paying Google approximately $1 billion annually for Gemini access, potentially reaching $5 billion over the deal’s duration. Apple decided against matching the infrastructure investments of its competitors because the economics were unworkable.
Gemini 3 is by far the most scalable top-tier model with several times lower cost per token than the competition. Any other model becomes extremely unsustainable for powering Siri at scale.
Economics defeated sovereignty.
The Core Trade-Off: Apple chose rental over ownership because building equivalent infrastructure would require 7x current spending with no guarantee of matching Google’s cost efficiency.
Why Distribution Alone Does Not Win in AI
Siri holds 45.1% of the smartphone voice assistant market. That translates to 86.5 to 87 million U.S. users. Apple Intelligence reached 82% adoption on iOS 18.
But only two features were recognized as having meaningful impact by more than 20% of users. Clean Up and photo search.
Distribution does not equal adoption in AI features. Apple has the reach. Apple does not have the engagement.
Craig Federighi previously stated Apple did not want to build a chatbot. Growing competitive pressure forced a strategic reversal. The popularity of ChatGPT, Gemini, and Claude made the chatbot format unavoidable.
Market adoption velocity defeated Apple’s preferred technical approach.
What This Means: Having devices in millions of hands means nothing if users do not activate the features. Apple needed a chatbot because the market decided chatbots are how people interact with AI.
How OpenAI Walked Away
OpenAI consciously declined to become Apple’s custom model provider in autumn 2025. OpenAI chose instead to focus on building its own AI hardware with former Apple design chief Jony Ive.
That move soured Apple on a deeper OpenAI partnership and forced the Google deal.
Apple tested an internal chatbot called Enchanté with employees in November 2025. It ran Apple Foundation Models, Anthropic’s Claude, and Google’s Gemini simultaneously. After technical assessment, Apple determined Google’s technology provides the most workable foundation.
Capability was not the only variable. Cost per token at Siri’s scale made the decision inevitable.
The Reality: OpenAI betting on hardware with Jony Ive closed the door on partnership. Apple was left choosing between Anthropic and Google, and Google’s cost structure won.
What Campos Actually Is
Campos will run on Apple Foundation Models version 11 at 1.2 trillion parameters, powered by a custom Google Gemini 3 model. Processing shifts to Google’s TPU servers rather than Apple’s Private Cloud Compute.
That represents a policy reversal on data sovereignty. Apple’s privacy positioning now contradicts the reality of where user queries are processed.
The partnership effectively makes billions of smartphones Gemini phones. Google’s models become a default layer across the mobile ecosystem. This creates unprecedented AI market concentration.
Google’s market cap briefly surpassed $4 trillion following the announcement, overtaking Apple for the first time since 2019.
Technical Specifications:
- 1.2 trillion parameter model
- Custom Google Gemini 3 base
- Processing on Google TPU infrastructure
- Launch after WWDC in June 2026
- Integrated across iOS 27, iPadOS 27, macOS 27
Translation: Apple owns the design. Google owns the engine, the fuel efficiency, and the ongoing operational costs.
The Structural Pattern Worth Watching
Apple is not even in the top five AI companies. Analysts note Apple is doing a poor job keeping up, let alone leading, despite being the most valuable company since 2011.
This is what happens when infrastructure shifts rewrite competitive dynamics faster than product innovation. Energy efficiency is the next computing moat. Adoption velocity defeats technical superiority in markets with network effects.
ChatGPT has 80% market share but loses on specialized tasks. That is category disaggregation, not competitive weakness. Market dominance and product superiority are increasingly orthogonal.
Apple chose distribution over infrastructure. Apple chose economics over sovereignty. Apple chose partnership over independence.
Pattern Recognition: When compute infrastructure becomes the bottleneck, companies with distribution but without infrastructure become renters. The value shifts to whoever owns the processing layer.
What Campos Delivers to Users
Campos launches after WWDC in June, following the new iPhone release. Integration across iOS 27, iPadOS 27, and macOS 27. Chat-style voice and text conversations. Web search, summarization, content generation, file analysis, on-screen context understanding, device control.
Accessible via the existing Siri wake word or side button. Integrated into Mail, Music, Photos, Xcode.
This is Apple’s most significant AI initiative to date. The goal is improving upon the weak reception of Apple Intelligence and competing with ChatGPT, Gemini, and Claude.
But Campos is built on rented infrastructure. The compute advantage belongs to Google. The data flows through Google’s systems. The cost structure is determined by Google’s pricing.
Apple owns the interface. Google owns the intelligence layer.
User Experience: You will interact with Apple’s design wrapped around Google’s intelligence. The experience feels like Apple. The processing happens on Google infrastructure.
What This Means for Your Next Five Years
You are watching the rebundling of who owns compute advantage. The companies controlling AI infrastructure at scale will extract value from everyone else who needs access.
Apple made a calculated bet. Apple will win on user experience and ecosystem lock-in. Apple will lose on infrastructure independence and long-term cost control.
The question is not whether Campos will be better than the current Siri. Campos will be better. The question is what happens when the company controlling your AI layer also controls the pricing, the data policies, and the strategic roadmap.
This is not about a chatbot.
This is about who owns the intelligence infrastructure powering the next decade of computing.
Apple chose to rent.
Frequently Asked Questions
When does Campos launch?
Campos will launch after WWDC in June 2026, following the new iPhone release. Integration across iOS 27, iPadOS 27, and macOS 27.
Why did Apple choose Google over OpenAI?
OpenAI declined to become a custom model provider in autumn 2025, choosing instead to build AI hardware with Jony Ive. Google’s Gemini 3 also offers several times lower cost per token than alternatives, making it the only sustainable choice at Siri’s scale.
Will Campos replace Siri completely?
Campos is the replacement for Siri as Apple’s primary AI assistant. Users will access Campos using the same Siri wake word or side button, but the underlying technology shifts from Apple’s systems to Google’s Gemini 3 infrastructure.
Does this change Apple’s privacy stance?
Processing on Google’s TPU servers rather than Apple’s Private Cloud Compute represents a policy reversal on data sovereignty. User queries will flow through Google’s infrastructure, contradicting Apple’s historical privacy positioning.
How much is Apple paying Google?
Apple pays Google approximately $1 billion annually for Gemini access, with the deal potentially reaching $5 billion over its full duration.
What features will Campos have?
Campos supports chat-style voice and text conversations, web search, summarization, content generation, file analysis, on-screen context understanding, and device control. Integration includes Mail, Music, Photos, and Xcode.
Is Apple still developing its own AI models?
Apple Foundation Models version 11 will run at 1.2 trillion parameters, but powered by a custom Google Gemini 3 base. Apple is developing interface and integration layers while relying on Google for core model infrastructure.
What does this mean for Google’s market position?
The partnership makes billions of smartphones effectively Gemini phones, with Google’s models becoming a default layer across the mobile ecosystem. Google’s market cap briefly surpassed $4 trillion following the announcement, overtaking Apple for the first time since 2019.
Key Takeaways
- Apple is replacing Siri with Campos, powered by Google’s Gemini 3 instead of proprietary infrastructure, because Google’s cost per token is several times lower than alternatives.
- The infrastructure gap between Apple ($12.7B spent) and Google ($90B planned) forced a strategic choice: rent rather than build equivalent compute capacity.
- Distribution without engagement is worthless in AI. Despite 82% iOS 18 adoption, only two Apple Intelligence features gained meaningful traction.
- OpenAI’s decision to pursue hardware with Jony Ive closed the door on deeper partnership, leaving Google as the only economically viable option.
- Processing shifts from Apple’s Private Cloud Compute to Google’s TPU servers, reversing Apple’s data sovereignty stance and contradicting privacy positioning.
- The partnership creates unprecedented AI market concentration, effectively converting billions of iPhones into Gemini-powered devices.
- Apple owns the interface and user experience. Google owns the intelligence layer, the cost structure, and the long-term strategic leverage.
