Apple Opens Siri to Rival AI Models, Monetizing What It Can’t Build
iOS 27's third-party LLM integration ends OpenAI exclusivity and transforms Siri into an AI marketplace—validating the commoditization thesis Apple spent $34 billion trying to avoid.
Apple will allow Siri to route queries to Google Gemini, Anthropic Claude, and OpenAI’s ChatGPT starting with iOS 27, ending OpenAI’s exclusive partnership and marking the company’s retreat from closed-garden AI. The announcement, expected at WWDC on June 8, 2026, transforms Siri from a proprietary assistant into a platform marketplace for large language models—a strategic pivot that acknowledges Apple’s own generative AI capabilities lack the differentiation to sustain platform lock-in.
Apple spent $34.55 billion on R&D in fiscal 2025—a 10% year-over-year increase directed largely toward on-device AI and Private Cloud Compute infrastructure. Despite this investment, the company is now opening Siri to the same third-party models it sought to avoid partnering with at premium prices just months ago.
The Economics of Strategic Retreat
Apple’s decision follows a failed negotiation with Anthropic in January 2026, where the AI startup demanded several billion dollars annually for Claude to power Siri—a figure that would double over three years, according to AppleInsider. Apple rejected the proposal and instead signed a $1 billion annual deal with Google for a custom Gemini model with 1.2 trillion parameters, per reporting by MLQ.ai citing the Financial Times.
The third-party integration solves Apple’s AI economics problem while creating a new revenue stream. The company already captured nearly $900 million in App Store fees from generative AI apps in 2025, charging a 30% commission on first-year subscriptions (15% thereafter), according to MacDailyNews. AI app revenue is projected to exceed $1 billion in 2026. By routing Siri queries through an Extensions system that treats LLMs like App Store products, Apple can monetize AI distribution without bearing the multi-billion-dollar costs of differentiated model development.
Validation for the Commodity Thesis
The move validates the distribution-first strategies of OpenAI and Anthropic, both of which bet that AI models would commoditize faster than platforms could build proprietary alternatives. Anthropic reached $19 billion in annualized revenue by early 2026—up from $4 billion mid-2025—by focusing on enterprise B2B partnerships rather than consumer apps, Sacra reported. OpenAI hit an estimated $25 billion in annualized revenue by February 2026 with 910 million weekly active users and 9 million paying business users.
Apple’s willingness to open Siri confirms what Anthropic’s pricing demands made explicit: LLM providers now command enough market power to extract platform-level economics. Anthropic captures 40% of enterprise LLM spending, according to Axios, while Apple planned just $14 billion in capex for 2026—less than 3% of the $650 billion combined AI infrastructure spending by Amazon, Microsoft, Meta, and Google, per Yahoo Finance.
“Extensions allow agents from installed apps to work with Siri, the Siri app and other features on your devices.”
— Mark Gurman, Bloomberg
The Extensions architecture—described by Bloomberg’s Mark Gurman and reported by MacRumors—allows users to select their preferred LLM through Siri settings, routing queries to Google Gemini, Anthropic Claude, or OpenAI ChatGPT. This mirrors the App Store model where Apple controls distribution and monetization while outsourcing differentiation to third parties.
Regulatory Cover and Competitive Pressure
The timing provides regulatory insulation. Elon Musk’s xAI sued Apple and OpenAI in August 2025, accusing the companies of conspiring to ensure continued dominance in the AI market through the exclusive ChatGPT integration. The lawsuit was proceeding as of November 2025, MacRumors reported. Opening Siri to multiple LLM providers defuses antitrust concerns while still capturing App Store-level economics on AI subscriptions.
The shift pressures Microsoft and Google to follow suit. If iOS users can switch between LLMs without changing platforms, Android and Windows face similar demands for model neutrality. This accelerates industry-wide AI commoditization—the exact outcome hyperscalers spending hundreds of billions on proprietary infrastructure sought to avoid.
Platform Control Erosion
Apple’s AI strategy now mirrors its historical response to technological transitions it couldn’t control: own the experience, outsource the commodity, capture the margin. The company followed this pattern with web browsers (Safari wrapping WebKit), cloud services (iCloud running on Google and AWS infrastructure), and now AI models.
But the pattern reveals eroding platform leverage. App Store economics worked when Apple controlled scarce distribution and developers needed access to iOS users. AI model providers—particularly Anthropic and OpenAI—command comparable scarcity through model quality and enterprise relationships. When Anthropic can demand several billion annually and walk away, Apple’s negotiating position weakens.
The asset-light approach may prove strategically sound if AI models commoditize as rapidly as Asymco analyst Horace Dediu argues. While competitors spend $650 billion building proprietary AI infrastructure, Apple can capture distribution economics without the capital risk. But that requires believing LLM differentiation collapses faster than Apple’s 30% commission rate becomes untenable—a bet that looks less certain as Anthropic’s revenue quintuples in nine months.
What to Watch
The WWDC announcement on June 8 will clarify revenue-sharing terms and whether Apple negotiated platform fees from LLM providers beyond standard App Store commissions. If Anthropic, Google, and OpenAI agreed to additional payments for Siri integration access, Apple successfully monetized AI distribution at scale. If they simply route through existing App Store subscriptions, the company’s leverage proved weaker than expected.
Monitor enterprise adoption rates for third-party LLMs via Siri Extensions versus direct app usage. If users prefer app-based AI access, the Extensions architecture fails to capture switching behavior and Apple’s App Store moat around AI remains intact. If Extensions become the primary interface, Apple successfully platformized AI distribution—but at the cost of admitting its own models can’t compete.
Competitive responses from Google and Microsoft will signal whether iOS 27’s openness becomes an industry standard or a uniquely Apple concession. If Android adopts similar LLM marketplace features within six months, the commoditization thesis accelerates. If Google resists, betting Gemini integration provides sustainable differentiation, the AI platform wars remain contested—and Apple’s retreat looks premature.