Apple has fundamentally altered its artificial intelligence strategy, moving away from intensive on-device AI development towards a platform model that positions the company as a marketplace operator rather than primary AI developer, according to Bloomberg.
The shift, which includes rare employee bonuses to acknowledge the strategic pivot, represents Apple’s acknowledgement that competing directly with OpenAI, Google, and Anthropic in foundational model development requires unsustainable capital expenditure. Instead, the company will leverage its existing App Store infrastructure to distribute third-party AI services whilst maintaining control over the customer relationship and revenue share.
This approach mirrors Apple’s historical playbook: create the distribution channel, set quality standards, and collect platform fees whilst others bear development costs. The company successfully employed this model when it transformed mobile software distribution in 2008, creating a £1.1 trillion app economy from which it extracts an estimated 15-30% commission.
For enterprise AI vendors, the implications are substantial. Access to Apple’s installed base of over 2 billion active devices provides immediate distribution scale that would take years to build independently. However, this access comes at a cost—not merely financial through revenue sharing, but strategic through Apple’s ability to set technical requirements, pricing guidelines, and user experience standards.
The pivot also signals Apple’s assessment that on-device AI processing, whilst technically impressive, offers insufficient competitive advantage to justify the investment required. Training large language models now costs hundreds of millions of dollars per iteration, with leading labs spending upwards of £500 million annually on compute infrastructure alone. By contrast, operating a marketplace requires primarily software engineering and content moderation—capabilities Apple already possesses.
Companies positioned to benefit include established AI vendors like OpenAI and Anthropic, which gain streamlined access to iOS and macOS users without building separate distribution channels. Smaller enterprise AI firms may find the platform essential for customer acquisition, though they’ll face Apple’s notorious approval processes and the company’s history of incorporating popular third-party features into native offerings.
The losers are likely Apple’s own AI engineering teams, who face reduced scope and resources, and potentially Google, whose £15 billion annual payment for default search placement becomes more vulnerable as Apple diversifies its services revenue through AI marketplace fees. Enterprise customers may also face fragmented AI tool access, paying separately for services that competitors like Microsoft bundle into existing subscriptions.
The rare employee bonuses mentioned by Bloomberg suggest internal recognition that this strategic shift required difficult trade-offs. Apple typically reserves such payments for extraordinary circumstances, indicating leadership views the pivot as both necessary and potentially contentious among staff who joined to build cutting-edge AI systems.
The marketplace model also provides Apple plausible deniability regarding AI safety and accuracy concerns. By positioning itself as a platform rather than AI provider, the company can deflect responsibility for model outputs to third-party developers whilst maintaining App Store-style content policies that protect brand reputation.
Market observers should monitor several indicators of this strategy’s success: the number and quality of AI services Apple approves for its platform, revenue share terms compared to standard App Store rates, and whether the company maintains any internal AI development for proprietary features. Additionally, regulatory responses matter—antitrust authorities already scrutinising Apple’s App Store practices will likely examine whether the company leverages its platform position to disadvantage AI competitors.
The broader significance extends beyond Apple. If the platform model proves commercially successful, other hardware manufacturers may follow, fragmenting enterprise AI distribution across multiple marketplaces with varying technical requirements and business terms. This would represent a marked departure from the current model where AI vendors distribute directly to customers through web interfaces and APIs, potentially adding friction and costs to enterprise AI adoption whilst consolidating power among platform operators.
