Apple may publicly frame its artificial intelligence strategy around careful partnerships and privacy-first principles, but behind the scenes the company is far more dependent on one external player than previously understood. According to Bloomberg journalist Mark Gurman, Apple’s internal product development now leans heavily on Anthropic and its Claude models.
Speaking in a recent interview on TBPN, Gurman offered a blunt assessment of Apple’s current AI setup. “Apple runs on Anthropic at this point,” he said, adding that Anthropic is powering much of what Apple is doing internally across product development and internal tooling. Gurman also claimed that Apple is running custom versions of Claude on its own servers, rather than relying on off-the-shelf access to public AI models.
Those remarks help explain Apple’s earlier interest in striking a major AI partnership with Anthropic before its recently announced agreement with Google. While Apple ultimately went public with a Google partnership, reporting over the past year has suggested that Anthropic and OpenAI were both seriously considered as potential long-term collaborators.
Gurman’s comments indicate that Anthropic was not just a theoretical option. Apple reportedly wanted to formalise a deeper relationship with the company, but negotiations broke down over cost. According to Gurman, Anthropic was seeking several billion dollars per year for the partnership, with contractual terms that would have allowed those fees to double over time. For a company as financially conservative as Apple, even in the AI era, that price tag appears to have been a deal-breaker.
By comparison, Apple’s deal with Google is said to cost around one billion dollars annually. That lower figure, combined with Google’s scale and existing infrastructure, likely made the partnership easier to justify internally, even if Apple was already relying heavily on Anthropic technology in practice.
The timing of these discussions also matters. Apple’s AI partner search unfolded against the backdrop of regulatory uncertainty around its long-running Safari search agreement with Google. With that deal facing scrutiny, Apple reportedly prioritised alternative AI partners earlier in the process, including Anthropic and OpenAI, before circling back to Google once the situation became clearer.
What stands out most from Gurman’s remarks is just how embedded Anthropic appears to be inside Apple’s internal workflows. Running customised Claude models on Apple-controlled servers suggests a level of integration that goes well beyond experimentation. It points to Anthropic models being used as foundational tools by Apple engineers and product teams, rather than as optional add-ons.
That reliance also adds context to Apple’s broader AI narrative. Publicly, the company has positioned itself as moving carefully, selectively, and with strong emphasis on user privacy. Privately, it appears to be leaning hard on external large language models to accelerate development, even as it works to maintain control by hosting and customising those systems internally.
The situation creates an interesting dynamic. On one hand, Apple is now partnered with Google for consumer-facing AI features. On the other, a significant amount of its internal AI work may still be powered by Anthropic. That dual-track approach reflects both Apple’s desire to avoid overdependence on any single partner and the reality that no single AI provider currently meets all of its needs.
If Gurman’s assessment is accurate, Anthropic’s influence on Apple’s products could be far greater than the public partnership announcements suggest. As Apple prepares to roll out more AI-driven features across iOS, macOS, and its wider ecosystem, the quiet role played by Anthropic may become increasingly difficult to ignore.
Discover the latest Business News, Sensex, and Nifty updates. Obtain Personal Finance insights, tax queries, and expert opinions on Moneycontrol or download the Moneycontrol App to stay updated!
