Apple steadfastly upholds a privacy-centric approach in its AI implementation, prioritizing local, on-device AI processing. Through the refinement of Google's Gemini model, Apple has successfully trained a streamlined, on-device AI variant. At one point, Apple even contemplated acquiring Liquid AI, a burgeoning startup dedicated to local AI operations. When it comes to cloud deployment, the full Gemini model's vast parameters and formidable computational demands pose a challenge for Apple's Private Cloud Compute infrastructure, making it difficult to support. Consequently, in the upcoming iOS 27 iteration of Siri, certain requests will be rerouted to Google Cloud. There, an authorized iteration of the Gemini model will step in to process these inquiries. To safeguard user privacy, Apple has given the green light for the integration of NVIDIA's 'confidential computing' technology. While this may introduce a slight lag in the speed of cloud-based AI queries, it substantially bolsters the protection of sensitive data. In the end, Apple may come to depend on both Google Cloud and NVIDIA's AI chips.