It looks like Apple will be leaning on Google for help when it comes to Siri. Bloomberg News said on Wednesday that Apple plans to use a 1.2 trillion-parameter artificial intelligence model developed by Alphabet’s Google to help power a revamp of its Siri voice assistant.
Apple’s Siri is a voice-activated virtual assistant developed by Apple Inc., launched in 2011 with the iPhone 4S. It uses natural language processing and machine learning to interpret spoken commands, helping users send messages, make calls, set reminders, play music, check weather, control smart-home devices via HomeKit, and answer questions. Siri works across Apple devices—iPhone, iPad, Mac, Apple Watch, Apple TV, HomePod, and Apple Vision Pro. While it supports multiple languages and regional dialects, feature availability varies, and some advanced capabilities are limited to certain regions. Apple emphasizes privacy, processing many commands on-device, though some requests still rely on the cloud. Siri adapts to usage patterns, but the highly context-aware AI improvements originally expected in 2025 have been delayed to 2026. Despite these limitations, Siri remains a key tool for hands-free tasks and smart-home integration.
The report said that after an evaluation, the companies are finalizing a deal that would have Apple pay about $1 billion a year for access to Google’s technology.
READ: Google adds Gemini AI to Chrome following antitrust court win (
Bloomberg said the iPhone maker will use Google’s Gemini model as a stopgap until its own systems are ready.
Apple plans to run the Gemini model on its own infrastructure (“Private Cloud Compute”) to maintain user privacy, though the partnership is not yet live. The feature rollout will vary by region and device, and Google’s contribution is intended as a temporary solution while Apple develops its own AI models.
Google’s Gemini AI, launched in 2025 by Google DeepMind/Google LLC, is a family of advanced multimodal models capable of processing text, images, audio, video, and code, with long-context inputs up to one to two million tokens in higher-end versions. The 2.5 series, including “Gemini 2.5 Pro,” is designed for complex reasoning, coding, and real-time assistance across Google services and the Gemini app, with features like object recognition and screen sharing. While Gemini represents cutting-edge AI, it remains prone to typical model limitations such as hallucinations, bias, and uneven reliability, and not all features are available uniformly across regions, devices, or subscription tiers. Privacy and data management are nuanced, as memory and personalization settings may retain interactions unless adjusted by users. Full adoption and consistent performance remain works in progress, with ongoing updates expected throughout 2025.
READ: Google rolls out Gemini 2.5 Pro, their ‘most intelligent AI model’ yet (
The report said that the deal is separate from earlier discussions about integrating Gemini directly into Siri as a chatbot.
The Apple–Google collaboration over Siri in 2025 reflects a strategic yet cautious approach to AI development. By leveraging Google’s Gemini AI, Apple can enhance Siri’s reasoning, context awareness, and multimodal capabilities while maintaining privacy through on-device processing and its Private Cloud Compute infrastructure. This partnership is temporary, bridging the gap until Apple’s own advanced AI models are ready. Feature rollout will be uneven across regions and devices, with some advanced functionalities remaining limited.
For Google, the deal provides significant annual revenue and expanded real-world deployment of Gemini across Apple’s ecosystem. Both companies face challenges, including technical integration, model limitations such as hallucinations, bias, and inconsistent reliability, and regulatory and regional constraints.
Overall, the partnership demonstrates a pragmatic, competitive, and cautious approach to AI innovation in 2025, balancing capability, privacy, and real-world deployment realities while positioning both firms strategically in the evolving AI landscape.