In November 2025 Bloomberg’s Mark Gurman reported that Apple is “planning to pay about $1 billion a year” for a custom 1.2 trillion-parameter AI model from Google’s Gemini to power a Siri overhaul. Reuters later confirmed this scoop: citing the same sources, it noted Apple would use Google’s 1.2T-parameter Gemini AI model as a stopgap while its own AI systems catch upreuters.com. Importantly, neither Apple nor Google have confirmed the deal – Reuters observed “Google had no comment, while Apple did not respond” – so as of now this remains a well-sourced rumor, not an official partnership.
The 1 billion per year figure can be directly traced to Bloomberg unnamed sources. These details are quoted by tech publications such as The Verge, AppleInsider and Toms Guide all referencing Bloomberg. Similar to most of Gurman scoops, the report is taken seriously but grounded in people who requested that they should not be identified. Briefly, the story has been reported by major media with confidence, yet the companies concerned remained silent. (Normally, Apple pays Google huge amounts of money to get services such as search, but in this case, it would be the reverse, i.e. Apple is paying Google, which is notable had it been the case)
Gemini AI vs. Other Options
As reported, the version of the Gemini LLM that would be licensed by Apple is a 1.2-trillion-parameter custom one. Comparatively, the current cloud-based model of Apple Intelligence has approximately 150 billion parameters, and therefore the Gemini model will be enormous. The architecture of Gemini is based on a mixture-of-experts design, and thus only a small proportion of its total parameters is fired when making a single query – allowing huge capacity at a cost that is not prohibitive.
Why Gemini AI, rather than, e.g., GPT of OpenAI or Claude of Anthropic? It was reported that Apple considered those as well. Sources indicate that Apple conducted extensive testing of the 1.2T -parameter Gemini model of Google but also tested ChatGPT and Claude of Anthropic. As it has been reported early, Apple even considered OpenAI or Anthropic models before choosing Google.
. The given reason is rather pragmatic: Apple found the terms and performance of Google to be the most optimal. Indicatively, insiders have reported that Apple considered the prices of Anthropic to be excessively high therefore, it caused the decision to shift to Google Gemini. In brief, Apple decided that Gemini was its best option since it tested the alternatives.
Siri’s AI Upgrades: Summaries, Reasoning and More
The new Siri – code-named Linwood (project Glenwood) – is expected to debut in Spring 2026 (likely in an iOS 26.4 release).
Bloomberg claims that this redesigned Siri will rely on Gemini to perform the most complicated tasks. Practically, Google AI would assume the summarization and planning roles: the AI would assume the summarizing and planning capabilities of Siri, which emphasizes the synthesis of information and the performance of the complex tasks. That is, Siri would become much more competent in multi-step reasoning and queries that are heavy in context.
According to Gurman, the new Siri will be capable of answering more intricate questions, and performing more complex tasks both within and between apps, thus it will be more like Claude or ChatGPT (Apple has no plans for a dedicated chatbot app).
As an illustration, a webpage summary with Siri on the hood could be more accurate than in the past or a multi-task (such as trip planning with multiple steps) could be completed more effectively. Such tasks regularly confuse Siri, which is today, but Apple thinks that the 1.2T model developed by Google will significantly increase the power of the system and the capabilities it has to read in contexts and analyze complex data.
The Siri upgrade is already delayed – Apple acknowledged in 2021 that Siri upgrades would be moved to 2026, so a Spring 2026 (along with iOS 26.4) release is within the timeframe. The sources of Bloomberg comment internally that the Siri project is known as Glenwood, and the new assistant is named Linwood.
Apple Private Cloud deployment.
More importantly, the Gemini would operate on Apple cloud servers and not Google cloud servers. The AI is said to be running on the new Apple infrastructure, the Private Cloud Compute. It implies that you request Siri to perform a particular query or task, the query and data remains in the secured cloud environment at Apple despite the heavy work load being handled by the brain in Gemini.
Apple specifically intends to retain the information of users in the hands of Apple and nowhere in Google infrastructure. Google is, in reality, merely a supplier in the background. The iPhone users would not have Gemini in the iPhone interface, but they would only be able to access it through Siri in the cloud. Such an arrangement ensures the privacy policies of Apple: your information is isolated on Apple hardware with the advantage of the Google AI.
The AI Roadmap and 1T-Param Model at Apple.
According to industry analysts, the Gemini AI deal is a transitory one in case it materialises. Apple regards it as a temporary measure until its models are powerful enough. As a matter of fact, Apple is already hastening to construct its own giant AI model. Several reports indicate that Apple also has its 1-trillion-parameter cloud model being developed and is expected to be launched in 2026.
In this consideration, Gemini would be used to replace the vacuum during the training of Apple internal LLM. According to Gurman, Apple is planning to continue to build new AI technology and in the future, it wants to substitute Gemini with its own.
This is in line with the overall strategy of Apple. The current cloud AI (the Siri / Apple intelligence system) at Apple has parameters of about 150B, which is much lower than the new deal. However, Apple has confirmed on device AI capabilities and the company is heavily investing in AI research and development.
Apple CEO Tim Cook has mentioned that the new Siri would be launched in the next spring, and the company has silently recruited talent and dedicated server hardware to AI.
Bloomberg reports Apple’s Vision Pro architect Mike Rockwell now leads Siri’s overhaul, under Craig Federighi. All signs point to Apple treating Google as a behind-the-scenes partner for now, while quietly building up its own trillion-parameter brains.
In sum, the $1B/yr Google Gemini AI deal is very much a rumor at this point – but it comes from highly credible outlets referencing inside sources. It’s not confirmed, yet multiple reports paint a consistent picture: Apple would use a custom 1.2T Gemini AI model for Siri (on Apple’s private cloud) and pay Google around $1B/year, as an interim boost to Siri’s reasoning and summarization abilities. Apple plans to keep developing its own AI (including a new 1T‑parameter model) and eventually switch off Google’s tech. If true, this deal would be a striking example of the old rivals collaborating – a patch to upgrade Siri in the short term, while Apple readies its next-generation AI.
