Apple’s long-promised, AI-heavy Siri update might be closer than we thought. According to a new report, Apple plans to ship a revamped Siri in iOS 26.4 this coming spring. The twist: the new Siri would not be powered by Apple’s own large language models. Instead, the report claims Apple contracted Google to build a custom model based on Gemini, and that model would run on Apple’s private cloud.
If true, this would be a surprising turn. Apple has made a big public push to build its own AI capabilities and to emphasize privacy and vertical control. But building a top-tier language model is expensive and time consuming. The claim says Apple tested multiple partners — including Anthropic’s Claude — and found the Claude models technically stronger. Yet, for financial and business reasons tied to the companies’ long-standing search deal, Apple allegedly chose Google’s Gemini-based option.
What that would look like in practice is straightforward to picture: Apple pays Google to create a custom Gemini variant tuned to Apple’s needs. That model is then hosted on Apple’s cloud servers and accessed by Siri. There would be no Gemini name or Google branding in the final product. Apple would present the experience as its own work — new Siri, new capabilities — while the underlying model would be a Google-built system running behind the scenes.
Why might Apple do this? The report’s logic is simple. Apple and Google already have a big commercial relationship: Google pays Apple to be the default search engine across Apple devices. That arrangement creates a channel for more complex collaborations. From Apple’s perspective, licensing a custom model could be faster and cheaper than building everything in house. From Google’s side, providing a bespoke Gemini instance to Apple would be a lucrative, high-value deal and a way to keep Gemini widely used even on rival hardware.
There are a few things to keep in mind. Reports like this often rely on anonymous sources, and companies sometimes test multiple partnerships before making a final choice. If Apple does rely on a Google-built model, it raises questions about marketing, transparency, and privacy. Will Apple disclose that a third party built the model? Will Google have access to user prompts or only to aggregated, privacy-preserving telemetry? The claim says the partnership would be quietly handled — no Gemini badge, no public acknowledgement from either side. That secrecy could upset users and privacy advocates who expect Apple to do its AI work internally.
There’s also a user-experience angle. A Gemini-based Siri could arrive with stronger conversational skills, better context handling, and broader knowledge. Apple’s tight integration across devices might let the company splice Siri’s voice, personalization, and on-device signals into that cloud model in a way that feels native. That would give users a better assistant without Apple having had to train every model piece from scratch.
Bottom line: this report, if accurate, describes a pragmatic but awkward solution. Apple would speed up a Siri overhaul by outsourcing core model work to a competitor while keeping the public story simple: Apple built Siri. That approach would likely keep the user experience solid and protect the public image Apple prefers. But it would also underline how complex and interdependent the AI world is — even big rivals can become quiet partners when technology and money line up.
Also Read | OpenAI Locks in $38 Billion AWS Deal to Run ChatGPT on Hundreds of Thousands of Nvidia GPUs
Source: Bloomberg
