Apple Is Paying Google $1 Billion a Year to Make Siri Not Terrible

The Billion-Dollar Admission

Here's something you don't see every day: the world's most valuable company paying a competitor $1 billion per year because it can't build a decent AI assistant on its own. That's the reality of Apple's deal with Google, announced in January 2026, to power the long-awaited Siri overhaul with Google's Gemini model. The new Siri runs on what Apple internally calls the "Apple Foundation Model version 10," which is really a custom build of Gemini, white-labeled so users never see Google's branding.

The deal flips the traditional Apple-Google financial relationship on its head. For years, Google has been paying Apple billions to remain the default search engine on Safari. Now Apple is the one writing the checks, and the reason is simple: after years of Siri being the butt of every AI joke, Apple decided it couldn't catch up to OpenAI, Google, and Anthropic on its own.

The Delays Are Piling Up

The problem is that even with Google's help, the new Siri keeps slipping. A Bloomberg report from February 11 revealed that internal testing has hit serious snags. Features originally planned for iOS 26.4 (the March update) are now being pushed to iOS 26.5 (May) or possibly even iOS 27 (September). When Apple released the iOS 26.4 beta on February 17, none of the new Siri features were included, all but confirming the delay.

Apple SVP Craig Federighi admitted that the company's first attempt at rebuilding Siri with a first-generation architecture "ended up being too limited" and that "Siri was not able to match Apple's high standards." That's corporate-speak for: it didn't work. The pivot to a second-generation LLM-based architecture explains both the Google partnership and the timeline slippage.

The delays also threaten four upcoming hardware products that depend on the new Siri: a smart home hub, a smart doorbell, AR glasses, and a new Apple TV. Without a working AI assistant, these products are significantly less compelling.

The March 4 Event: More Questions Than Answers

Apple is hosting a "special Apple Experience" on March 4 across three cities simultaneously: New York, London, and Shanghai, starting at 9:00 AM ET. The event is expected to showcase new hardware, including the iPhone 17e, an affordable MacBook, and an M4 iPad Air. But the elephant in the room is Siri.

Will Apple demo the new AI assistant even if it's not shipping yet? Will the hardware event distract from the AI delays? Or will Apple try to set expectations, showing off limited Siri improvements while promising the full overhaul later? The multi-city format suggests Apple wants to make a big splash, but the substance of the AI story may not match the spectacle of the presentation.

The Privacy Paradox

Apple has built its brand on privacy, and the Gemini deal creates an awkward tension. Tim Cook has promised that the new Siri will run "on-device and in Private Cloud Compute," Apple's system of Apple Silicon servers designed to process data without exposing it to third parties. The idea is that data gets anonymized before it ever touches Google's infrastructure, and nothing is stored or used for model training.

But then Google CEO Sundar Pichai described Google as Apple's "preferred cloud provider" for the same partnership, which sounds a lot like data is going to Google's servers. The contradictory CEO statements haven't been reconciled, and privacy advocates are already raising red flags. If advanced Siri features eventually need to bypass Private Cloud Compute and run directly on Google's TPU hardware for performance, Apple's core privacy promise starts to look hollow.

What the Competition Thinks

The deal is a validation of Google's position in the foundation model market, and a blow to OpenAI and Anthropic, both of which were reportedly in the running for the partnership. Elon Musk called the arrangement "an unreasonable concentration of power," though given his own AI ambitions with xAI, his objection is hardly disinterested.

The broader industry takeaway is striking: even Apple, with its $3 trillion market cap and world-class engineering talent, concluded it couldn't build frontier AI models competitive with what Google, OpenAI, and Anthropic are producing. That says something profound about how concentrated the AI model-building capability has become. If Apple can't do it alone, very few companies can.

The Strategic Bet

Apple's defenders argue this is classic Apple strategy: don't invent the technology, perfect the integration. Apple didn't build the first MP3 player, smartphone, or tablet, but it built the best versions of each by focusing on user experience. The argument is that by white-labeling Gemini and wrapping it in Apple's privacy architecture and ecosystem integration, the result will be a better product than what Google offers directly.

The skeptics see it differently. Paying $1 billion a year to a competitor for core functionality makes Apple a distribution layer, not a technology leader. If the AI powering your assistant is someone else's model, your competitive moat shrinks to hardware design and ecosystem lock-in. And with the new Siri repeatedly delayed, even the integration story isn't convincing yet.

What to Watch

March 4 is the immediate focus. Watch for whether Apple actually demos the new Siri or sticks to hardware announcements. The iOS 26.5 beta cycle (expected in April) will reveal whether the delayed Siri features are actually ready. And keep an eye on the privacy story: as more details emerge about how data flows between Apple's PCC and Google's infrastructure, the "privacy-first AI" narrative will either hold up or collapse. For 2.5 billion Apple device owners, the stakes couldn't be higher.