Please sign up to get our Binary Response articles directly in your inbox!
Google’s I/O 2026 keynote felt less like a routine developer keynote and more like a rescue mission for Apple’s AI ambitions. For four years, Apple has largely whiffed on the generative AI moment, culminating in that overproduced Apple Intelligence demo at WWDC 2024 that, as we later learned, was closer to a concept video than a working product. The promised new Siri still hasn’t arrived, with Apple repeatedly delaying core AI features until 2026, leaving users and developers wondering if the company had simply missed the AI turn.
Now, thanks to the Gemini partnership, Google is effectively backfilling Apple’s homework. You can see why Apple caved. Reports over the last two years chronicled Apple’s internal struggles, from failed attempts to use large language models and aging code to the repeated slippage of its AI roadmap.
The big 2024 promise was a more contextual Siri that could act across apps, but those capabilities were quietly pushed out, then delayed again, and are now targeted for sometime in 2026. Apple announced back in January that it would lean on Google’s Gemini models and cloud infrastructure for Apple Foundation Models and a major Siri revamp looked like a bold partnership as well as an admission that it couldn’t catch up alone.
Yesterday’s I/O keynote put a spotlight on exactly what Apple is paying them for. Google introduced the Gemini 3.5 family, led by Gemini 3.5 Flash, a lighter-weight model of intelligence while being fast and cheap enough to serve as the new default in the Gemini app and AI mode in Search. Google emphasized not just raw IQ but the ability to coordinate agentic workflows, with 3.5 Flash tuned for long-running, multi-step tasks rather than just chat replies.
Crucially, Google framed all of this as part of a broader shift by showing that Gemini isn’t just a single assistant anymore; it’s a platform. Displaying the ecosystem logic forming. Third-party hardware like Walmart’s upcoming Gemini-powered smart speaker is being teased, suggesting that Google is happy to have Gemini show up in places where Hey Google historically dominated.
It’s tempting to look at the Apple deal and assume this platform pivot is driven by the prospect of powering the next Siri. But the timing and the regulatory backdrop tell a different story. Over the last few years, Google has lost two major U.S. antitrust cases — one over its dominance in search and another over digital advertising — and courts have started imposing remedies that specifically restrict exclusive defaults and bundling across products like Search, Chrome, Google Assistant, and now the Gemini app. Regulators explicitly want Google to stop locking up distribution and to make its data and services more accessible to rivals.
Gemini as a platform looks less like a nice-to-have and more like a survival strategy. By turning Gemini into something any app, device maker, or even cloud customer can plug into — with configurable agents, open APIs, and the option to run on non‑Google infrastructure — Google can argue it’s enabling competition rather than crushing it. The Apple partnership fits that narrative perfectly too; instead of paying Apple to make Google Search the default, Google now licenses models that Apple can wrap in its own UX, branding, and privacy guarantees.
For Apple, the appeal is obvious. The company has been under pressure for dragging its feet in AI, and its own Apple Intelligence stack has stumbled both technically and in public perception. At the same time, Apple still wants to preserve its story about privacy and on‑device intelligence, which is why the Gemini deal is structured so that models run either on device or through Apple’s Private Cloud Compute rather than a generic Google cloud endpoint. Apple gets a state‑of‑the‑art model family without giving up the narrative that your data stays your data.
Where this really matters is Siri itself. The last time Siri felt meaningfully more powerful to power users was in 2018, when Apple introduced Siri Shortcuts and later the Shortcuts app, letting people script multi-step automations and trigger them via voice. That was clever, but it always put the burden on users to design the logic. Gemini’s architecture is built for the opposite since the model does the planning, decomposing a request into steps, executing them across tools and apps, and adapting as it goes.
Gemini 3.5 is optimized for long‑horizon tasks, multi‑agent coordination, and multimodal input — text, images, video, even code — which matches almost exactly what Apple has been promising and failing to ship with the Siri update back in 2024. Som now Apple can plug that into their ecosystem, with deep hooks into Messages, Mail, Photos, and Shortcuts, and you can imagine a version of Siri that finally lives up to their 2024 demo which found the flight info buried in your email, cross‑checking it against your calendar, and then texting your friend that you’re running late, all from a single natural language request.
Here’s the thing, folks: If Apple pulls this off, it will be the first time since the original Shortcuts era that Siri feels like a platform in its own right rather than a glorified voice remote. The irony, of course, is that the platform muscles won’t really be Apple’s — they’ll belong to Google’s Gemini, wrapped in Apple’s UX and privacy story. And the deeper irony is that Google isn’t doing this just to help Apple; it’s doing it because regulators have made it clear that the age of tightly controlled, default‑only distribution is over.
With that… Google is “saving” Apple from its AI detour, handing over the model architecture that can finally make Siri feel modern again. But in the process, Apple is also helping Google make its best possible case to governments around the world because the can use Gemini to defend against being a monopoly, but instead being an open platform that even its fiercest rival is free to use.
When you work with technology everyday you will always understand the business side of things differently!