At its WWDC 2026 keynote on June 8, Apple unveiled a rebuilt Apple Intelligence platform powered by a new family of Apple Foundation Models developed together with Google using the technology behind Gemini, alongside a revamped assistant now called Siri AI. The models run both on the device and on Apple servers through Private Cloud Compute. The arrangement traces back to a multiyear partnership the two companies announced on January 12, 2026, reportedly worth roughly one billion dollars a year for access to a custom Gemini model said to carry about 1.2 trillion parameters.

What did Apple actually announce at WWDC 2026?

Apple introduced the third generation of its Apple Foundation Models, a lineup of five models that for the first time is built on Google Gemini technology rather than entirely in house. Two of those models live directly on iPhone, iPad, and Mac, led by a compact 3 billion parameter model that handles everyday tasks, while heavier reasoning runs in the cloud. Craig Federighi, Apple's senior vice president of software engineering, described the result as a dramatically more capable and conversational assistant.

The headline product is Siri AI. The assistant finally behaves like a full chatbot, with its own dedicated app where conversations are saved and synced privately across iPhone, iPad, Mac, and Apple Vision Pro through iCloud. Apple also pushed the new intelligence into core apps, adding realistic image creation, advanced photo editing, and visual question answering across the system.

Does this mean Siri now runs on Google?

Not exactly, and Apple was careful to draw that line. Apple says the models were built through distillation and training on Gemini technology, not by dropping Google code into iOS. Training happened on Google cloud TPUs, but the finished models are Apple's own, and hosting runs through Apple's Private Cloud Compute. When you talk to Siri AI, you are not touching Google Search or Gemini agents directly.

Privacy is the spine of the whole pitch. Apple insists user data is never stored or shared, not even with Apple, and Google has confirmed it receives no Apple user data from the arrangement. That framing matters, because the only way Apple could sell a rival's models to its privacy conscious base was to wall off the data completely.

Why does this matter for the rest of the AI industry?

This is the clearest admission yet that building a frontier model from scratch is brutally hard, even for the richest company on earth. Apple spent years promising a smarter Siri and kept missing, and rather than keep losing, it paid a competitor to supply the foundation. That is a pragmatic move, and it tells you the moat in AI right now is the base model, not the polish on top.

It also reshapes the competitive map. Google now sits underneath the two largest mobile platforms in the world, since it powers both Android intelligence and, quietly, Apple's. For everyone watching where AI is headed, the takeaway is simple. Distribution still belongs to Apple, but the engine increasingly belongs to whoever trains the best model, and right now Apple decided that was not going to be Apple.

Check out what else is trending at Hacker News