Apple’s Siri AI Needs 12GB RAM — iPhone 16 Pro and iPhone 17 Base Miss the Cut at WWDC 2026

The Google collaboration deserves closer attention. In January 2026, Apple and Google announced a multi-year partnership. Apple’s machine learning team used Google’s Gemini technology as a training signal — sometimes called distillation — to refine the AFM 3 models. Gemini is not running live inside your iPhone; the final production models are Apple’s own. But for AFM 3 Cloud Pro specifically, training infrastructure, NVIDIA inference hardware, and cloud hosting all involve Google. Apple confirmed the collaboration but has been precise in saying the models are “custom-built in collaboration with Google,” not Gemini deployments.

For users, the collaboration matters less than what the models can actually do. Siri AI in iOS 27 can read what’s on your screen and answer questions about it. It can pull context from texts and emails — so asking “Where’s Jeff’s new place?” pulls the address from a recent message. Apple VP of Siri Engineering Mike Rockwell demonstrated asking Siri for directions to a landmark from a photo, with a stop at a friend’s house along the way. These are everyday tasks, not enterprise workflows. That positions Siri AI differently from Anthropic’s Claude and OpenAI’s ChatGPT, both of which are built around business and productivity use cases.

Apple’s financial position heading into this AI push is strong. The company posted $111.2 billion in revenue for the March 2026 quarter — its best March quarter ever — with iPhone revenue of $57 billion, up 22% year over year, driven by iPhone 17 demand. Tim Cook described the quarter’s results in prepared remarks to investors. Services, which includes iCloud+, reached an all-time revenue record of $31 billion, up 16%.

Despite that, analysts are still watching for evidence that AI features drive actual behaviour changes — upgrades to newer hardware, iCloud+ subscriptions, or both. Morgan Stanley analysts noted in a research note following WWDC 2026 that Apple’s announcements provided “clearer paths to monetizing AI,” though they also described Apple’s AI progress as a “marathon, not a sprint.” Barclays analysts were less generous, writing that “updates felt more evolutionary vs revolutionary” and questioning Apple’s monetisation strategy in a research note after the event.

Some Siri AI features tied to iCloud+ subscriptions include higher limits on image generation and descriptions of footage from HomeKit-compatible smart home cameras. For features like expressive voices and advanced dictation, the gate is hardware, not subscription. More than 1.3 billion iPhones currently in use lack the computing power or memory to run those two features, according to estimates from Bloomberg Intelligence analyst Anurag Rana. Even among users who upgrade this fall, the buying decisions will likely be driven by battery life and performance — not AI features specifically, according to Paul Schell, a senior analyst covering AI for ABI Research.

The stickiness argument is different. Analysts at IDC have noted that Siri AI’s growing personal context — history of your conversations, addresses, preferences, calendar — creates a switching cost. As Nabila Popal, a senior director at IDC, put it regarding users who build up Siri AI context: “You’re going to leave that behind when you switch completely.” The AI model landscape across platforms has sharpened considerably in 2026, with competition from Anthropic and other AI ventures intensifying.