Apple’s Siri AI is finally good enough, but good enough for what exactly?
The new Siri pulls Apple out of its AI crisis. That might be true, but it raises more questions than it answers.
Bloomberg’s Mark Gurman says Apple’s new Siri might have finally done enough. In his latest Power On newsletter, Gurman wrote that the rebuilt assistant is “just good enough” to pull the company out of what he called an AI crisis. That framing feels about right.
Apple has been the quietest member of big tech’s AI conversation for a while now. Google, Microsoft, and Meta have been pouring hundreds of billions into AI infrastructure, while Apple spent roughly $13 billion on capital expenditures across all of fiscal 2025. So instead of building a frontier model, it signed a deal with Google in January 2026, paying around $1 billion a year for a custom Gemini model to power Siri AI, which was then unveiled at WWDC 2026 alongside iOS 27.
The new assistant is a hybrid setup. Lighter requests run on-device, and more complex ones route through Apple’s Private Cloud Compute infrastructure on the Gemini backend. Apple says user data isn’t stored externally, and its deal with Google prevents Siri queries from being used to train future Gemini models.
What you actually get with Siri AI is a back-and-forth conversation that holds context, access to your emails, messages, and photos, multi-app actions, and a dedicated standalone app. For everyday tasks, it’s genuinely useful in a way old Siri never was.
The harder question is what people will actually reach for it for. Setting timers and sending messages worked before. The new stuff, cross-app actions, personal context, conversational memory, requires a habit that most users haven’t built even with more capable tools already on their phones.
And then there’s the on-device model itself. It’s probably the best on-device AI running on any smartphone right now, but that’s still a pretty low ceiling. On-device models are smaller by necessity, and no matter how well Apple optimizes for its own hardware, they’re not going to match what a full cloud model can do. Anything complex or creative is still better handled by ChatGPT or Gemini, and most people already have those on their phones.
On the other hand, Apple is playing it safe. It gives its users enough AI to stay relevant without betting the house on infrastructure that may or may not pay off. If the bubble ever bursts, Apple will be the one holding the least debt from it.