The world's largest hardware company just announced that selling you a rectangle is no longer the point — the chips inside it are.
The Summary
- Apple's new CEO John Ternus positioned the iPhone 18, first-ever foldable Duo, Apple Watch Series 12, and AirPods 5 as an "intelligent personal hub" built explicitly for on-device AI processing.
- The iPhone 18 Pro line runs a dual 16-core Neural Engine on the A20 Pro chip — double the on-device AI compute of previous models — to power the redesigned Siri.
- Apple's AI-powered Siri, originally announced in Spring 2024, was delayed until 2026 due to "developmental roadblocks and performance issues" — a rare public admission of an AI stumble.
- Hardware specs are now secondary to whether your phone can run your personal AI agent without phoning home to the cloud.
The Signal
Apple spent two years learning what Google and Microsoft learned faster: you can't ship AI assistants that work like 2015 chatbots and call it innovation. The company announced its AI-powered Siri overhaul in Spring 2024, then went silent as users waited. Now we know why: the thing didn't work. Developmental roadblocks is corporate speak for "our agents hallucinated your calendar into oblivion." That admission matters because Apple doesn't usually telegraph its failures.
The hardware announcements — iPhone 18 Pro, the foldable Duo, Apple Watch Series 12 and Ultra 4, AirPods 5 — are less interesting than the silicon strategy underneath them. Apple added a second Neural Engine to the A20 Pro chip, delivering dual 16-core processing dedicated entirely to running AI models locally. This isn't about speed. It's about where inference happens.
"Apple has been adding to the power of the chips inside its devices so they can run more AI processes on the device itself, rather than in the cloud."
On-device inference means your agent doesn't need to ask Apple's servers for permission to read your texts, summarize your emails, or understand what's on your screen. It means:
- Lower latency for context-aware responses
- Privacy by default, not by policy
- Agents that work on airplanes and in network dead zones
The new Siri can draw on personal context across messages, emails, and photos; use onscreen awareness to act on what you're looking at; and answer questions about virtually any topic. That last part — "virtually any topic" — is doing heavy lifting. If it's true, Apple just turned Siri from a kitchen timer into an agent that competes with ChatGPT and Claude, except it lives in your pocket and already knows who your dentist is.
New CEO John Ternus framed the entire product line as purpose-built vehicles for AI, not as phones that happen to run some AI features. That's a positioning shift. Apple is no longer selling you a camera that makes calls. It's selling you an inference engine that fits in your hand. The foldable Duo is less about screen real estate and more about sensor surface area — more ways for the device to understand context, capture input, and feed your agent.
The Implication
If Apple can actually deliver on-device agents that work reliably, every other hardware maker is now playing catch-up on silicon, not software. The race isn't who has the best cloud model anymore. It's who can fit the best model into a phone chip without melting the battery. Watch what happens to inference chip startups in the next six months. And watch whether developers start building for Apple Intelligence the way they once built for the App Store. The foldable is a footnote. The real product is the Neural Engine.