Apple’s self-driving car never shipped. But the silicon it forced Apple to invent is now the quiet foundation of the company’s entire AI strategy, according to The Verge AI, which broke down Mark Gurman’s latest Power On reporting. Early in the car program, Apple engineers hit a wall: autonomous driving needs heavy on-device AI processing, and no existing chip could handle it. The car processor never got finished. The Neural Engine that came out of that work did, and it changed everything.
That’s the story worth sitting with. A failed moonshot left behind the one piece of hardware Apple now leans on hardest.
From FaceID to the whole Mac lineup
The Neural Engine debuted in 2017 with the iPhone X and the A11 Bionic. Back then it did narrow work: computer vision for FaceID, Animoji, and early augmented reality. Modest stuff. The Verge AI reports that this early bet is what let Apple bring on-device AI to desktops through its M-series chips, putting the company ahead of most rivals on hardware.
What stands out here is the split between Apple’s hardware and its software. Its AI models and Siri have clearly lagged the rest of the industry. Its chips have not. That gap is the whole tension in Apple’s AI story right now.
Why on-device processing is the real strategy
Running AI on the chip instead of in the cloud isn’t just a speed play. It’s Apple’s privacy pitch. Less data leaves your device, so Apple can market security as a feature competitors struggle to match. That’s a hardware advantage dressed up as a values argument, and it works because the silicon backs it up.
Here’s why it matters now. The industry is split into two camps:
- Cloud-first AI (OpenAI, Google, most chatbots): powerful, but your data travels to someone else’s servers.
- On-device AI (Apple’s direction): more private, lower latency, no connection required, but limited by what the chip can hold.
Apple committed to the second camp years ago, almost by accident, because a car needed it.
The M6 skip and the M7 bet
The forward-looking news is the sharpest part of Gurman’s reporting. According to The Verge AI, Apple is skipping the Pro, Max, and Ultra versions of its upcoming M6 chip entirely. Instead it’s accelerating the M7, expected in the first half of 2027 with major Neural Engine upgrades.
There’s more. The M7 Ultra is expected to power a new Apple server product, with support for up to 1.5TB of RAM. That’s a signal. Apple has mostly kept its silicon inside consumer devices. A server chip means Apple wants to run its own AI infrastructure on its own hardware, rather than renting cloud capacity from Nvidia-powered data centers like everyone else.
Gurman has one of the most accurate track records on Apple’s roadmap in the business, so these aren’t idle rumors. When he flags a product skip and a server pivot, it’s worth planning around.
What to watch and what to do
For practitioners and businesses, a few takeaways:
- If you build for Apple platforms, the on-device AI runway is getting longer, not shorter. Design for local inference. The Neural Engine is only getting bigger.
- If you’re weighing privacy as a product feature, Apple is about to make it cheaper to deliver, because more compute stays on the phone or Mac.
- If you track the AI infrastructure race, an Apple server chip is a new front. It could reduce Apple’s dependence on Nvidia and reshape how it trains and serves models.
- Watch the software gap. Great chips don’t fix a weak Siri. Apple’s 2027 challenge is matching its hardware lead with models people actually want to use.
The lesson underneath all of this is that failed projects aren’t always wasted. Apple’s car burned years and billions and produced no car. It also produced the exact chip architecture the AI era rewards. By 2027, we’ll see whether Apple can finally pair that hardware with software worthy of it.
More detail is available in the original report from The Verge AI.