Apple's AI chips may owe their strength to a project that never shipped: the company's canceled self-driving car. Apple's automotive effort never really got off the ground, but the on-device processing it demanded helped seed the Neural Engine that now powers the company's silicon, according to Bloomberg's Mark Gurman.
Early in the car's development, Apple realized it would need serious on-device AI horsepower to handle real-time driving tasks. The vehicle's processor was never finished, but as Gurman details in his Power On newsletter, that work fed directly into the Neural Engine, the backbone of Apple's on-device AI today. Follow the bigger picture in our AI coverage.
From car project to the Neural Engine
The effort, long rumored under the codename Project Titan, was ultimately shelved, but its technical requirements outlived it. The most lasting result was the Neural Engine, which made its debut with the iPhone X and the A11 Bionic. In those early days it handled computer-vision tasks, powering Face ID, Animoji, and augmented-reality features rather than the generative AI we associate with the term now.
The connection makes sense. Self-driving systems lean heavily on real-time perception — reading roads, signs, and obstacles on the fly — which is exactly the kind of machine-learning workload a dedicated neural processor is built to accelerate. The car never launched, but that requirement pushed Apple to invest early in silicon that could think locally.
By building that groundwork, Apple positioned itself as an early leader in edge AI. It later brought the Neural Engine to the desktop with its M-series chips, extending on-device processing from phones to Macs. You can track that lineage in our laptop and Mac coverage.
Hardware strength, software questions
Apple's AI story has been lopsided. Its software efforts have lagged much of the industry, but its hardware has been consistently impressive. That silicon advantage is also what lets Apple lean on privacy as a selling point: the more work a chip can do locally, the less user data has to travel to the cloud.
In practice, on-device processing means features like photo analysis, dictation, and language tasks can run on the phone or Mac itself, a pitch that resonates as rivals push cloud-first AI. It's a rare case of a scrapped product leaving behind an advantage that outlived the thing it was built for.
What comes next: M7 and a server chip
Apple is making AI hardware a cornerstone of its roadmap. According to Gurman, the company is skipping the Pro, Max, and Ultra versions of its upcoming M6 chip and accelerating the M7 instead, which should arrive in the first half of 2027 with significant Neural Engine upgrades.
The most eye-catching detail is the M7 Ultra. Apple is reportedly positioning it as the basis for a new server product, with support for up to 1.5TB of RAM, a sign the company wants to run heavier AI workloads on its own chips rather than someone else's. A server-grade Apple chip would let the company keep more of its AI stack in-house instead of leaning on third-party hardware, the same self-reliance that has defined its move away from other suppliers. For how this shapes its phones, see our mobile coverage.
What is Apple's Neural Engine?
It's the dedicated AI block inside Apple's chips, first introduced with the A11 Bionic in the iPhone X. It accelerates machine-learning tasks, from Face ID to on-device language features, without offloading everything to the cloud.
When is the Apple M7 coming?
Gurman reports the M7 is targeted for the first half of 2027, with Apple skipping the higher-end M6 variants to get there. An M7 Ultra with up to 1.5TB of RAM is expected to anchor a new Apple server product.
