Apple’s abandoned self-driving car work appears to have left behind something more useful than another Silicon Valley robotaxi ghost story: a chip roadmap built around local AI processing.
Bloomberg’s Mark Gurman reported in his Power On newsletter that Apple is speeding development of the M7 chip family, with major upgrades planned for the Neural Engine, the part of Apple silicon used for on-device AI work. Gurman also reported that Apple is skipping Pro, Max and Ultra variants of the coming M6 generation and pushing ahead to M7 instead.
The reported schedule puts the M7 in the first half of 2027. Gurman said an M7 Ultra is also expected to underpin a new Apple server product and could support up to 1.5TB of RAM. Apple has not announced those chips or the server product.
Project Titan’s useful wreckage
Apple’s self-driving car program, widely known as Project Titan, did not produce a shipping car. According to Gurman, though, the effort helped produce the Neural Engine, now central to Apple’s on-device AI pitch.
The reason is not mysterious. A self-driving platform needs to interpret sensor data quickly and locally. Early in the car project, Apple recognized that it would need strong AI processing on the device rather than relying on remote servers for every decision. The dedicated car processor was not completed, but the work fed into Apple’s broader chip design, Gurman reported.
The Neural Engine first appeared in the A11 Bionic chip used in the iPhone X. Apple initially used it for computer-vision features including Face ID, Animoji and augmented reality. Those were not the large language model features that now dominate AI marketing, but they required the same basic bet: run machine-learning workloads close to the user, on Apple-controlled hardware.
Apple later brought the Neural Engine to Macs through its M-series processors. That gave the company a hardware base for local AI before its software strategy caught up. Apple’s AI services have trailed rivals in visibility and rollout, but its silicon has given the company a credible technical story for keeping more processing on the device.
Local AI is also a privacy argument
Apple has leaned on that architecture in its privacy claims. If more AI work happens on an iPhone, iPad or Mac, less user data has to be sent to cloud systems. That does not make every AI feature private by default, and Apple still uses server-side processing for some tasks. It does make the Neural Engine more than a benchmark bullet point.
Gurman’s report suggests Apple is treating that hardware advantage as a core part of its next chip cycle. The M7 line, if the timeline holds, would make the Neural Engine a larger focus across consumer devices and, through the reported M7 Ultra server plan, Apple’s own infrastructure.
The car project may have failed at building a car. Its chip work is apparently still steering Apple’s AI hardware, which is a very Apple outcome: cancel the vehicle, keep the silicon, and pretend the detour was part of the map.
This story draws on original reporting from The Verge.
