Apple's Neural Engine, the core of its on-device AI processing, was born from Project Titan, the company's defunct self-driving car program. Engineers building the car platform recognized they needed serious on-device AI compute. The car processor was never finished, but the work produced the Neural Engine, which debuted in the A11 Bionic inside the iPhone X in 2017.
Mark Gurman's latest Power On newsletter traces this lineage forward through Apple's current silicon roadmap, including details on the M6, M7 Pro, M7 Max, M7 Ultra, and M8. The early Neural Engine started narrow, handling Face ID and Animoji via computer vision. What it became is the reason Apple can run AI models locally at all.
The full piece is worth reading for Gurman's specific chip naming details and the roadmap timeline, not just the origin story. The argument running underneath is that Apple's biggest AI hardware advantage came from a program that publicly failed. That tension is the reason to go read the original.
[READ ORIGINAL →]