On September 12, 2017, Apple announced the iPhone X with the A11 Bionic chip, which included what Apple called a “neural engine.” In its press release Apple said “the new A11 Bionic neural engine is a dual-core design and performs up to 600 billion operations per second for real-time processing,” and described it as “designed for specific machine learning algorithms” that enable “Face ID, Animoji and other features.” It was one of the first dedicated neural-network accelerators shipped at consumer scale inside a phone.
The Neural Engine ran the math behind facial recognition, augmented reality and on-device image processing without sending data to the cloud. By doing inference locally, Apple could promise that biometric data used for Face ID stayed on the device, framing on-device AI as both a performance and a privacy feature.
The A11’s accelerator was the start of a line that Apple expanded every year and carried into its Mac silicon, and it set the template for neural accelerators that became standard in smartphones across the industry. Apple later built its Apple Intelligence features on the foundation of years of Neural Engine hardware.
Why business readers should care: the Neural Engine showed that AI compute would not live only in data centers - a meaningful share of inference would run on the device in a customer’s pocket. That split between cloud training and on-device inference shapes product design, privacy positioning and the chip strategy of every consumer hardware company.