Apple's A20 Pro Chip Doubles AI Power in Two Completely Different Ways
Apple's A20 Pro chip, unveiled September 9, 2026, doubles its AI processing power through two distinct hardware upgrades rather than a single improvement. The chip features a 32-core Neural Engine (up from 16 cores) and new GPU-based Neural Accelerators that run AI math twice as fast, both working in parallel to handle different types of AI tasks on the iPhone 18 Pro.
What Exactly Did Apple Double in the A20 Pro?
The confusion around Apple's AI performance claims stems from conflating two separate pieces of silicon. The Neural Engine, Apple's dedicated AI processor since 2017, now runs 32 cores organized as two 16-core blocks instead of a single block. Separately, Apple added new Neural Accelerators built directly into the GPU cores, designed specifically for FP8 (8-bit floating point) math, the low-precision arithmetic that large language models rely on for fast inference.
Both numbers represent real doublings, but they serve different purposes. The Neural Engine handles fixed AI tasks like Face ID, photo processing, and Apple Intelligence features that the chip was purpose-built for. The GPU accelerators add general-purpose low-precision compute that can handle larger, more flexible AI model workloads. This distinction matters because it shows Apple is attacking the AI performance problem from multiple angles simultaneously.
Why Does This Hardware Matter for Your iPhone?
The expanded AI capacity directly addresses a critical gap in Apple's AI strategy. Earlier in 2026, Apple rebuilt Siri on Google's Gemini because its own on-device models couldn't power a full conversational assistant. More Neural Engine and GPU headroom is Apple's answer to that problem: the more capable the on-device model, the less every Siri request needs to leave the phone and travel to a cloud server.
This shift has real implications for speed, privacy, and reliability. Requests that stay on-device return answers faster, work offline, and don't depend on internet connectivity or remote infrastructure. iOS 27, arriving Monday, September 14, includes a rebuilt Siri AI that draws on personal context, app actions, and onscreen awareness, with conversations synced across Apple devices through iCloud.
What Other Improvements Come With the A20 Pro?
The Neural Engine and GPU accelerators aren't the only upgrades. The A20 Pro also includes a 7-core GPU delivering roughly 40% faster graphics performance, a 6-core CPU with new performance cores offering about 20% faster CPU performance, and 50% more memory bandwidth compared to the A19 Pro. The memory bandwidth increase is particularly important because it determines whether the larger Neural Engine can actually process data fast enough to be useful in practice.
The A20 Pro is Apple's first smartphone chip built on TSMC's 2-nanometer process, a smaller manufacturing node that allows more transistors to fit in the same die area. This process shrink creates the physical space for the larger Neural Engine and bigger GPU without increasing the overall chip size.
How to Understand Apple's AI Hardware Strategy
- On-Device vs. Cloud Trade-off: Larger on-device AI capacity means fewer requests need to travel to cloud servers, improving speed and privacy while reducing dependence on internet connectivity and remote infrastructure.
- Dual Hardware Approach: Apple is using both dedicated Neural Engine cores and GPU-based accelerators because different AI tasks benefit from different hardware designs, allowing the chip to handle multiple workloads efficiently.
- Manufacturing Process Advantage: The move to TSMC's 2-nanometer process provides the transistor density needed to pack more AI hardware without making the chip physically larger or consuming more power.
Apple's move reflects a broader industry pattern. Google's Frozen v2 chip targets cutting the cost of running Gemini at scale, and OpenAI's Jalapeño chip with Broadcom focuses on cheaper inference in data centers. Every major AI player is now designing custom silicon around inference cost and speed, not just training larger models.
When Will These AI Features Actually Arrive?
Siri AI ships as a beta on September 14 in English only, with French, Japanese, Korean, Portuguese, and Spanish following in October. The feature is not available to anyone under 13. Apple Intelligence features like Image Playground, which now produces photorealistic images, Calendar event creation from text, and Call Context in the Phone app are also arriving with iOS 27.
The expanded AI hardware in the A20 Pro represents Apple's bet that on-device AI will become increasingly important as users demand faster, more private, and more reliable AI assistants. By doubling two different types of AI silicon simultaneously, Apple is signaling that the future of AI on smartphones isn't about a single breakthrough, but about thoughtfully combining multiple hardware approaches to handle the diverse demands of modern AI workloads.