Apple’s chip announcements usually lead with a camera feature or a battery claim. This week’s was different. On September 9, 2026, Apple used its keynote to introduce the A20 Pro, the silicon inside the new iPhone 18 Pro and iPhone 18 Pro Max, and the headline spec had nothing to do with photography. It’s the manufacturing process. The A20 Pro is Apple’s first iPhone chip built on a 2-nanometer node, and it pairs that shrink with a 7-core GPU, a 32-core Neural Engine, and what Apple calls a 50% jump in memory bandwidth over the A19 Pro, according to Notebookcheck.
That combination matters more than it might look on a spec sheet. A node shrink alone is routine; Apple has done one nearly every year since 2023. Pairing it with a wider memory pipe and a doubled Neural Engine, in the same generation, is the kind of architectural bet Apple typically spreads across two or three product cycles. This article breaks down what actually changed inside the A20 Pro, why the 2-nanometer process is the real story, and what it signals about the race between Apple, Qualcomm, and Samsung to control the most advanced chip manufacturing in the smartphone industry.
Don't miss new tech stories on Google
Add Tech Insider once in the Google app and our stories appear in your news suggestions.
A20 Pro Specs: 7-Core GPU, 32-Core Neural Engine, 6-Core CPU
The core layout of the A20 Pro breaks down into three blocks that all grew from the A19 Pro. The GPU moves from six cores to seven, and Apple says the new graphics architecture delivers up to 40% faster performance than its predecessor, per reporting from Wccftech. The Neural Engine doubles in size, from a single 16-core block on the A19 Pro to two 16-core blocks running in parallel on the A20 Pro, for 32 cores total. And the CPU keeps its six-core layout but adds new “Super” performance cores, contributing to a CPU speed gain that Wccftech pegs at roughly 20% generation over generation.
None of these gains happen in isolation. Apple’s chip team has structured the A20 Pro so that GPU, Neural Engine, and memory controller improvements reinforce each other rather than compete for the same bandwidth budget. That’s the part outlets like 9to5Mac and MacRumors both flagged in their coverage of the announcement: the memory bandwidth increase isn’t a standalone feature, it’s what makes the GPU and Neural Engine gains sustainable under real workloads instead of throttling within minutes.
What TSMC’s 2-Nanometer Process Actually Changes
The A20 Pro is built on a 2-nanometer node widely reported to be TSMC’s N2 process, marking the first time an iPhone chip has shipped on this class of manufacturing, according to Tom’s Hardware. Node names get thrown around a lot in chip coverage, so it’s worth being precise about what actually changes when a foundry moves from 3nm to 2nm. TSMC’s N2 process swaps the FinFET transistor structure used since the 7nm era for gate-all-around (GAA) nanosheet transistors, which give engineers finer control over current leakage at smaller sizes. In practical terms, that translates to two possible trade-offs a chipmaker can choose: run the chip at the same clock speed for meaningfully less power, or push clock speed higher within the same power envelope.
Industry estimates cited by outlets covering the A20 Pro launch put the N2 advantage at roughly 10 to 15% faster performance at equivalent power, or a 25 to 30% cut in power draw at equivalent performance, compared with TSMC’s N3E process used in earlier iPhone generations. Apple appears to have split the difference, taking some of that headroom as raw speed and some as efficiency, which lines up with separate reporting that frames the A20 Pro as up to 15% faster and roughly 30% more power efficient than the A19 chip family. For iPhone 18 Pro owners, that shows up less as a benchmark number and more as a phone that stays in its top performance tier longer before thermal throttling kicks in during sustained gaming or video export sessions.
The Memory Bandwidth Story: A Wider Bus Behind the Numbers
The “50% more memory bandwidth” figure Apple is promoting sounds abstract until you see where it comes from. Supply-chain reporting points to the A20 Pro moving to a 96-bit memory bus, up from the 64-bit interface used on the A19 Pro and earlier chips, which is exactly the 50% increase Apple is citing. A wider bus means more data lanes moving between the chip and its RAM simultaneously, rather than a faster clock pushing the same number of lanes harder. Some pre-launch leaks pointed toward a shift to LPDDR6 memory, though Apple’s own materials, as parsed by outlets covering the keynote, describe the gain in terms of bus width rather than confirming a memory standard change outright.
Tom’s Hardware described the result as the widest memory interface Apple has ever shipped in an iPhone. That’s a meaningful claim because memory bandwidth is often the quiet bottleneck behind stuttering frame rates and slow AI inference, even when the compute cores themselves have plenty of headroom. A 7-core GPU or a 32-core Neural Engine can only process data as fast as it arrives, so widening the pipe is arguably a bigger real-world upgrade than the core count increases themselves. It’s also consistent with reports that Apple is exploring wafer-level multi-chip module (WMCM) packaging, which places memory physically closer to the compute die to cut latency further, though that detail has circulated mainly in pre-launch leaks rather than Apple’s own announcement.
GPU Performance: What 40% Faster Graphics Means for Gaming
Apple’s graphics claims tend to run ahead of what third-party benchmarks eventually confirm, so the 40% figure should be read as a ceiling under ideal conditions rather than a guaranteed everyday gain. Still, the seventh GPU core combined with the wider memory bus gives the A20 Pro a genuinely different profile than a simple core-count bump would. Wccftech’s reporting also notes the newest generation of Neural Accelerators embedded in the GPU pipeline delivers double the FP8 performance of the previous design, a detail that matters increasingly for games and creative apps that lean on machine-learning-assisted rendering techniques like upscaling and frame interpolation rather than pure rasterization.
For the iPhone 18 Pro specifically, this positions Apple to court the same class of mobile game studios that have spent the past two years building titles around on-device AI upscaling on Android flagships. A 7-core GPU with meaningfully more bandwidth to draw on is the kind of upgrade that shows up in sustained frame rates during 20-minute gaming sessions, not just in a 30-second synthetic benchmark run. Whether third-party test suites validate Apple’s 40% figure once review units ship widely will be the real test, and Tech Insider will be watching independent GPU benchmarks as they land in the coming weeks.
CPU Gains: New Super Cores Behind the 20% Claim
The CPU side of the A20 Pro gets less attention than the GPU and Neural Engine changes, but it’s not standing still. Apple kept the six-core layout from the A19 Pro but introduced new “Super” performance cores, part of what Wccftech’s reporting frames as a roughly 20% CPU speed improvement generation over generation. That’s a smaller headline number than the GPU’s 40%, which tracks with how Apple has allocated its node-shrink gains over the past several generations: graphics and AI silicon tend to get the more dramatic year-over-year jumps, while CPU cores see steadier, more incremental gains.
That pattern isn’t unique to Apple. Across the smartphone chip industry, CPU performance has plateaued relative to GPU and NPU gains simply because most day-to-day phone tasks, from scrolling to messaging, stopped being CPU-bound years ago. Where the CPU improvements matter most now is in sustained multi-core workloads: video editing exports, code compilation for developers testing on-device, and increasingly, running larger on-device AI models that lean on CPU threads alongside the Neural Engine.
A20 Pro vs A19 Pro: Full Spec Comparison
The table below lines up the confirmed specification changes between the A19 Pro and the A20 Pro, based on Apple’s announcement and reporting from Notebookcheck, Wccftech, and Tom’s Hardware.
| Specification | A19 Pro | A20 Pro |
|---|---|---|
| Process node | TSMC 3nm-class | TSMC 2nm (N2) |
| GPU cores | 6-core | 7-core |
| GPU performance claim | Baseline | Up to 40% faster |
| Neural Engine | 16-core (single block) | 32-core (dual 16-core blocks) |
| Neural Accelerator FP8 throughput | Baseline | 2x |
| CPU cores | 6-core | 6-core with new Super cores |
| CPU performance claim | Baseline | Up to 20% faster |
| Memory bus width | 64-bit | 96-bit |
| Memory bandwidth claim | Baseline | 50% higher |
| Debut device | iPhone 17 Pro / Pro Max | iPhone 18 Pro / Pro Max |
Apple’s Node Shrink Timeline: From A17 Pro to A20 Pro
Context helps explain why the A20 Pro’s 2-nanometer jump is being treated as a bigger deal than a typical annual refresh. Apple’s Pro-tier chips have moved through the 3-nanometer generation in stages rather than a single leap: the A17 Pro in 2023 was the first iPhone chip on TSMC’s initial 3nm process, the A18 Pro in 2024 and A19 Pro in 2025 both used refined variants of that same 3nm family. The A20 Pro is the first chip to break out of that node entirely, moving Apple’s most advanced iPhone silicon to a new transistor structure three years after the initial 3nm transition. That cadence, three years on one node family followed by a full node change, mirrors how Apple has historically staged its biggest architectural bets, including the original transition to 3nm itself.
| Chip | Year | Process Node | Debut Device |
|---|---|---|---|
| A17 Pro | 2023 | TSMC 3nm (first generation) | iPhone 15 Pro |
| A18 Pro | 2024 | TSMC 3nm (refined) | iPhone 16 Pro |
| A19 Pro | 2025 | TSMC 3nm (refined) | iPhone 17 Pro |
| A20 Pro | 2026 | TSMC 2nm (N2) | iPhone 18 Pro |
How A20 Pro Stacks Up Against Snapdragon and Exynos
Apple is not manufacturing in a vacuum, and the A20 Pro’s 2-nanometer status is really a story about foundry timing as much as chip design. TSMC’s N2 process has been positioned as the most advanced node in mass production for smartphone silicon, and by claiming the first commercial iPhone chip on it, Apple gets a window, likely measured in months rather than years, where its GPU and Neural Engine architecture runs on more advanced transistors than the competing flagship chips shipping in most Android devices. Tech Insider has covered Qualcomm’s own roadmap moves in detail, including its dual Snapdragon 8 Elite Gen 6 chip strategy, and the node race between Apple, Qualcomm, and Samsung Foundry is increasingly the real battleground, more than any single core-count spec.
Samsung Foundry has its own 2-nanometer ambitions for future Exynos chips, and reports throughout 2026 have described the company racing to close the gap with TSMC on advanced node yields. The practical effect for consumers is less about a single “who wins” headline and more about a compressed timeline: node advantages that used to separate Apple from Android flagships by a full year are narrowing to a matter of months as more foundries bring 2nm-class capacity online. That compression is also why memory bandwidth and packaging innovations, not just raw node numbers, are becoming a bigger differentiator between competing chip families, since they’re harder for a rival foundry to simply replicate once a process node becomes broadly available.
Neural Engine Architecture: Why Two 16-Core Blocks Beat One 32-Core Block
It’s worth pausing on how Apple chose to double the Neural Engine, because the implementation detail matters. Rather than designing a single monolithic 32-core Neural Engine, Apple built two 16-core blocks that operate in parallel. That’s a meaningfully different engineering decision than it sounds. Running two independent blocks in parallel can offer better fault isolation and more flexible workload scheduling, letting the chip route different AI tasks to different blocks simultaneously rather than queuing everything through one pipeline. It also tends to be a more conservative, lower-risk path to doubling throughput than redesigning a single unit from scratch, which likely explains why Apple was comfortable pairing it with an equally ambitious node shrink in the same generation. Tech Insider has covered the AI-compute implications of this specific design choice in more depth in a separate look at what it means for Apple Intelligence performance.
Market and Supply Chain Impact: TSMC’s Lead Widens
Apple committing its highest-volume chip order to TSMC’s N2 process is a significant vote of confidence for the foundry, and it reinforces a supply chain dynamic that’s been building for several years: Apple gets first access to TSMC’s newest nodes, which in turn helps TSMC justify the enormous capital expenditure required to bring 2nm capacity online at volume. That dynamic has knock-on effects across the rest of the chip industry, since every other TSMC customer, from AMD to Nvidia to Qualcomm, effectively waits in line behind Apple’s initial allocation before broader N2 capacity opens up.
Tech Insider has separately reported on Apple’s broader component supply strategy this year, including its multi-year NAND memory agreement, and the A20 Pro’s 2nm commitment fits the same pattern: Apple is increasingly locking in capacity and pricing years ahead of launch to insulate its highest-margin product line from the kind of component shortages that have periodically disrupted other manufacturers, including the CPU and memory pricing pressure Tech Insider has tracked in the PC and GPU market this year around chips like Intel’s upcoming Nova Lake generation.
What This Means for iPhone 18 Pro Buyers
For someone deciding whether to upgrade, the practical question is less “how big are these numbers” and more “will I feel the difference.” The honest answer is that most of the A20 Pro’s gains show up in sustained, demanding tasks rather than everyday app switching, where recent-generation iPhones have felt roughly similar for a couple of years now. Gamers running graphically intensive titles for extended sessions, video editors exporting 4K footage on the go, and anyone running larger on-device AI models locally are the users most likely to notice the combination of a 7-core GPU, a wider memory bus, and a doubled Neural Engine working together.
Tech Insider covered the pricing and broader hardware reveal from Apple’s September 9 keynote in a separate report on the iPhone 18 Pro launch event, which is the better read if the starting price and camera and display changes are the deciding factors. This piece is specifically about what changed under the hood, and on that front, the A20 Pro represents one of the more structurally significant chip generations Apple has shipped since the original 3-nanometer transition in 2023.
Predictions: Where Apple Silicon Goes Next
- Expect Apple to carry the same 2nm process and dual-block Neural Engine architecture into the next generation of M-series Mac and iPad chips within the following year, following its usual pattern of bringing iPhone silicon innovations to its other product lines roughly a generation later.
- Third-party benchmarks will likely land somewhere below Apple’s 40% GPU claim in real-world testing, consistent with how Apple’s own performance percentages have historically compared to independent lab results in past launches.
- Samsung Foundry and Qualcomm will both push harder on their own 2nm-class roadmaps in response, likely compressing the node-lead window Apple currently holds to well under a year by the 2027 flagship cycle.
- Memory bandwidth and packaging innovation, rather than core counts alone, will become the more prominent marketing angle for smartphone chipmakers over the next two to three product cycles, following the template Apple set with the A20 Pro.
- On-device AI performance claims tied to Neural Engine core counts will keep climbing across the industry, but real-world differentiation will increasingly depend on software optimization and memory bandwidth rather than raw NPU core totals.
The Bigger Picture: Node Races Are Becoming the Real Chip War
Step back from the individual spec numbers and the A20 Pro tells a broader story about where competitive advantage in mobile silicon actually comes from in 2026. A decade ago, chipmakers competed primarily on core counts and clock speeds that consumers could easily compare. Today, the more consequential competition happens at the foundry level, in who can secure the most advanced manufacturing capacity first and design an architecture that actually exploits what that capacity makes possible. Apple’s 2-nanometer commitment, paired with a memory architecture built specifically to keep the new GPU and Neural Engine cores fed with data, is a clearer example of that shift than almost any spec Apple has published in recent years.
Whether that translates into a durable competitive edge depends on how quickly Samsung Foundry and TSMC’s other major customers close the node gap, and how much of Apple’s 40% GPU claim and 50% bandwidth claim survive contact with independent benchmarking once iPhone 18 Pro units reach reviewers broadly. Those numbers will be worth revisiting once third-party labs publish their own test results in the coming weeks.
Frequently Asked Questions
What is the Apple A20 Pro chip?
The A20 Pro is Apple’s new system-on-chip powering the iPhone 18 Pro and iPhone 18 Pro Max, announced on September 9, 2026. It features a 7-core GPU, a 32-core Neural Engine split across two 16-core blocks, a 6-core CPU with new Super cores, and a claimed 50% increase in memory bandwidth over the A19 Pro.
Is the A20 Pro really Apple’s first 2-nanometer chip?
Yes, based on reporting from Tom’s Hardware, Wccftech, and Notebookcheck, the A20 Pro is the first Apple iPhone chip manufactured on a 2-nanometer process, widely reported to be TSMC’s N2 node, following three straight years of 3-nanometer chips from the A17 Pro through the A19 Pro.
How much faster is the A20 Pro than the A19 Pro?
Apple claims up to 40% faster GPU performance and roughly 20% faster CPU performance compared with the A19 Pro, according to Wccftech’s coverage of the announcement. Real-world gains will depend on independent benchmark testing once iPhone 18 Pro units are widely available.
What does 50% more memory bandwidth mean in practice?
Apple’s memory bandwidth increase comes from widening the chip’s memory bus from 64-bit to 96-bit. A wider bus moves more data simultaneously between the chip and RAM, which helps sustain GPU and Neural Engine performance during demanding tasks like gaming or on-device AI processing, rather than one component starving for data while others sit idle.
Which devices use the A20 Pro?
The A20 Pro debuts in the iPhone 18 Pro and iPhone 18 Pro Max, announced alongside the rest of Apple’s iPhone 18 lineup at its September 9, 2026 event.
How does the A20 Pro compare to Qualcomm’s Snapdragon chips?
The A20 Pro’s move to a 2-nanometer process gives it a manufacturing lead over most current Snapdragon chips, which remain on 3-nanometer-class nodes as of the A20 Pro’s launch. Qualcomm has its own next-generation chip strategy in motion, including a dual-chip approach for upcoming Snapdragon 8 Elite Gen 6 devices, but a direct benchmark comparison against the A20 Pro isn’t yet available from independent testing.
Why did Apple double the Neural Engine to 32 cores?
Apple built the A20 Pro’s Neural Engine as two parallel 16-core blocks rather than a single larger unit, doubling total AI compute capacity. This supports more demanding on-device AI processing, including running larger AI models locally without sending data to external servers.
When will independent benchmarks of the A20 Pro be available?
Independent lab and reviewer benchmarks typically follow within days to a few weeks of an iPhone’s official launch, once review units and retail devices reach testers. Apple’s own performance percentages, including the 40% GPU and 50% bandwidth claims, are based on internal testing and have not yet been independently verified.
Related Coverage
- Cherry K63W Pro vs Keychron Q1 Ultra: 0.125ms Gap [2026]
- Apple Sept. 9 Event: iPhone 18 Pro Hits $1,199 [2026]
- Xiaomi 18 Fold: Arm Debuts New GPU China-Only [2026]
- Ryzen AI Max PRO 400: 192GB Memory Runs 300B LLMs [2026]
- Rapid Trigger Keyboard Setup: 12 Steps, 90 Min [2026]
