Apple used its September 9 keynote to confirm the A20 Pro, the chip that will run the new iPhone 18 Pro lineup, and the numbers Apple put on stage are the biggest jump the company has claimed in several generations. The headline figures: a 7-core GPU delivering up to 40% faster graphics, a dual 16-core Neural Engine totaling 32 cores, and a 50% increase in memory bandwidth over the A19 Pro. Apple built the chip on TSMC’s 2-nanometer (N2) process, making the A20 Pro the first Apple silicon shipped at that node.
Outlets including Wccftech, Tom’s Hardware, MacRumors, 9to5Mac, AppleInsider, Gadgets 360, and Beebom all published coverage within hours of the announcement, a sign of how closely the smartphone chip market is watching Apple’s move to 2nm ahead of its rivals. This piece breaks down what Apple confirmed, what’s still leak-only, how the A20 Pro stacks up against its own predecessor, and what the jump to 2nm means for the wider chip supply chain heading into 2027.
What Apple actually announced about the A20 Pro
The A20 Pro is a system-on-chip built for the iPhone 18 Pro and iPhone 18 Pro Max, according to the announcement. Apple’s pitch centers on four numbers: a 6-core CPU, a 7-core GPU, a 32-core Neural Engine split across two 16-core blocks, and a memory subsystem Apple describes as the widest memory interface ever shipped in an iPhone. Each of those figures represents a real architectural change from the A19 Pro, not just a clock-speed bump.
That’s a notable shift in tone for Apple’s chip marketing. In recent years the company leaned on single-digit CPU gains and let the GPU and Neural Engine carry the story. This time, memory bandwidth gets equal billing, which points to how much AI workloads running locally on the phone (image generation, on-device model inference, camera computational photography) depend on getting data to the processor fast enough to keep the compute cores fed.
Built on TSMC’s 2-nanometer process
The A20 Pro is manufactured on TSMC’s N2 (2nm) process node, the first time Apple has shipped a phone chip at this density. Apple has historically been TSMC’s lead customer for each new node, and the A20 Pro extends that pattern by roughly a year ahead of most Android flagship silicon, which remains on 3nm or 4nm nodes going into 2027. A smaller node generally means more transistors packed into the same die area, translating into either higher performance at the same power draw or the same performance at lower power draw, and Apple appears to be splitting the difference here rather than maximizing either extreme.
Six CPU cores, two of them built for peak performance
The CPU configuration stays at 6 cores, split between 2 high-performance cores and 4 efficiency cores, the same core count Apple has used since the A17 Pro. What changes generation over generation is the design of those cores, not the count, and Apple’s own framing for the A20 Pro emphasizes efficiency gains alongside raw speed rather than simply adding cores. That’s consistent with how Apple has approached CPU design for the past several A-series cycles: hold the core count steady, push the design further inside the die shrink.
The 7-core GPU and Apple’s 40% graphics claim
Apple’s GPU jumped from 6 cores on the A19 Pro to 7 cores on the A20 Pro, and the company says that translates into up to 40% faster graphics performance compared with the previous generation. A single extra GPU core sounds modest on paper, but combined with the 2nm process and the wider memory path, it lines up with the kind of graphics uplift Apple has delivered in stronger generational cycles.
The practical effect shows up in two places: sustained frame rates in graphically demanding mobile games, and the rendering pipeline for computational photography, where the GPU handles portrait-mode depth mapping, Photonic Engine processing, and the growing list of on-device video effects Apple has added to the Camera app over the past two years. A faster GPU with more bandwidth to draw from also matters for how long a phone can sustain peak graphics performance before thermal throttling kicks in, which is where the 2nm efficiency gains do quiet, unglamorous work.
Neural Engine doubles to 32 cores
The most eye-catching architectural change is the Neural Engine, which Apple built as two 16-core blocks for a combined 32 cores, doubling the AI-dedicated silicon compared with the previous generation. Apple and outlets covering the launch describe this as roughly doubling the chip’s AI throughput year over year, which is a larger jump than the CPU or GPU gains.
That’s not a coincidence. Apple has spent the past two years building out on-device Apple Intelligence features, and every one of those features (writing tools, image generation, Genmoji, on-device Siri processing) runs through the Neural Engine rather than the GPU or CPU. Doubling that block gives Apple headroom to run larger local models without offloading to the cloud, and it puts pressure on how Apple frames its AI story relative to Google’s Tensor chips and Qualcomm’s Hexagon NPUs, both of which have also grown their dedicated AI silicon generation over generation.
Memory bandwidth jumps 50%: what that actually fixes
Apple says the A20 Pro delivers a 50% increase in memory bandwidth over the A19 Pro, calling it the widest memory interface Apple has shipped in an iPhone. Bandwidth, not raw memory capacity, has quietly become one of the biggest bottlenecks for on-device AI: a chip can have plenty of compute cores, but if data can’t move from memory to those cores fast enough, the extra cores sit idle waiting on data.
Some of the more granular detail behind that 50% figure is still leak-territory rather than Apple-confirmed. Reports following the announcement describe the memory interface moving from a 64-bit bus on the A19 Pro to a 96-bit bus on the A20 Pro, and separate reports point to 12GB of LPDDR5X memory in the iPhone 18 Pro models. Neither of those two specifics has been confirmed directly by Apple’s own materials, so treat the bus width and RAM capacity as reported-but-unconfirmed until Apple or a teardown outlet like GSMArena verifies them against retail hardware.
A20 Pro vs A19 Pro: the generational comparison
Here’s how the confirmed specs stack up against last year’s chip. Figures marked “unconfirmed” come from reports and leaks rather than Apple’s own announcement and should be read with that caveat.
| Spec | A19 Pro (2025) | A20 Pro (2026) |
|---|---|---|
| Process node | 3nm (TSMC N3-family) | 2nm (TSMC N2) |
| CPU configuration | 6-core (2 performance + 4 efficiency) | 6-core (2 performance + 4 efficiency) |
| GPU cores | 6-core GPU | 7-core GPU |
| GPU performance claim | Baseline | Up to 40% faster graphics |
| Neural Engine | 16-core | 32-core (dual 16-core blocks) |
| AI throughput claim | Baseline | Roughly double |
| Memory bandwidth | Baseline | +50% vs A19 Pro |
| Memory bus width | 64-bit (reported) | 96-bit (reported, unconfirmed by Apple) |
| RAM | LPDDR5X | LPDDR5X, 12GB reported (unconfirmed) |
| Device | iPhone 17 Pro / Pro Max | iPhone 18 Pro / Pro Max |
How the A20 Pro fits Apple’s longer chip history
Zooming out, the A20 Pro is the fourth straight generation to hold the CPU at 6 cores while pushing gains through process shrinks and per-core design work rather than adding cores. The A17 Pro launched on TSMC’s first 3nm node in 2023, the A18 Pro moved to a refined 3nm process in 2024, the A19 Pro pushed that same process family further in 2025, and now the A20 Pro jumps straight to 2nm in 2026. That’s a faster node cadence than Apple used through most of the 2020s, when it stretched a single process generation across two or three chip cycles.
| Chip | Year | Process node | Neural Engine |
|---|---|---|---|
| A17 Pro | 2023 | 3nm (N3B) | 16-core |
| A18 Pro | 2024 | 3nm (N3E) | 16-core |
| A19 Pro | 2025 | 3nm (N3P) | 16-core |
| A20 Pro | 2026 | 2nm (N2) | 32-core |
The jump from a single 16-core Neural Engine to a dual 16-core, 32-core design breaks that steady pattern, and it’s the clearest signal in the whole announcement that Apple is designing this generation of silicon around on-device AI first, with CPU and GPU gains arriving as a byproduct of the smaller process node rather than as the headline.
Competitive picture: Apple’s 2nm head start
Apple reaching 2nm before its Android-side competitors matters because of how TSMC allocates early capacity on a brand-new node. Apple has been TSMC’s first customer on every major node transition going back to the original A11’s move to 10nm, and it typically buys up a large share of the earliest, most expensive wafers before yields improve and costs drop for smaller customers. That head start means Qualcomm’s and MediaTek’s next flagship mobile chips are widely expected to remain on 3nm-class processes into 2027, giving Apple roughly a full generation of process-node advantage on paper, even if real-world performance differences end up smaller once software and thermal design are factored in.
That advantage is not guaranteed to show up evenly across every workload. Android flagships built around Snapdragon and Dimensity silicon have closed much of the sustained-performance gap with Apple in recent generations by leaning on larger vapor chambers and more aggressive power management, so a smaller node number for Apple doesn’t automatically translate into a proportionally faster phone in everyday use. What it does guarantee is a cost and efficiency advantage for Apple at the chip level, which shows up over time in battery life claims and in how much thermal headroom Apple has to work with when adding new camera or AI features.
Market impact: what a 2nm iPhone chip means for the supply chain
Being first on a new TSMC node is expensive. Early 2nm wafers cost significantly more per unit than mature 3nm wafers, and that cost pressure typically flows through Apple’s component budget before it ever reaches a retail price tag. Component cost increases at the chip level are one of the reasons industry watchers pay close attention to Apple’s supplier relationships every time a node transition happens, since TSMC, memory suppliers, and camera-module makers all adjust pricing and allocation around Apple’s volume commitments.
The memory side of the equation compounds that pressure. LPDDR5X supply has already been tight across the industry in 2026 as AI PC and AI phone makers compete for the same memory fabs that also supply data-center customers. A wider, higher-bandwidth memory interface on the A20 Pro means Apple needs memory that can sustain higher throughput, which typically comes at a premium versus standard-bin parts, adding another cost variable on top of the 2nm wafer premium itself.
What’s confirmed versus what’s still a leak
It’s worth separating what Apple has actually confirmed from what’s circulating in supply-chain reports, since the two get blended together quickly once a chip launch story spreads across dozens of outlets.
- Confirmed by the announcement: 2nm (TSMC N2) process, 6-core CPU, 7-core GPU with up to 40% faster graphics, 32-core Neural Engine (dual 16-core), and a 50% increase in memory bandwidth versus the A19 Pro.
- Reported but not confirmed by Apple: the exact memory bus width (reports point to 96-bit, up from 64-bit), and RAM capacity (reports point to 12GB of LPDDR5X).
- Not yet public: individual core clock speeds, cache sizes, and exact TOPS figures for the Neural Engine’s AI throughput.
That distinction matters for anyone deciding whether to upgrade based on spec sheets alone. The confirmed numbers already represent a meaningful generational jump, particularly on the Neural Engine and memory side, and buyers don’t need the unconfirmed details to know the iPhone 18 Pro is a bigger architectural change than the iPhone 17 Pro was over its predecessor.
Why this matters for iPhone 18 Pro buyers
For most people, the A20 Pro’s real-world impact shows up in three places: how long the phone lasts on a charge under heavy use, how well it handles on-device Apple Intelligence features without hitting the cloud, and how long the phone stays fast years into ownership. The 2nm process should help with the first, the doubled Neural Engine should help with the second, and Apple’s historical pattern of over-provisioning Neural Engine capacity ahead of software needs should help with the third, since past A-series chips with generous AI silicon (like the A17 Pro) have aged better against new software features than chips Apple designed with tighter margins.
Gamers and creators get the most direct benefit from the GPU and memory bandwidth gains specifically, since those are the two subsystems most directly tied to frame rates and video export times. Everyday users who mostly browse, message, and take photos will notice the difference less in raw speed and more in battery life and in how smoothly newer Apple Intelligence features run without lag or cloud round-trips.
Predictions: where the A20 Pro story goes next
A few things look likely to play out over the next 12 to 18 months based on how Apple has handled previous node transitions and how competitors have historically responded.
- Expect the standard A20 (non-Pro) chip, likely arriving in the iPhone 18 and iPhone 18e, to inherit a trimmed-down version of the same 2nm architecture with fewer GPU and Neural Engine cores, following the pattern set by every prior A-series generation.
- Qualcomm and MediaTek are likely to respond with their own 2nm designs, but probably not until their late-2027 flagship chips, since TSMC’s early 2nm capacity is heavily allocated to Apple through most of 2026 and 2027.
- Watch for independent teardown and benchmark verification of the reported 96-bit memory bus and 12GB RAM figures once retail iPhone 18 Pro units reach reviewers, which should settle whether those leak-sourced specs hold up.
- Apple’s Mac and iPad lines are likely to see a 2nm M-series chip within the next generation or two, following the same pattern where iPhone silicon leads and Mac silicon follows roughly a year later on any new process node.
- Expect continued upward pressure on component costs across the smartphone industry as more manufacturers compete for early access to 2nm capacity at TSMC, which could show up in flagship Android pricing before it shows up in iPhone pricing given Apple’s typically fixed retail price points.
The bigger picture: Apple’s AI strategy runs through silicon
The A20 Pro announcement reads less like a routine annual refresh and more like Apple building the hardware foundation for a multi-year AI push. Doubling Neural Engine core count in a single generation is a bigger jump than Apple has made to that block in years, and pairing it with a 50% memory bandwidth increase suggests Apple’s engineering priorities have shifted to match what its software teams need to ship increasingly capable on-device Apple Intelligence features without leaning on cloud compute for every request.
That strategy carries risk as well as upside. Analysts at firms like Counterpoint Research have tracked how on-device AI silicon investment across the smartphone industry has grown faster than consumer willingness to pay a premium for AI features specifically, meaning Apple is betting that better hardware will make Apple Intelligence noticeably faster and more capable rather than betting that AI alone will drive iPhone 18 Pro upgrades. The A20 Pro gives Apple the hardware headroom to make that bet. Whether Apple Intelligence’s software roadmap catches up to the silicon is the open question heading into 2027.
Frequently asked questions
What is the Apple A20 Pro?
The A20 Pro is Apple’s newest system-on-chip, built for the iPhone 18 Pro and iPhone 18 Pro Max and announced alongside the phones on September 9, 2026. It’s the first Apple chip built on TSMC’s 2-nanometer (N2) process.
How many GPU cores does the A20 Pro have?
The A20 Pro has a 7-core GPU, up from the 6-core GPU in the A19 Pro. Apple says that delivers up to 40% faster graphics performance than the previous generation.
How much bigger is the A20 Pro’s Neural Engine?
The A20 Pro packs a 32-core Neural Engine, built as two 16-core blocks, doubling the core count of the A19 Pro’s 16-core Neural Engine.
Is the A20 Pro’s 96-bit memory bus confirmed by Apple?
No. Apple has confirmed a 50% increase in memory bandwidth over the A19 Pro, but the specific claim that the bus width grew from 64-bit to 96-bit comes from supply-chain reports and leaks, not from Apple’s own announcement.
Does the A20 Pro come with 12GB of RAM?
That figure is reported by multiple outlets covering the launch but has not been confirmed directly by Apple. Treat 12GB of LPDDR5X as an unconfirmed spec until independent teardowns verify it.
How does the A20 Pro compare to the A19 Pro?
The A20 Pro keeps the same 6-core CPU configuration as the A19 Pro but moves to a 2nm process, adds a seventh GPU core for up to 40% faster graphics, doubles the Neural Engine to 32 cores, and increases memory bandwidth by 50%.
Which process node is the A20 Pro built on?
The A20 Pro is built on TSMC’s N2 process, a 2-nanometer node. It’s the first Apple silicon chip to ship on this process.
Will other iPhone 18 models get the A20 Pro?
The A20 Pro is confirmed for the iPhone 18 Pro and iPhone 18 Pro Max. Based on Apple’s usual naming and rollout pattern, a standard A20 chip with fewer GPU and Neural Engine cores is expected for other iPhone 18 models, though Apple has not detailed that chip separately.
