Apple has made the jump to 2 nanometers with the A20 Pro, the processor powering the new iPhone 18 Pro, iPhone 18 Pro Max, and the foldable iPhone Duo. The chip once again puts TSMC behind one of the most important pieces in Apple’s lineup, and it marks the arrival of TSMC’s most advanced manufacturing technology in the smartphone market. Beyond the new iPhone, the launch shows just how much the relationship between the two companies is shifting because of the growth of AI chips.
The key facts about Apple’s A20 Pro and TSMC, in 20 seconds
- The A20 Pro uses TSMC’s new 2nm manufacturing generation.
- It packs a 6-core CPU, a 7-core GPU, and two 16-core Neural Engines.
- Apple says it delivers 50% more memory bandwidth than the A19 Pro.
- TSMC still relies on major customers, but AI is reshaping how its business breaks down.
- NVIDIA is gaining weight as demand for data center accelerators grows.
Apple unveiled the A20 Pro on September 9 alongside the iPhone 18 Pro. The company doesn’t explicitly name TSMC as the manufacturer in its announcement, which is typical of its product presentations, but the manufacturing relationship between the two companies and the A20 Pro’s jump to the 2nm generation put the Taiwanese foundry at the center of the new generation.
The shift from the 3nm process used previously shouldn’t be read as a literal one-nanometer reduction in physical size. The commercial names given to modern nodes no longer directly describe the size of any specific transistor. They represent process generations with changes in density, performance, power consumption, and architecture.
In TSMC’s case, N2 introduces nanosheet transistors, replacing the FinFET architecture used in previous generations.
The A20 Pro Debuts an Architecture Built for More AI in the iPhone
Apple used the process change as an opportunity to redesign several parts of the processor.
The A20 Pro packs a six-core CPU, made up of two high-performance cores and four efficiency cores, along with a seven-core GPU, matching earlier leaks that had already pointed to a seven-core GPU and a wider memory bus.
Artificial intelligence gets special attention.
Apple has included two 16-core Neural Engines, 32 cores in total, plus Neural Accelerators built into the CPU. According to the company’s figures, this configuration doubles the Neural Engine’s AI processing capacity compared with the A19 Pro.
Apple also says the new processor has 50% more memory bandwidth than its predecessor and that the GPU can be up to 40% faster. These are comparisons provided by Apple and can vary depending on the workload used.
| Feature | Apple A20 Pro |
|---|---|
| Manufacturing process | 2nm |
| CPU | 6 cores |
| GPU | 7 cores |
| Neural Engine | 2 × 16 cores |
| Memory bandwidth | +50% vs. the A19 Pro, per Apple |
| GPU speed | Up to +40% vs. the A19 Pro, per Apple |
| Products | iPhone 18 Pro, 18 Pro Max, and iPhone Duo |
The processor also shows up in the iPhone Duo, Apple’s first foldable phone. The company uses the same A20 Pro along with a thermal solution adapted to the new form factor.
There’s also a less visible change than the jump to 2nm: packaging.
Apple explains that the new design, inspired by its M-series processors, places the silicon next to the memory and moves the memory off the main thermal path. That lets the processor connect directly to a new vapor chamber designed to sustain performance during extended workloads.
This matters because in mobile devices, peak performance doesn’t count for much if the processor has to quickly throttle its clock speed to stay within its thermal limits.
Apple Is Still Huge for TSMC, but AI Is Changing the Split
For years, the relationship between Apple and TSMC has been relatively simple: Apple designs its processors, and TSMC manufactures them.
That model let Apple progressively move away from third-party components in areas like its computer processors, and at the same time turned the Cupertino company into the Taiwanese foundry’s single biggest customer.
TSMC’s official reports now reveal a shift.
TSMC doesn’t name its biggest customers in its annual report, but it does disclose the percentage of revenue each one represents.
In 2023, its largest customer generated 25% of revenue and the second-largest 11%. A year later, those figures were 22% and 12%, respectively.
In 2025, the gap narrowed dramatically: 19% for the top customer and 17% for the second.
| Year | TSMC’s largest customer | Second-largest customer |
|---|---|---|
| 2023 | 25% | 11% |
| 2024 | 22% | 12% |
| 2025 | 19% | 17% |
TSMC doesn’t provide the names in this document. The industry has traditionally associated the first with Apple and the second with NVIDIA, but that identification should be treated as an outside attribution rather than data the foundry itself has officially published.
What really matters here is the trend.
The gap between the two went from 14 percentage points in 2023 to just two in 2025.
Artificial intelligence explains a good part of the structural shift TSMC is going through. Large data center accelerators require advanced nodes and are also consuming a growing share of packaging capacity like CoWoS, capacity that TSMC is reportedly looking to double by 2028 to keep up with AI chip demand.
TSMC acknowledges in its report that customer concentration is partly tied to the industry’s shift toward high-performance computing (HPC) applications and smartphones.
The A20 Pro Also Shows Why Apple Remains Strategic for TSMC
NVIDIA’s growth, and that of AI accelerators generally, doesn’t turn Apple into a secondary customer.
The company still moves enormous volumes of processors and, above all, is usually among the first to adopt TSMC’s most advanced manufacturing generations.
That has considerable industrial value.
A new technology needs to progressively ramp up its manufacturing volumes and yields. A customer capable of selling tens of millions of devices provides a very different kind of demand than specialized server processors do.
Apple needs TSMC too.
Designing a processor like the A20 Pro is only part of the challenge. Manufacturing it at scale with one of the most advanced processes available requires facilities that cost tens of billions of dollars.
This specialization lets Apple invest in chip architecture without becoming a semiconductor manufacturer itself.
TSMC, meanwhile, can spread its enormous industrial investments across smartphone customers, computer makers, AI accelerator designers, data center processor makers, and custom chip customers.
That diversification is gaining importance precisely because the market is changing.
A few years ago, the annual iPhone launch could set much of the calendar for TSMC’s most advanced nodes. Now those same processes have to coexist with growing demand from NVIDIA, AMD, and the custom processors developed by major cloud companies.
The A20 Pro confirms that Apple still has access to TSMC’s most advanced technology, but it also arrives at a moment when the foundry no longer depends on smartphones the way it used to.
The stock-market comparison some analysts draw between Apple and TSMC belongs to a different conversation. Growth forecasts, valuation multiples, and buy recommendations shift with share prices and depend on assumptions about future earnings. The A20 Pro launch alone doesn’t tell us which of the two companies is the better investment.
What it does confirm is something more tangible: TSMC’s 2nm process has now reached a commercial product with enormous potential volume, while the same company simultaneously has to expand its capacity to keep up with the AI infrastructure race.
Apple and NVIDIA may compete for capacity and importance within TSMC, but they represent two distinct markets the foundry needs to serve. One puts its chips in hundreds of millions of personal devices; the other is helping build data centers whose consumption of advanced silicon and packaging keeps climbing.
The A20 Pro is the latest proof that, even amid the AI boom in data centers, the smartphone remains one of the main showcases for the most advanced semiconductor manufacturing technologies.
