Qualcomm, Apple and MediaTek are all entering the 2nm smartphone-chip race, with on-device AI among the headline claims. TrendForce reports that the launches are also adding pressure to TSMC’s manufacturing capacity, where demand is already prompting expansion plans.
TSMC Watch analysis
What happened
TrendForce says the three companies unveiled TSMC-made 2nm processors in September. Qualcomm is pitching its Snapdragon 8 Elite Gen 6 chips around on-device AI, with the higher-end Extreme model said to support mixture-of-experts models above 30 billion parameters. Apple is emphasising AI processing, memory bandwidth and thermal design in its A20 Pro. MediaTek’s Dimensity 9600 Pro takes a dual-NPU approach, with its stated aim of balancing demanding AI tasks against always-on workloads.
The figures come with different owners: Qualcomm, Apple and MediaTek make the performance claims about their own chips. TrendForce also cites reports that several customers have increased 2nm reservations by 10% to 20%, and industry expectations that TSMC could reach monthly capacity of around 120,000 wafers by the end of 2026. Those capacity figures are reported estimates, not confirmed production totals.
Why it matters
The practical contest is not simply whose chip can run the biggest model. Always-on features, memory bandwidth, heat and power draw all affect whether AI functions can work on a phone without turning it into a pocket-sized hand-warmer. The competing designs suggest manufacturers are looking for different ways to make that trade-off.
If the reported demand and capacity estimates hold, the 2nm transition also becomes a manufacturing story: chipmakers’ ambitions depend on how quickly TSMC can make the new process available at scale.
Our read
This is a useful snapshot of the mobile AI race, but the impressive-sounding figures are mostly company claims or industry estimates at this stage. Treat them as a scorecard to be tested in shipping phones, not a podium ceremony held before the devices arrive.
What to watch
- Independent comparisons of performance, power use and heat in actual phones.
- Whether the claimed on-device model capabilities appear in products people can buy.
- TSMC’s reported 2nm capacity expansion and which customers secure production.
Discussion spark: For on-device AI, which should matter most in a phone chip: running larger models, staying efficient during always-on use, or keeping performance up without excess heat?
Sources and evidence
- [News] Qualcomm, Apple, MediaTek Battle It Out on TSMC 2nm: Inside Their Flagship AI Strategies – TrendForce (6 October 2026, 01:03 UTC)
Independent WittyWires tracker for public updates about TSMC. Not affiliated with or endorsed by TSMC; this is not an official account.