[News] Qualcomm, Apple, MediaTek Battle It Out on TSMC 2nm: Inside Their Flagship AI Strategies
The flagship mobile chip race is entering the 2nm era, with on-device AI emerging as a key battleground. According to ijiwei, Qualcomm, Apple, and MediaTek all unveiled TSMC-made 2nm processors in September 2026, marking the industry’s shift to GAA nanosheet transistors. While all three are putting AI at the center of their latest chips, their approaches differ significantly, from larger on-device models and always-on AI agents to higher memory bandwidth and improved thermal design.
Notably, the simultaneous shift to 2nm is also increasing demand for TSMC’s manufacturing capacity. According to Economic Daily News, Apple, NVIDIA, AMD, Qualcomm, and MediaTek have reportedly increased their reservations for TSMC’s 2nm family by 10% to 20%, prompting the foundry to accelerate its capacity expansion. Industry sources cited by the report expect TSMC’s 2nm monthly capacity to reach around 120,000 wafers by the end of 2026, above previous market estimates of 90,000 to 100,000.
Qualcomm Pushes Larger AI Models On-Device
Qualcomm’s Snapdragon 8 Elite Gen 6 and higher-end Snapdragon 8 Elite Extreme Gen 6 are both built on a 2nm process, with on-device AI at the center of the upgrades. According to Qualcomm, the redesigned Hexagon NPU and Sensing Hub are designed to support more personalized AI agents that can learn and remember user preferences. The Extreme Gen 6 goes further, supporting mixture-of-experts (MoE) models with more than 30 billion parameters directly on-device, Qualcomm notes. The two chips will first appear in flagship devices from brands including Honor, iQOO, Motorola, OnePlus, OPPO, Redmi, RedMagic, vivo, and Xiaomi.
Apple Focuses on AI Computing, Memory and Cooling
Apple is taking a different approach with the 2nm A20 Pro, emphasizing AI compute, memory bandwidth, and thermal performance, ijiwei notes. According to Apple, its new 32-core Dual Neural Engine delivers twice the AI processing power of the A19 Pro, while memory bandwidth increases by 50%. Apple has also introduced new M-series-inspired packaging that places the silicon die and memory side by side, moving memory out of the chip’s thermal path and allowing the A20 Pro to connect directly to a next-generation vapor chamber. Apple says the design improves heat dissipation and sustained performance, an increasingly important consideration for demanding on-device AI workloads. The A20 Pro powers the iPhone 18 Pro lineup as well as the iPhone Duo.
MediaTek Targets Always-On AI With Dual NPUs
MediaTek’s Dimensity 9600 Pro is also built on a 2nm process, but takes a dual-NPU approach to on-device AI. According to MediaTek, its second-generation high-efficiency NPU cuts power consumption by 40% for always-on AI, while the higher-performance NPU 1090 delivers 51% faster LLM prefill performance and 55% more token generation per watt. The Dimensity 9600 Pro supports on-device models with up to 30 billion parameters, with the dual-NPU design aimed at balancing demanding AI workloads with continuously running AI agents.
MediaTek also introduced the Dimensity 9600M for a broader range of high-end smartphones. Unlike the 9600 Pro, however, the 9600M uses TSMC’s 3nm process, according to Reuters.
Read more
- [News] TSMC Reportedly Targets 22% 2nm, 16%+ 3nm Capacity Boost by Mid-2027; CoWoS to Double by 2028
- [News] Qualcomm Reportedly Enlists Samsung, SK hynix as HBC Memory Partners; TSMC Eyes Packaging Role
(Photo credit: Qualcomm)
