Compare AIFind AIAI NewsAI How-To
About Us
PrivacyTermsFAQContactContact
AIB Inc.Company info
© 2026 AIB Inc.

Qualcomm, Apple and MediaTek Unveil 2nm AI Chips

Qualcomm, Apple and MediaTek Unveil 2nm AI Chips

TrendForce·Tuesday, October 6, 2026
  • •Qualcomm, Apple and MediaTek unveiled TSMC-made 2nm flagship mobile chips in September 2026, with distinct on-device AI designs.
  • •Five chipmakers raised TSMC 2nm capacity reservations by 10% to 20%; monthly output is forecast at 120,000 wafers.
  • •Apple cites 2x AI processing power; MediaTek reports 40% lower power use and 51% faster LLM prefill on its Pro chip.
  • •Qualcomm, Apple and MediaTek unveiled TSMC-made 2nm flagship mobile chips in September 2026, with distinct on-device AI designs.
  • •Five chipmakers raised TSMC 2nm capacity reservations by 10% to 20%; monthly output is forecast at 120,000 wafers.
  • •Apple cites 2x AI processing power; MediaTek reports 40% lower power use and 51% faster LLM prefill on its Pro chip.
  • •Qualcomm, Apple and MediaTek unveiled TSMC-made 2nm flagship mobile chips in September 2026, with distinct on-device AI designs.
  • •Five chipmakers raised TSMC 2nm capacity reservations by 10% to 20%; monthly output is forecast at 120,000 wafers.
  • •Apple cites 2x AI processing power; MediaTek reports 40% lower power use and 51% faster LLM prefill on its Pro chip.
  • •Qualcomm, Apple and MediaTek unveiled TSMC-made 2nm flagship mobile chips in September 2026, with distinct on-device AI designs.
  • •Five chipmakers raised TSMC 2nm capacity reservations by 10% to 20%; monthly output is forecast at 120,000 wafers.
  • •Apple cites 2x AI processing power; MediaTek reports 40% lower power use and 51% faster LLM prefill on its Pro chip.

Qualcomm, Apple and MediaTek unveiled TSMC-made 2nm mobile processors in September 2026, as on-device AI became a central focus of their flagship chip strategies. The transition also marks a move to gate-all-around (GAA) nanosheet transistors, in which the gate surrounds the transistor channel. Each company is pursuing different AI priorities, including larger on-device models, continuously running AI agents, greater memory bandwidth and improved cooling.

Apple, NVIDIA, AMD, Qualcomm and MediaTek reportedly raised reservations for TSMC’s 2nm production family by 10% to 20%, according to Economic Daily News, prompting the foundry to speed capacity expansion. Industry sources cited in the report expect monthly 2nm capacity to reach around 120,000 wafers by the end of 2026, above earlier market estimates of 90,000 to 100,000 wafers.

Qualcomm’s Snapdragon 8 Elite Gen 6 and higher-end Snapdragon 8 Elite Extreme Gen 6 use a 2nm process. Qualcomm says their redesigned Hexagon NPU and Sensing Hub support personalized AI agents that learn and remember user preferences. The Extreme model supports mixture-of-experts (MoE) models with more than 30 billion parameters directly on the device. Both chips are set to appear first in flagship phones from Honor, iQOO, Motorola, OnePlus, OPPO, Redmi, RedMagic, vivo and Xiaomi.

Apple’s 2nm A20 Pro emphasizes computing power for AI, memory bandwidth and thermal performance. Its 32-core Dual Neural Engine delivers twice the AI processing power of the A19 Pro, while memory bandwidth rises 50%, Apple says. A new package places the silicon die and memory side by side, moving memory out of the chip’s thermal path and connecting the A20 Pro directly to a next-generation vapor chamber. Apple says this improves heat dissipation and sustained performance. The chip powers the iPhone 18 Pro lineup and iPhone Duo.

MediaTek’s 2nm Dimensity 9600 Pro uses two NPUs: the second-generation high-efficiency unit cuts power consumption 40% for always-on AI, while the NPU 1090 delivers 51% faster LLM prefill and 55% more token generation per watt, the company says. It supports on-device models with up to 30 billion parameters. MediaTek also introduced the Dimensity 9600M for other high-end phones; Reuters reports it uses TSMC’s 3nm process.

Qualcomm, Apple and MediaTek unveiled TSMC-made 2nm mobile processors in September 2026, as on-device AI became a central focus of their flagship chip strategies. The transition also marks a move to gate-all-around (GAA) nanosheet transistors, in which the gate surrounds the transistor channel. Each company is pursuing different AI priorities, including larger on-device models, continuously running AI agents, greater memory bandwidth and improved cooling.

Apple, NVIDIA, AMD, Qualcomm and MediaTek reportedly raised reservations for TSMC’s 2nm production family by 10% to 20%, according to Economic Daily News, prompting the foundry to speed capacity expansion. Industry sources cited in the report expect monthly 2nm capacity to reach around 120,000 wafers by the end of 2026, above earlier market estimates of 90,000 to 100,000 wafers.

Qualcomm’s Snapdragon 8 Elite Gen 6 and higher-end Snapdragon 8 Elite Extreme Gen 6 use a 2nm process. Qualcomm says their redesigned Hexagon NPU and Sensing Hub support personalized AI agents that learn and remember user preferences. The Extreme model supports mixture-of-experts (MoE) models with more than 30 billion parameters directly on the device. Both chips are set to appear first in flagship phones from Honor, iQOO, Motorola, OnePlus, OPPO, Redmi, RedMagic, vivo and Xiaomi.

Apple’s 2nm A20 Pro emphasizes computing power for AI, memory bandwidth and thermal performance. Its 32-core Dual Neural Engine delivers twice the AI processing power of the A19 Pro, while memory bandwidth rises 50%, Apple says. A new package places the silicon die and memory side by side, moving memory out of the chip’s thermal path and connecting the A20 Pro directly to a next-generation vapor chamber. Apple says this improves heat dissipation and sustained performance. The chip powers the iPhone 18 Pro lineup and iPhone Duo.

MediaTek’s 2nm Dimensity 9600 Pro uses two NPUs: the second-generation high-efficiency unit cuts power consumption 40% for always-on AI, while the NPU 1090 delivers 51% faster LLM prefill and 55% more token generation per watt, the company says. It supports on-device models with up to 30 billion parameters. MediaTek also introduced the Dimensity 9600M for other high-end phones; Reuters reports it uses TSMC’s 3nm process.

Read original (English)·Oct 6, 2026
Infra#qualcomm#apple#mediatek#tsmc#2nm process#snapdragon 8 elite gen 6#a20 pro#dimensity 9600 pro#on device ai