MediaTek Dimensity 9600 Pro has launched as the company’s first mobile processor built on TSMC’s 2-nanometre process, giving the Taiwanese chip designer a new flagship platform for phones that run more generative-AI work locally.

 

MediaTek’s September 15 launch introduces three immediate changes:

  • A 2nm flagship chip for premium smartphones
  • A new NPU with a claimed 51% prompt-processing gain
  • A 3nm Dimensity 9600M for a broader flagship tier

 

MediaTek Dimensity 9600 Pro Moves to TSMC 2nm

The manufacturing shift is the clearest distinction in the new lineup. MediaTek says the Dimensity 9600 Pro is its first mobile system-on-chip produced with TSMC’s 2nm technology, while the first smartphones using it are due to launch soon.

 

Smaller process nodes can support denser transistor designs and improve the balance between speed and power consumption. The label does not guarantee a particular battery-life or performance result by itself, however, because those outcomes also depend on the chip architecture, software and each phone maker’s thermal design.

 

MediaTek and TSMC disclosed the technical groundwork a year earlier. In September 2025, MediaTek said it had taped out a flagship system-on-chip using TSMC’s enhanced N2P process, which replaces traditional transistor structures with nanosheet technology.

 

At that stage, the companies said N2P could offer up to 18% more performance at the same power, about 36% lower power at the same speed and 1.2 times the logic density of TSMC’s N3E process. Those are foundry-level comparisons, not independently measured results for the finished 9600 Pro.

 

The New NPU Targets On-Device Generative AI

The 9600 Pro includes a dedicated neural processing unit designed to handle more complex generative-AI workloads on the handset. MediaTek said the NPU improves by 51% over the previous generation when processing a user’s prompt before a model starts generating its response.

 

MediaTek did not disclose the benchmark, model, precision or power conditions behind that percentage. The claim is therefore best read as the company’s measure of one stage of AI inference, rather than evidence that every chatbot, image generator or assistant will run 51% faster.

 

Local inference can reduce the amount of data sent to remote servers and may lower response latency when a model fits within a phone’s memory and compute limits. It can also keep selected features working without a continuous network connection, although developers still decide which tasks remain on-device.

 

Further Reading

 

Dimensity 9600M Extends the Launch to 3nm

MediaTek also introduced the Dimensity 9600M, manufactured on TSMC’s 3nm process. The company positions it for a wider part of the flagship smartphone market, creating a two-chip strategy rather than limiting the launch to its most advanced and likely most expensive silicon.

 

The first phones powered by both processors will arrive soon, according to MediaTek, but the announcement did not identify models, launch markets or pricing. MediaTek has previously supplied leading Chinese handset makers including Xiaomi, Oppo and Vivo, making those partnerships important to watch as the new chips reach commercial devices.

 

The 9600M also gives handset manufacturers another way to manage component costs. Advanced packaging, memory and leading-edge fabrication have become more expensive as AI demand tightens supply, and a 3nm option may help vendors offer new AI features below the highest-priced flagship tier.

 

MediaTek’s Smartphone Push Meets Its Data-Centre Expansion

The launch places MediaTek in a more direct contest with Qualcomm for premium Android phones, a segment where higher selling prices and differentiated AI features can produce stronger margins. MediaTek’s market value passed Qualcomm’s earlier in 2026, adding financial weight to a rivalry long defined by handset design wins.

 

MediaTek executive JC Hsu said the company is working with phone makers to limit the effect of rising component costs on consumers. He also argued that the industry’s shift toward pricier hardware leaves room for MediaTek to gain share in flagship devices.

 

The mobile processors arrive while MediaTek is expanding beyond consumer chips. Its first AI accelerator for a major U.S. cloud provider is expected to enter mass production in the fourth quarter of 2026, linking the company’s edge-AI ambitions with a growing custom data-centre business.

 

That expansion is backed by fresh capital and strategic partners. MediaTek raised $3.9 billion through a convertible-bond sale in August, with Nvidia investing $3.5 billion and Alphabet also participating. The new phone chips are a separate product event, but they show how broadly MediaTek now wants to compete across AI computing.

 

The next evidence will come from shipping phones rather than launch claims. Device reviews can test sustained performance, power use, thermal limits and real application speed, while handset announcements will show whether the 2nm Pro model wins designs beyond MediaTek’s established customer base.