Mark RadarMARK RADAR
About
EN
Sign in

TSMC Adopts NVIDIA AI and Accelerated Computing to Transform Chipmaking

1 reports · First detected 2026-06-01 · Last active 2026-06-01

TSMC is the world's leading chip foundry, and its advanced processes involve highly complex work spanning computational lithography, transistor design and process control. Its adoption of NVIDIA's CUDA-X accelerated computing services and AI technology could help shorten development cycles for next-generation AI chips, improve production yields and reduce the energy consumed by computation-intensive manufacturing processes.

TSMC said it would deeply integrate NVIDIA CUDA-X and AI systems into its core semiconductor operations, including computational lithography, transistor simulation and process control. The initiative is intended to accelerate the development and production of next-generation chips while improving yields and energy efficiency. Available reports did not disclose the announcement date, investment amount, mass-production schedule or expected scale of the improvements.

All Coverage

1 original reports

The Backstory

The history behind this event
After this
Nvidia's Next-Generation Rosa CPU Reportedly to Use TSMC's Advanced 2nm Processfirst seen 2026-07-09 · 1 reports · similarity 0.76 · same topic: NVIDIA

The rise of agentic AI is driving increasingly stringent demands for AI inference efficiency and single-core performance. Nvidia, the world's leading graphics-chip company, is seeking to maintain its lead in AI computing through continued close collaboration with TSMC, the world's largest contract chipmaker. Process-node advances can significantly reduce power consumption and increase computing speeds, directly affecting the performance of next-generation data centers and AI servers.

Supply-chain sources said in July 2026 that Nvidia plans to manufacture its next-generation Rosa CPU using TSMC's advanced A16 process, part of the company's 2nm family. The chip will feature Arm v9.2 cores and is designed to improve single-core performance and instruction-delivery efficiency for agentic AI inference workloads. If confirmed, Nvidia would join Apple and other customers as a major AI client for TSMC's 2nm process family.

Mark Radar|MARK RADAR

If you search news on Google, you can set Mark Radar as a preferred source—our coverage will show up more often in your results. Set as preferred source on Google →

All times are in Taipei time (GMT+8)