Rapidus 2nm is Japan’s flagship next-generation logic process node, under development by Rapidus Corporation in collaboration with IBM and imec, targeting high-performance, energy-efficient, leading-edge semiconductor manufacturing. The goal is to achieve mass production readiness by 2027, marking Japan’s re-entry into the forefront of advanced logic manufacturing after decades of decline.
Built upon Gate-All-Around (GAA) nanosheet transistor architecture, the 2nm node represents a leap beyond FinFET scaling limits and is designed for applications in AI, HPC, automotive, and 6G communications.
Key Features
Feature | Description |
---|---|
Transistor Type | GAA nanosheet (also known as Nanosheet FETs) |
Gate Length | Sub-20nm, multi-stack sheets |
Node Scaling | True 2nm-class dimensions based on metal pitch and contacted gate pitch |
EUV Lithography | Uses extreme ultraviolet (EUV) lithography throughout the BEOL and FEOL |
High Performance | Expected >45% performance gain compared to 7nm FinFET |
Energy Efficiency | Estimated >75% power reduction at equivalent performance |
Density | >2x transistor density improvement over 7nm |
Backside Power Delivery (Planned) | Pathfinding for future node extensions, not first-gen |
Foundational Technology
Based on IBM 2nm Research Platform
-
Developed originally by IBM Research Albany (announced in 2021).
-
First public demonstration of stacked nanosheets with 12nm gate length.
-
Rapidus licensed IBM’s device architecture, process flow, and DTCO methodology.
imec Collaboration
-
Supports process integration, mask development, and material modeling through imec’s platform in Leuven, Belgium.
Timeline
Year | Milestone |
---|---|
2021 | IBM announces 2nm GAA prototype |
2022 | Rapidus founded and licenses IBM 2nm tech |
2023 | Construction of Chitose R&D fab begins |
2024 | EUV tools ordered, pathfinding with imec begins |
2025 | Pilot line operational in Hokkaido |
2026 | Process qualification |
2027 | Volume production of 2nm logic devices begins |
Technical Architecture
Gate-All-Around (GAA) Nanosheet
-
Multi-Bridge Channel FET structure: several nanosheets stacked vertically and fully surrounded by a gate.
-
Offers improved electrostatic control, lower leakage current, and greater scalability than FinFET.
-
Sheet width modulation enables tuning of performance and leakage tradeoffs.
Interconnect
-
EUV patterning with 1D gridded interconnects for tighter pitch.
-
Advanced middle-of-line process integration for signal routing.
Device Design Targets
Metric | Target |
---|---|
Gate pitch | ~42nm |
Metal pitch | ~28nm |
Effective transistor density | >300 million/mm² |
Supply voltage | ~0.6–0.75V |
Frequency scaling | >4GHz design feasibility |
Applications
The Rapidus 2nm node is being designed for:
-
AI accelerators
-
High-Performance CPUs & GPUs
-
Automotive SoCs (ASIL-D target)
-
5G/6G and edge compute
-
Low-power mobile/embedded systems
Competitive Landscape
Company | Node | Target Year |
---|---|---|
TSMC | N2 | 2025–2026 (risk prod), 2026–2027 (HVM) |
Intel | Intel 20A / 18A | 2024 (pre-2nm, ribbonFET) |
Samsung | SF2 / SF1.4 | 2025+ |
Rapidus | 2nm | 2027 |
Although Rapidus trails TSMC and Samsung by 1–2 years in ramp, it is targeting a “leapfrog” approach through ecosystem co-development and tight design-technology co-optimization (DTCO) with global partners.
Challenges
Area | Details |
---|---|
Workforce | Recruiting and training engineers for GAA and EUV is ongoing |
Ecosystem | Japan lacks native EDA, IP, and 2nm tapeout flow vendors |
Yield Ramp | High defect sensitivity due to small feature sizes |
Capital Investment | 2nm fabs cost $15–20B; needs sustained government support |
Government Support
-
Supported by METI (Ministry of Economy, Trade and Industry)
-
Expected $5–10 billion in subsidies over multiple years
-
Aligned with Japan’s economic security and semiconductor sovereignty initiatives
Summary
The Rapidus 2nm process represents a bold national effort by Japan to re-enter the top tier of semiconductor manufacturing. With IBM’s GAA nanosheet platform, imec collaboration, and Japanese industrial strength, Rapidus seeks to become a trusted, high-performance foundry for advanced AI and compute semiconductors in the post-FinFET era.
Also Read:
Share this post via:
TSMC N3 Process Technology Wiki