Book a demo

Cut patent&paper research from weeks to hours with PatSnap Eureka AI!

Try now

Buried Power Rail IR Drop at 2nm — PatSnap Eureka

Buried Power Rail IR Drop at 2nm — PatSnap Eureka
2nm Power Delivery · Patent Intelligence

Buried Power Rail Integration for IR Drop Reduction at 2nm

At sub-3nm standard cell heights, conventional M1/M2 power straps are geometrically unviable. Buried power rail (BPR) technology repositions VDD/VSS conductors below the active device plane — enabling parallel-path shunting, upsized cross-sections, and placement-aware current normalization to eliminate IR drop hotspots.

BPR IR Drop Suppression: 10% single-chip, 59% four-chip stack — Kobe University 2021, 3.9M-gate crypto core at 30 MHz Dynamic IR drop suppression achieved by backside buried metal PDNs with distributed capacitance. Single-chip configuration achieves 10% suppression; four-chip stacked configuration achieves 59% suppression. Source: Kobe University, 2021, PatSnap Eureka literature analysis. 60% 45% 30% 15% 0% 10% Single-chip 59% 4-chip stack Dynamic IR Drop Suppression — Backside Buried Metal PDN
15+
BPR patent families analysed across Intel, IBM, Samsung, ARM, NVIDIA & more
59%
Dynamic IR drop suppression in 4-chip stacked BPR configuration (Kobe University, 2021)
10%
Dynamic IR drop suppression in single-chip backside buried metal PDN
>1.0
Height-to-width aspect ratio for Qualcomm's high-aspect-ratio voltage rails
Physical Architecture

How Buried Power Rails Reduce Effective Rail Resistance

At the heart of BPR technology is the repositioning of VDD and VSS conductors from front-end-of-line (FEOL) metal layers into trenches carved within the device layer itself. As described in Intel Corporation's 2023 patent, the BPR resides within the device layer neighboring the drain structure, with its uppermost surface below the uppermost surface of the drain. A top-side power rail sits vertically above the BPR, connected to it via a conductive structure.

Intel explicitly articulates the resistance-reduction mechanism: "the buried power rail acts as a shunt for the top-side power rail… Since the buried power rail and top-side power rail are effectively infinitely long in a logic block relative to a single transistor or single logic cell, the resistance of the top-side power rail in parallel [with the BPR] is reduced." This parallel-path shunting directly lowers the effective rail resistance, which is the primary driver of static and dynamic IR drop at high current densities.

Samsung Electronics takes a process-centric approach, describing formation of BPRs within power rail trenches in a dielectric layer on a first semiconductor substrate, followed by wafer bonding of a donor wafer carrying the active device layer. This fabrication flow enables BPR widths to be set lithographically during the substrate preparation stage, decoupled from the standard cell layout pitch constraints that govern front-side metal. Learn more about patent landscape analysis for semiconductor processes on PatSnap.

The geometric constraint problem at tight cell-to-cell pitches — a critical concern at 2nm where adjacent standard cells may share boundary edges — is addressed by IBM's dual-BPR stack: a first BPR with a larger critical dimension (CD) contacts a second BPR with a smaller CD. This stacked geometry allows the effective cross-sectional area of the power conductor to exceed what would be possible through a single trench patterned within the tight cell boundary.

M1/M2
Metal layers displaced by BPR, freeing routing resources for signal wires
<5
Fin pitches defining 2nm standard cell height — below which front-side power straps become unviable
2023
Intel BPR structural patent year — defining the buried-rail-plus-top-rail shunting architecture
Dual
IBM's BPR stack layers — larger CD first BPR contacts smaller CD second BPR
  • BPR positioned below active device plane — zero routing-layer consumption
  • Parallel shunting reduces effective rail resistance proportional to combined cross-section
  • Wafer bonding decouples BPR width from cell pitch constraints
  • Dual-BPR stack overcomes tight cell-to-cell spacing at 2nm
Patent Landscape Data

BPR Innovation by Assignee and Technical Focus

Distribution of buried power rail patent families across major semiconductor organizations at sub-3nm nodes, derived from PatSnap Eureka analysis of 15+ directly relevant patent families.

BPR Patent Families by Assignee

ARM, NVIDIA, and Intel lead BPR patent activity at advanced nodes, with ARM holding 3 active US patent families covering bitcell and dual-function buried signal routing.

BPR Patent Families by Assignee: ARM 3, NVIDIA 3, Intel 2, IBM 2, Samsung 2, Qualcomm 2, Synopsys 1, AMD 1 Horizontal bar chart showing count of buried power rail patent families per assignee. ARM and NVIDIA each hold 3 families; Intel, IBM, Samsung, and Qualcomm each hold 2; Synopsys and AMD each hold 1. Source: PatSnap Eureka patent analysis, 2025. 0 1 2 3 ARM 3 NVIDIA 3 Intel 2 IBM 2 Samsung 2 Qualcomm 2 Synopsys 1 AMD 1

Dynamic IR Drop Suppression: Single vs. Stacked BPR

Backside buried metal PDNs with distributed capacitance suppress dynamic IR drops by 10% (single-chip) and 59% (four-chip stack) in a 3.9M-gate crypto core at 30 MHz (Kobe University, 2021).

Dynamic IR Drop Suppression: Single-chip 10%, Four-chip stack 59% — Kobe University 2021 backside buried metal PDN, 3.9M-gate crypto core at 30 MHz Two donut charts comparing dynamic IR drop suppression percentages. Left donut shows single-chip configuration at 10% suppression; right donut shows four-chip stacked configuration at 59% suppression. Source: Kobe University, 2021, via PatSnap Eureka literature analysis. 10% Single-chip IR Suppression 59% 4-chip stack IR Suppression

Search the full BPR patent landscape with AI-powered analysis in PatSnap Eureka.

Search BPR Patents in Eureka
Cross-Sectional Engineering

Upsizing BPR Width Beyond Inter-Cell Spacing

IR drop is inversely proportional to rail cross-sectional area. Four complementary strategies from Synopsys, Qualcomm, AMD, and GlobalFoundries address this from different angles — from spatial relocation to aspect ratio control and variable density routing.

Synopsys · 2025

BPR Width Greater Than Vertical Space Width

Synopsys explicitly claims an integrated circuit structure where the upsized BPR has a width greater than the width of the vertical space separating adjacent component-bearing structures. The contiguous cavity formed adjacent to the vertical space accommodates a wider conductor without displacing active device regions. By increasing cross-sectional area beyond the inter-cell spacing, the rail resistance per unit length drops proportionally, reducing the IR voltage drop experienced by transistor sources connected to the BPR.

Width > inter-cell gap → resistance ↓
Qualcomm · 2024

High-Aspect-Ratio Voltage Rails (H/W > 1.0)

Qualcomm's invention provides first and second high-aspect-ratio voltage rails with height-to-width ratio greater than 1.0, where the height of each rail exceeds its width. The increased height allows a larger cross-sectional area within a constrained lateral footprint — directly limiting resistance and the corresponding IR drop. This principle is architecturally equivalent to BPR upsizing but implemented via aspect ratio control rather than spatial relocation below the device layer.

H:W > 1.0 → larger cross-section
AMD · 2023

Dual-Height Cells with Reduced CPP Power Posts

AMD's method employs dual-height standard cells at half the width of single-height cells, placing power posts with 1 contacted gate poly pitch (CPP) to align power rails across cell rows. By increasing the number of power delivery contacts per unit length of the power rail, the effective via resistance is reduced, further lowering IR drop without requiring changes to the back-end metal stack. This cell-architecture-level approach is complementary to process-level BPR strategies.

More contacts per rail length → via R ↓
GlobalFoundries · 2014

Variable Power Rail Density via Current-Aware Routing

GlobalFoundries' variable power rail design creates standard cells with M2 power rails, then computes power rail current density after routing. Cells with current density below a predetermined threshold are replaced with functionally equivalent cells lacking M2 rails. The design iterates until convergence. This selective de-population of over-provisioned rails restores routing resources in low-activity regions, while high-current zones retain thicker rail infrastructure — a variable-density philosophy that informs modern BPR placement algorithms.

Current-density-aware rail provisioning
PatSnap Eureka

Map the full BPR cross-section IP landscape

Identify white spaces, freedom-to-operate risks, and key claim boundaries across Synopsys, Qualcomm, AMD, and GlobalFoundries filings.

Analyse BPR IP with Eureka
Multi-Layer Strategies

Front-Side Augmentation, Jumper Connections & Backside PDNs

Before and alongside full BPR adoption, TSMC, Samsung, ARM, and Kobe University developed complementary multi-layer and backside power delivery techniques that remain relevant at 2nm.

Organization Technique IR Drop Mechanism Year
TSMC Jumper connections between vertically separated power/ground rails Creates additional parallel current paths, reduces rail impedance, distributes current uniformly — mitigates both IR drop and electromigration 2017
Samsung Dual-metal-layer routing: timing-critical cells powered from lower-resistance second metal layer Routes high-priority cells to lower-resistance rails, reducing peak noise and IR drop on most timing-sensitive paths 2020
ARM Dual-function buried supply rail: backside power network + signal path for critical nets Reduces front-side routing congestion while providing low-resistance power delivery — synergistic benefit of buried-layer paradigm 2022
ARM Bitcell buried ground rail: ROM cells connected between bitlines and buried VSS rail via backside IVs Reduces length of ground current path, directly reducing resistive voltage drop along VSS rail 2022
Kobe University Backside buried metal PDN with distributed capacitance over full-chip backside area Suppresses dynamic IR drops by 10% (single-chip) and 59% (four-chip stack) in 3.9M-gate crypto core at 30 MHz 2021
IBM Power staple avoidance: shift transistor gates to prevent pin-to-staple alignment during routing Reduces via count in congested areas while maintaining current delivery, managing resistance along local rail segments 2025

Need freedom-to-operate analysis on multi-layer PDN patents?

PatSnap Eureka maps claim scope across TSMC, ARM, Samsung, and IBM filings in seconds.

Run FTO Analysis in Eureka
EDA Co-Design

Placement-Aware IR Drop Mitigation at the Algorithm Level

BPR effectiveness at 2nm depends not only on physical rail geometry but on intelligent standard cell placement that prevents current density spikes along rails. NVIDIA and IBM lead the EDA-side innovation.

NVIDIA: Rail Power Density Aware Placement (2023–2025)

NVIDIA's methodology reorganizes cell placement within bounding box regions before fabrication. Power density across the power rails within each bounding box is normalized based on spatial and temporal power density characteristics of each cell. IR drop is mitigated because distributing current consumption between supply rails reduces current spikes and IR drops. The reorganization achieves minimal timing impact. This EDA-level approach is complementary to BPR geometry: even a low-resistance buried rail will suffer localized IR drop if high-switching-activity cells are clustered at a single rail segment.

🔩

IBM: Power Staple Avoidance for Via Reduction (2025)

IBM's 2025 technique shifts transistor gates within rows to avoid alignment of cell pins and power staples during routing. By preventing pin-to-staple alignment, the methodology reduces via count in congested areas while maintaining adequate current delivery, directly managing resistance along local rail segments. This technique is particularly relevant at 2nm where the number of available routing tracks per cell is severely limited. For further context on international patent filings in this space, WIPO data confirms accelerating BPR-related WO applications from 2022 onward.

🔒
Unlock BPR Co-Design Insights
See how process, cell architecture, and EDA converge at 2nm — plus Intel's TVB self-alignment methodology details.
TVB self-alignment Co-design convergence map + more
Explore in PatSnap Eureka →
Key Players

Who Owns the BPR Innovation Landscape at 2nm?

A small set of vertically integrated IDMs and major IP licensors dominate BPR patent activity. Understanding their claim boundaries is essential for R&D teams targeting 2nm and below. Explore the full PatSnap customer case studies to see how IP teams navigate this landscape.

Intel Corporation

Core BPR Structural Architecture

Intel leads in core BPR structural patents, defining the buried-rail-plus-top-rail shunting architecture and the trench contact via (TVB) self-alignment methodology. Two 2023 WO/US patent families establish the foundational claim that the BPR acts as a shunt for the top-side power rail, with resistance reduced in parallel.

2 patent families · 2023
IBM

Tight Cell-to-Cell Spacing Integration

IBM advances BPR integration under tight cell-to-cell spacing constraints via dual-BPR stacking, as seen in 2023 WO and US filings. IBM also holds the 2025 power staple avoidance patent for routing via reduction at congested cell boundaries.

2 patent families · 2023–2025
ARM Limited

Dual-Function Buried Signal Routing

ARM extends BPR architecture to memory bitcells and dual-function buried signal routing, with three 2022 active US patent families. ARM's buried metal technique enables at least one buried supply rail to serve as a backside signal path for critical signal nets — providing architectural justification beyond IR drop alone for BPR adoption.

3 patent families · 2022
NVIDIA Corporation

EDA-Side Placement Normalization

NVIDIA owns the EDA-side of the BPR problem, with a patent family spanning 2023–2025 on rail-power-density-aware placement. Cell reorganization within bounding boxes normalizes temporal and spatial power density, preventing IR hotspots even on geometrically optimized buried rails. The PatSnap platform tracks all three NVIDIA continuations.

3 patent families · 2023–2025
Integration Flow

Four Interconnected BPR Strategies at 2nm

The dominant technical approaches converge across four interconnected strategies that must be co-designed simultaneously for effective IR drop reduction at the 2nm node.

BPR Co-Design Strategy Flow: Process → Cell → EDA → Verification

All four strategies must be implemented simultaneously. A low-resistance buried rail geometry is ineffective without placement-aware EDA; upsized cross-sections require process integration advances to achieve.

BPR Co-Design Strategy Flow: 4 strategies from process integration to EDA verification 1 Bury Bury VDD/VSS below active plane 2 Upsize Upsize BPR beyond inter-cell spacing 3 Shunt Dual-rail parallel shunting path 4 Normalise Placement-aware current density EDA

Explore how leading semiconductor IP teams monitor BPR patent filings in real time.

Monitor BPR Filings Live
Frequently asked questions

Buried Power Rail IR Drop at 2nm — key questions answered

Still have questions? Let PatSnap Eureka answer them for you.

Ask Eureka About BPR Patents
PatSnap Eureka

Accelerate Your 2nm Power Delivery R&D with AI Patent Intelligence

Join 18,000+ innovators already using PatSnap Eureka to accelerate their R&D. Map BPR claim boundaries, monitor new filings from Intel, IBM, ARM, and NVIDIA, and identify freedom-to-operate risks before tape-out.

References

  1. Integrated circuit structure with buried power rail — Intel Corporation, 2023 (Patent ID: ef41c3bb)
  2. Integrated circuit structure with buried power rail — Intel Corporation, 2023 (Patent ID: 2ae82049)
  3. Buried power rail at tight cell-to-cell space — International Business Machines Corporation, 2023 (Patent ID: edc3e27e)
  4. Buried power rail at tight cell-to-cell space — International Business Machines Corporation, 2023 (Patent ID: 7733b928)
  5. Integrated circuit with buried power rail and methods of manufacturing the same — Samsung Electronics Co., Ltd., 2020
  6. Upsizing buried power rails to reduce power supply resistance and boost cell density scaling — Synopsys, Inc., 2025
  7. Rail power density aware standard cell placement for integrated circuits — NVIDIA Corporation, 2023
  8. Rail power density aware standard cell placement for integrated circuits — NVIDIA Corporation, 2024 (continuation 1)
  9. Rail power density aware standard cell placement for integrated circuits — NVIDIA Corporation, 2024 (continuation 2)
  10. Standard cell circuits employing high aspect ratio voltage rails for reduced resistance — Qualcomm Incorporated, 2024
  11. Standard cell circuits employing high aspect ratio voltage rails for reduced resistance — Qualcomm Incorporated, 2019
  12. Standard cell design architecture for reduced voltage droop utilizing reduced contacted gate poly pitch and dual height cells — Advanced Micro Devices, Inc., 2023
  13. Buried Metal Technique for Critical Signal Nets — ARM Limited, 2022 (Patent ID: 1afe38ec)
  14. Buried metal technique for critical signal nets — ARM Limited, 2022 (Patent ID: b15ec552)
  15. Bitcell architecture with buried ground rail — ARM Limited, 2022
  16. Apparatus and method for mitigating dynamic IR voltage drop and electromigration affects — Taiwan Semiconductor Manufacturing Co., Ltd., 2017
  17. Apparatus and method for mitigating dynamic IR voltage drop and electromigration affects — Taiwan Semiconductor Manufacturing Co., Ltd., 2014
  18. Variable power rail design — GlobalFoundries Inc., 2014
  19. Layout connection isolation technique for improving immunity to jitter and voltage drop in a standard cell — Samsung Electronics Co., Ltd., 2020
  20. Power staple avoidance for routing via reduction — International Business Machines Corporation, 2025
  21. 3-D CMOS Chip Stacking for Security ICs Featuring Backside Buried Metal Power Delivery Networks With Distributed Capacitance — Kobe University, 2021
  22. IEEE — Institute of Electrical and Electronics Engineers — Semiconductor process and EDA standards body
  23. WIPO — World Intellectual Property Organization — International patent filing data and PCT applications
  24. Semiconductor Industry Association (SIA) — Industry roadmap and process node definitions

All data and statistics on this page are sourced from the references above and from PatSnap's proprietary innovation intelligence platform.

Ask PatSnap Eureka
Ask PatSnap Eureka
AI innovation intelligence · always on
Ask anything about buried power rail IR drop at 2nm.
PatSnap Eureka searches patents and research to answer instantly.
Try asking
Powered by PatSnap Eureka