Dorsavi Ltd. · ASX:DVL
Dorsavi — Semiconductor Conference March 2026
27 slides from Dorsavi's conference presentation, Mar 25, 2026
Slide 1: Semiconductor Conference Presentation - March 2026
For personal use only SEMICONDUCTOR CONFERENCE PRESENTATION - MARCH 2026
Slide 2: Important Information & Disclaimer
Important Information & Disclaimer For personal use only
Slide 3: Intelligence at the Ultra Edge: dorsaVi is evolving
For personal use only Intelligence at the Ultra Edge dorsaVi is evolving
Slide 4: DorsaVi Today
DorsaVi Today Overview of dorsaVi’s human movement and sensor business For personal use only
Slide 5: Sensor Technology Overview
Sensor Technology Overview For personal use only 6 1 5 2 4 3
Slide 6: Insight That Changed the Strategy
Insight That Changed the Strategy Capturing data at the ultra-edge is not enough For personal use only • Real value is created when decisions are made and acted upon locally, without round-tripping to the cloud • Latency, bandwidth and connectivity constraints make cloud-dependent inference a liabil
Slide 7: World-Class Intellectual Property
World-Class Intellectual Property For personal use only License and Technology Parter: Manufacturing Partner: Architecture and arrays emulating Commercial IP readiness and biological neurons manufacturing Technical Advisory Board License and Technology Partner: Global Leaders in: Custom RRAM technol
Slide 8: An Overview of Our Partners
An Overview of Our Partners For personal use only *www.topuniversities.com/university-subject-rankings/electrical-electronic-engineering
Slide 9: Mathew Regan - Group CEO
Mathew Regan - Group CEO Experienced Technology Operator For personal use only • Former Chief Executive Officer of Artrya Limited (ASX:AYA) • Driving commercialisation across RRAM, robotics, neuromorphic computing and edge-AI platforms • Led transformation, operations and growth across ASX-listed an
Slide 10: Technical Advisory Board
For personal use only • Former VP Engineering & Manufacturing – • Dean, Electrical & Computer Engineering – • Professor of Physics – Nanyang Omron Robotics Technion – Israel Institute of Technology Technological University • 25+ years in robotics, mechatronics and • Global leader in memristor/RRAM l
Slide 11: Global memory bottlenecks are constraining AI scaling
Global memory bottlenecks are constraining AI scaling For personal use only • AI workloads are increasingly memory-bandwidth constrained rather than compute-limited • Large models consume vast quantities of high-performance memory in data centres • Moving intelligence toward the edge and ultra-edge
Slide 12: At the Sensor:
For personal use only At the Sensor:
Slide 13: Ultra-edge Semiconductor Computing
Ultra-edge Semiconductor Computing Overview of dorsaVi’s ultra-edge enabling technology For personal use only
Slide 14: Neuromorphic Computing
`Traditional For personal use only Computing Types of Event Driven Low-latency Computing Neuromorphic Computing • Information encoded in timing and events, not continuous Edge Computation Ultra-low power values • Neural behavior is implemented directly in hardware, not simulated • Asynchronous, even
Slide 15: The “Energy Wall” at the Edge
The “Energy Wall” at the Edge For personal use only
Slide 16: In-memory inference at the edge
For personal use only • Event-triggered execution rather than continuous clock-driven processing • Millisecond-class response enabled by local, in-memory inference • Sub-microsecond inference paths at the processing-in-memory (PIM) level • Nanojoule-class inference enabled by in-memory processing •
Slide 17: Traditional vs event-driven computing
For personal use only • Continuous clock-driven execution • Event-driven execution instead of continuous clock cycles • Constant power draw - even when idle • Near-zero idle power consumption • Processes all signals, regardless of relevance • Activates only when meaningful signals occur • Optimized
Slide 18: Problems and solutions at the edge
For personal use only Problem: Problem: Problem: Continuous Digital Signal Proce exceeds Cloud-dependent training introduces Camera-based systems <1mW power budgets. latency and cost. increase privacy exposure. Solution: Solution: Solution: Unsupervised on-chip learning detects Unsupervised on-chip
Slide 19: dorsaVi and NTU - RRAM
dorsaVi and NTU - RRAM For personal use only • >2× density scaling from 40 nm → 22 nm • Enables higher density compute-in-memory arrays • Reduces energy per inference • ~20–30% lower write voltage, reducing energy per operation • Faster access and switching enabled by advanced-node integration • >10
Slide 20: Business Division
For personal use only Business Division • Medical-grade wearable • Provides non-volatile, low- • Event-driven processing sensors capture real- power memory tightly coupled inspired by biological world movement at the to compute systems. point of origin • Enables Compute-in-Memory • Processes sensor
Slide 21: Tying Everything Together
Tying Everything Together For personal use only • Sensors + RRAM Improves the intelligence and efficiency of our existing sensor platforms • RRAM + Neuromorphic Enables new ultra-edge AI hardware markets beyond traditional sensing •Sensors + Neuromorphic Creates human-centric intelligent systems cap
Slide 22: Ultra-Edge Target Markets
Ultra-Edge Target Markets For personal use only Vision & Perception Systems [5- Human-Centric Systems [1-4] 7] Robotics [8-10] IoT Devices [14-15] Industrial & Infrastructure [16] Autonomous Systems [11-13] [1] Grand View Research. "Medical Devices Market Size, Share & Trends Analysis Report.”|[2] G
Slide 23: Our Competitive Advantage
Our Competitive Advantage For personal use only 02
Slide 24: Our Three Strategies
Our Three Strategies Key developments in parallel For personal use only Evolve The Existing dorsaVi Develop 22nm RRAM with NTU Unlock Neuromorphic Platform and ITRI Opportunities Move our FDA approved Create the memory layer Apply closed-loop sensing closer to the required for in-memory intelligence
Slide 25: Execution Plan
Execution Plan dorsaVi captures data at the ultra edge and processes it either at the edge device or in the cloud For personal use only Ultra Edge Modular Design Develop Advanced RRAM with NTU Sensor v6.5 Enhancement Product Software and Commercialise and build + ITRI SDK Build Neuromorphics • Integ
Slide 26: Board and Management
Board and Management For personal use only 02
Slide 27: Thank You
\V dorsaVi Thank You Gernot Abl Mathew Regan Executive Chairman @ @@&m Group CEO ga@dorsaVi.com @ =_ N N \Q +61 427 477 298 XQ +61 419 802 653 DX] mregan@dorsaVi.com www.dorsavi.com