Slides
Page 1
AMBARELLA.COM COPYRIGHT AMBARELLA 2026 AMBARELLA.COM COPYRIGHT AMBARELLA 2026 PROPRIETARY AND CONFIDENTIAL Product & Technology Briefing Las Vegas, Nevada January 6th, 2026
Page 2
AMBARELLA.COM COPYRIGHT AMBARELLA 2026AMBARELLA.COM COPYRIGHT AMBARELLA 2025 This presentation contains forward-looking statements that are subject to many risks and uncertainties. All statements made in this presentation other than statements of historical facts are forward-looking statements, including, without limitation, statements regarding Ambarella’s strategy, future operations, financial targets, future revenues, projected costs, prospects, plans and objectives for future operations, future product introductions, future rate of our revenue growth, the size of markets addressed by the company's solutions and the growth rate of those markets, technology trends, our ability to address market and customer demands and to timely develop new or enhanced solutions to meet those demands, our ability to achieve design wins, our ability to build and deliver products to customers, and our ability to retain and expand our customer and partner relationships. In some cases, you can identify forward-looking statements by terms such as "may," "will," "should," "could," "would," "expects," "plans," "anticipates," "believes," "estimates," "projects," "predicts," "potential," or the negative of those terms, and similar expressions and comparable terminology intended to identify forward- looking statements. We have based forward-looking statements largely on our estimates of our financial results and our current expectations and projections about future events, markets and financial trends that we believe may affect our financial condition, results of operations, business strategy, short term and long-term business operations and objectives, and financial needs as of the date of this presentation. Although these forward-looking statements are based upon information available at the time the statements are made and reflect management’s good faith beliefs, forward-looking statements inherently involve known and unknown risks, uncertainties and other factors that may cause actual results, performance or achievements to differ materially from anticipated future results. Important factors that could cause actual results to differ materially from expectations are disclosed in Ambarella's annual reports on Form 10-K and quarterly reports on Form 10-Q filed with the Securities and Exchange Commission (the “SEC”), particularly in the sections titled “Risk Factors” and “Management’s Discussion and Analysis of Financial Condition and Results of Operations.” You should not place undue reliance on forward-looking statements, which speak only as of the date on which they are made. We do not undertake to update or revise any forward-looking statements after they are made, whether as a result of new information, future events, or otherwise, except as required by applicable law. Moreover, we operate in a very competitive and rapidly changing environment. New risks emerge from time to time. It is not possible for management to predict all risks, nor can we assess the impact of all factors on our business or the extent to which any factor, or combination of factors, may cause actual results to differ materially from those contained in any forward-looking statements we may make. In light of these risks, uncertainties and assumptions, the forward-looking events and circumstances discussed in this presentation may not occur and actual results could differ materially and adversely from those anticipated or implied in the forward-looking statements. Before you invest, you should read the annual and quarterly reports and other documents Ambarella has filed with the SEC for more complete information about the company and its ordinary shares. Additional information will also be set forth in Ambarella's future quarterly and annual reports and other filings made with the SEC from time to time. You may access these documents for free by visiting EDGAR on the SEC web site at www.sec.gov. Forward-Looking Statements 22
Page 3
AMBARELLA.COM COPYRIGHT AMBARELLA 2026 Introduction & Overview Louis Gerhardy, VP Corporate Development Edge AI Platform Leadership Fermi Wang, CEO Evolving Edge AI Landscape From Cloud to Edge Muneyb Minhazuddin, Customer Growth Officer Market View Expert Panel Hosted by Muneyb Minhazuddin AI Architecture Bob Kunz, Chief Architect Evolution of Software and Agents Malhar Palkar, Software Principal AD SW Stack Advances and Learnings for Mobile Robotics Alberto Broggi, GM VisLab Agenda 3
Page 4
AMBARELLA.COM COPYRIGHT AMBARELLA 2026 Agenda Portfolio Update Muneyb Minhazuddin, Customer Growth Officer IoT Portfolio & Roadmap Update Auto Portfolio & Roadmap Update GTM & Developer Zone Scaled Operations Chan Lee, COO Summary and Q&A Louis Gerhardy, VP Corporate Development Reception Drinks + appetizers served in the bar area, mingle with market and investment analysts 4
Page 5
AMBARELLA.COM COPYRIGHT AMBARELLA 2026 AMBARELLA.COM COPYRIGHT AMBARELLA 2026 Dr. Fermi Wang Chief Executive Officer & Co-Founder Chan Lee Chief Operating Officer Today’s Presenters Fermi is a seasoned executive with deep domain expertise in the video processing industry .Before co-founding Ambarella in 2004, Fermi was CEO and co-founder of Afara Websystems, which pioneered throughput computing for servers and was acquired by Sun Microsystems. Prior to Afara, Fermi held several executive management positions at C-Cube. As vice president and general manager of the home media division, he had P&L responsibility for silicon solutions to the DVD , digital video recorder, and digital video distribution and production markets. Fermi holds several digital video-related patents, including one of MPEGLA's core MPEG-2 and MPEG-4/AVC patents. In 2016, he received the National Taiwan University Distinguished Alumni Award. He is therecipient of the 2018 Thomas Egleston Medal from the School of Engineering and Applied Science at Columbia University for his distinguished engineering achievement, and he is currently a Trustee of Columbia University . Prior to Ambarella, Chan Lee was the directorof VLSI at Afara, responsible for managing both the design and verification team, and driving key design and verification methodologies. Lee spent a significant part of his career at Intel, including: manager of the silicon debug team, design manager of the first-generation Pentium 4, and manager and technical lead on the P6 (Pentium Pro). Bob Kunz Chief Architect Bob Kunz is Chief Architect at Ambarella, where he has defined and led the company’s AI architecture since its first generation of products. His work spans image and video processing, computer vision, and edge AI platforms for smart cameras, automotive, and embedded systems. Before Ambarella, Bob spent nearly a decade at Stanford University , working under John Hennessy on large-scale systems. Today , he guides the next generation of Ambarella’s AI silicon. 5
Page 6
AMBARELLA.COM COPYRIGHT AMBARELLA 2026 AMBARELLA.COM COPYRIGHT AMBARELLA 2026 Today’s Presenters Muneyb Minhazuddin Customer Growth Officer Muneyb Minhazuddin is a proven growth leader with a track record of incubating and scaling new B2B businesses at the intersection of emerging technologies and market inflection points. He has successfully driven innovation at global technology leaders, including Intel, VMware, and Avaya, building entirely new multibillion-dollar product lines and opening markets beyond their core business portfolios. Before joining Ambarella, Muneyb served as Chief Marketing Officer for Intel’s Network and Edge Group, where he led the go-to-market strategy and helped establish Intel’s commercial Edge AI software business. Louis Gerhardy Vice President, Corporate Development Louis draws on 30 years of semiconductor industry strategic finance experience bridging both the industry as well as Wall Street. Prior to Ambarella, Louis was responsible for Corporate Development & Strategy at Freescale (formerly Motorola) until its acquisition by NXP . At Freescale, Louis developed and executed 13 acquisitions and strategic investments. Prior to this, Louis was a Senior Semiconductor Industry Analyst at Morgan Stanley for 10 years where he cumulatively covered more than 50 publicly traded semiconductor companies and vetted hundreds of private companies yielding 17 lead managed IPOs (including Broadcom, Cavium, Nvidia and Silicon Laboratories) and other capital market transactions valued at more than $50 billion. Louis has the unique distinction of #1 team rankings, as determined by Institutional Investor, for both sell-side research as well as corporate rankings. 6
Page 7
AMBARELLA.COM COPYRIGHT AMBARELLA 2026 AMBARELLA.COM COPYRIGHT AMBARELLA 2026 Today’s Presenters Alberto Broggi General Manager, VisLab Malhar Palkar Software Principal Malhar Palkar is a Senior Principal Engineer at Ambarella, where he has spent over a decade advancing AI/ML and multimodal technologies. He is a seasoned expert in A/V systems and serves as a key technical voice for developers using Ambarella’s AI/ML tools and software stack. Prior to Ambarella, Malhar led HP’s printer engineering team in India. Earlier , following LSI Logic’s acquisition of C-Cube Microsystems, he was a founding member of LSI Logic in India. During his tenure at C-Cube, Malhar held multiple senior leadership roles and played a pivotal role in the development of DVD players and recorders. Alberto Broggi is the General Manager at VisLab srl (a spinoff of the University of Parma, acquired by Ambarella, Inc. in 2015) and a professor of computer engineering at the University of Parma in Italy . Widely considered a leading pioneer in the use of machine vision for automotive applications,Broggi has been actively engaged in autonomous driving research and development for over twenty-five years. He is both an IEEE and IAPR Fellow, was awarded two European Research Council (ERC) recognitions for his contributions to the development of perception for autonomous driving, and in 2017 was honored with the IEEE Medal for Environmental and Safety Technologies “for leadership in vehicular environmental perception, and for setting worldwide milestones in safe and reliable intelligent vehicles. ” In 2018, P opular Mechanics called Broggi’s patent for a self-driving car one of the “15 Patents That Changed the World. ” 7
Page 8
AMBARELLA.COM COPYRIGHT AMBARELLA 2026 PROPRIETARY AND CONFIDENTIAL AMBARELLA.COM COPYRIGHT AMBARELLA 2026 Dr. Fermi Wang Chief Executive Officer Edge AI Platform Leadership
Page 9
AMBARELLA.COM COPYRIGHT AMBARELLA 2026 Evolution of AI (1998 – Present) 9 AMBARELLA Founded 2004 EdgeBox/Appliance Physical EdgeServer Edge AI IoTEndpoints Multi-sensor, Multi-modality, Central AI Processor 3rd Gen CVflow® 2nm, 4nm, 5nm (CV7x, CV3-AD, N1) Human Viewing Video Processors (A1, A3, A5S, S2, H1) 2nd Gen CVflow® 10nm, 5nm (CV2x, CV5x) Edge AI CNN Networks Automotive (ADAS) Ambarella’s Edge Advantage in the Golden Age of AI Early conviction and sustained innovation position Ambarella to lead the next era of edge intelligence 1998 CNN 2004 Picture sharing on Internet 2012 AlexNet 2017 Transformer 2020 GPT-3 2022 ChatGPT 2023 GPT-4V 2026 o1 2025 DeepSeek-R1 2006 2018 2022 2025 DeepSeek Live Demo on Ambarella SoCs ISC West 2025
Page 10
AMBARELLA.COM COPYRIGHT AMBARELLA 2026 10 CV3-AD family of central domain controllers Feb 2021P ¼ of CV22 AI Jul 2019P ½ of CV22 AI Dec 2018P Baseline (CVflow DNN AI) Apr 2021P Aug 2019P 4x CV22 AI Apr 2021P 2x CV22 AI 2022P 3.5x CV22 AI ✓ 5nm Samsung “New Products” ✓ 20xx = Calendar Year P = Production S = Sample 2026P ~6x CV22 AI ✓ 2026 ~13x CV22 AI ✓ 2026S ~27x CV22 AI ✓ 2023S ~80x CV22 AI ✓ 2022S ~150x CV22 AI ✓ 2026P ~1.5x CV22 AI ✓ C2025S ~27x CV22 AI ✓ 2nm Samsung 5nm Samsung10nm Samsung New 2nm SoC AI Inference SoCs 3rd generation; AI (e.g., Transformer and CNN AI networks) Video Processors (e.g., Human Viewing) 2nd generation; AI (e.g., CNN AI networks) More than 40 Million AI SoCs Shipped Cumulatively ASP double digits to triple digits ASP range $10 to $75 ASP range single digits Our ASP is Rising with New Generations of our AI SoCs Opportunity & Strategy by Market 2026S ~8x CV22 AI ✓ 4nm Samsung Broad Edge AI SoC Portfolio Continues to Expand
Page 11
AMBARELLA.COM COPYRIGHT AMBARELLA 2026 AMBARELLA.COM COPYRIGHT AMBARELLA 2025 We Are Capturing More Value per Edge AI Unit Ambarella’s average selling price increased ~20% in Q3 F2026 (October 2025) 1. Serial CPU workloads → parallel AI workloads = more edge AI “horsepower” required 2. Workloads moving from the Cloud and Data Center to the Edge 3. Larger AI models = more parameters and data to be supported 4. Continued support for CNN networks but incremental processing required for the more sophisticated transformer network applications 5. Higher resolution sensors = more data per sensor more sensors per system = more data per system All of Ambarella’s new edge AI SoCs in development are expected to command an ASP well above the current ASP Video Processor AI Processor Growing importance and proportion of edge AI inference processing in our SoCs More edge AI performance, intelligence and efficiency required AI accelerator becoming a larger and larger proportion of our AI SoCs 11
Page 12
AMBARELLA.COM COPYRIGHT AMBARELLA 2026AMBARELLA.COM COPYRIGHT AMBARELLA 2026 12 Our Platform of Edge AI SoCs and SW Provide Customers with a High Degree of Programmability and Flexibility Systems Edge Endpoints Edge Infrastructure Portable Video Auto Drones Silicon Video Processors 2nd Gen Edge AI SoCs 3rd Gen Edge AI SoCs L2+ Passenger Vehicles Enterprise / Public Security Robotics Perception Module Enterprise Video-Conferencing Crowdsourcing 360◦ 8K camera Scalable platform with minimal incremental investment to address new markets Market Platform Software Application Software AI ISP AI Libraries Robotics Cooper Developer Platform Cooper SDK Customer #1 SW Opensource AI Toolchain L2+ to L4 SW Customer #2 SW Customer #3 SW Security Telematics Edge InfraCustomer #n SW Ambarella developed Customer developed Customer #4 SW Customer #5 SW Drone 12
Page 13
AMBARELLA.COM COPYRIGHT AMBARELLA 2026 Rapid Expansion in the Edge AI Use Cases Edge AI Design Win & Production Scheduled Access Control Edge Endpoint Edge Infra. Security/SurveillancePortable Video Auto Safety/Telematics Auto L2+ to AutonomyOther (includes emerging) Emerging Body Worn Action/Sports Glasses Cinematic Crowdsourcing Enterprise Retail Smart Home Smart City/Public Traffic Mgmt. Video Conferencing Medical-surgery Trail Instrumentation Humanoid Robot Perc. Module Machine Vision AMR - Delivery 360o Aerial Drone AI Telematics Front ADAS eMirror Driver/Passenger Monitoring Drive Recorder L2+ Domain Controller L4 Domain Controller L2+ Software L4/L5 Software 4D Radar Software Developer AI Workstation AI Vision Box On-Premise Servers
Page 14
AMBARELLA.COM COPYRIGHT AMBARELLA 2026 Edge AI 5-Year Revenue CAGR 64% AI growth offset major (geopolitical) revenue loss from 3 major customers commencing FY20 +12% as reported total revenue CAGR (AI +64%)* +18% total revenue CAGR (AI +71%) excluding FY20 entity list impact $- $100 $200 $300 $400 FY20/ CY19 FY21/ CY20 FY22/ CY21 FY23/ CY22 FY24/ CY23 FY25/ CY24 FY26/ CY25 1) 5-Year CAGR FY2021→FY2026 2) *F2026e is based on midpoint of Ambarella’s F2026 revenue guidance range estimating +36% to +38% revenue growth in fiscal 2026 3) October 2019 (FY20) Entity List negative revenue impact (depicts revenue decline fromDahua, DJI & Hikvision) $390M ~80% $285M $226M $338M$332M $223M$229M Video ProcessorsAI 14
Page 15
AMBARELLA.COM COPYRIGHT AMBARELLA 2026AMBARELLA.COM COPYRIGHT AMBARELLA 2026 Ambarella Key Statistics ~40M Edge AI SoCs Shipped 400M SoCs Shipped (Cumulatively) $1.3B Invested in AI R&D (Cumulatively over the last decade) 2026 Small-Cap Companies America’s Most Successful N1 Edge GenAI SoC Family >370 Unique AI Customer Projects in Production (Cumulative) >200 Unique AI Model Architectures Ported 15
Page 16
AMBARELLA.COM COPYRIGHT AMBARELLA 2026 PROPRIETARY AND CONFIDENTIAL AMBARELLA.COM COPYRIGHT AMBARELLA 2026 Muneyb Minhazuddin Customer Growth Officer Product & GTM Growth Initiatives
Page 17
AMBARELLA.COM COPYRIGHT AMBARELLA 2026AMBARELLA.COM COPYRIGHT AMBARELLA 2026 Edge : Next Evolution of Digital Transformation Driver Workloads / Devices Enabler Information Technology App Economy The rise of machines Client-Server 200M Internet Cloud Native 300M Mobile device 3-7B • Enterprise IT • Process Automation • Developers defined • SaaS Edge Native 250M Edge Native 7-20B Mobile/Developer SDKs • Vertical Line of Business • Agentic Automation Pervasive connectivity/Edge AI 1990-2010 Data Center 2010-2020 Mobile/Cloud First 2020-30 Intelligent Edge Transforming how companies operate Transforming how people operate Transforming how “everything” operates 17
Page 18
AMBARELLA.COM COPYRIGHT AMBARELLA 2026AMBARELLA.COM COPYRIGHT AMBARELLA 2026 18 Cloud and Edge AI Requirements by Location Proximity Cloud Data Center IoT Endpoints Edge Devices Physical Edge Box / Appliance Far Edge Edge Server Near Edge LLM 500B – 100T+ LLM 100B – 400B MLM 7B – 70B VLM / SLM 500M – 7B MLLM 10B – 50B Edge AI Model Size
Page 19
AMBARELLA.COM COPYRIGHT AMBARELLA 2026AMBARELLA.COM COPYRIGHT AMBARELLA 2025 Edge AI Demands a Fundamentally Unique Processing Architecture Higher levels of efficiency and new capabilities enabled with edge AI Edge AI enables: • Privacy and security • Realtime processing • Reduced network bandwidth requirement • Low power consumption • Synergies between cloud and edge AI processing – AI processing throughout the network hierarchy 19
Page 20
AMBARELLA.COM COPYRIGHT AMBARELLA 2026AMBARELLA.COM COPYRIGHT AMBARELLA 2026 Expanding TAM for Edge AI *IDC ’27 Edge Infrastructure Spending Guide (excludes Apps) Edge Spend $360B* Hardware Services Software $126B Auto IoT Current TAM ‘27 + 75 Unchanged SAM 20
Page 21
AMBARELLA.COM COPYRIGHT AMBARELLA 2026AMBARELLA.COM COPYRIGHT AMBARELLA 2026 Requires a Full Stack Edge AI Solution Silicon Systems Software Applications Near Edge Far Edge Physical Edge Devices 21
Page 22
AMBARELLA.COM COPYRIGHT AMBARELLA 2026 222323 av Systems Software Silicon Applications Video Processors Agentic AI Tools Open Source Frameworks 2nd Gen Edge AI SoCs 3rd Gen Edge AI SoCs Ambarella Algorithm Stacks AI Tool Chain Cooper Model Garden es et ViT PCIe CardsCooper Developer Kits Edge AI Servers Edge InfrastructureEdge Endpoints Automotive Modules Security Camera Modules Machine Vision Camera Modules Edge AI Boxes Operating Systems T free SDK and Tool Chain (ISP & AI) Developer Framework ASIL Safety Framework Agentic Framework Cooper Developer Platform Real-Time Framework Functional Safety Platform Embedded RTOS Platform
Page 23
AMBARELLA.COM COPYRIGHT AMBARELLA 2026 PROPRIETARY AND CONFIDENTIAL AMBARELLA.COM COPYRIGHT AMBARELLA 2026 Panel Bob Kunz Chief Architect Malhar Palkar Software Principal Alberto Broggi General Manager, VisLab Panel Discussion
Page 24
AMBARELLA.COM COPYRIGHT AMBARELLA 2026 AI Accelerator Evolution • Flexibility OR Scale Data Center Power Trend Training Flops Over Time • AI Architect OR Environmental Scientist • Constraint OR Balance • Accelerator OR SoC DSP AI CPU + Compute Intensity vs Model Size 24
Page 25
AMBARELLA.COM COPYRIGHT AMBARELLA 2026 Ambarella’s Differentiation Flexibility AND Scale AI Architect AND Environmental Scientist Constraint AND Balance Accelerator AND SoC 25
Page 26
AMBARELLA.COM COPYRIGHT AMBARELLA 2026 AMBARELLA.COM COPYRIGHT AMBARELLA 2026 AI Model & System Evolution Designing AI Systems for Efficiency, Reliability, and Real-World Deployment Transformers Everywhere Transformers unify vision, language, audio, replacing fragmented task-specific models Multimodal Edge Inference is the New Bottleneck Decoding, reasoning, & long context drive latency, memory, and rate challenges Latency & DRAM Bandwidth Autoregressive Decoding Long Context Windows Optimization is Architectural Essential techniques make large models deployable through architectural efficiency Quantization Mixed Precision Speculative Decoding LoRA Token Pruning Distillation Model Size Efficiency Frameworks Enable Real Applications Abstracting model churn to accelerate reliable application development Serving RAG Agentic Workflows Tool Use Evaluation Model Churn Reliable AI Apps Safety, Security & Trust Are Mandatory Required for responsible, enterprise-ready AI systems Guardrails Prompt Security Controlled Agent Behavior Privacy Protection 26
Page 27
AMBARELLA.COM COPYRIGHT AMBARELLA 2026 HiveMind CLIAgentic CLI LangGraph UI Cooper Home (Perf. Analytics) Sample Applications Lychee UI Cooper Vision Radar/Lidar T oF | NIR | Structured Light Visible Sensors (Monocular, Stereo) Cooper Foundation EazyAI Framework Model Optimizer ML Training Toolkit ML Porting Toolkit SpongeTorch Ambarella Extension QAT Training Pruning Training CNNGen ONNX Converter PC Simulator VLMGen VLM/LLM Weights Convertor AmbaGPTQ CNN RetinaFace YoloX ResNet50 Facenet ViT OWL-ViT LongCLIP LLM/VLM LLaVA OV 7B Cooper Core Lychee System Fedora On-chip Compiler Desktop Cooper SDK AI GenAI Shepherd Inference Engine NN Inference Image Pre-Process Vision ISP Decoder Encoder AI MCP Transformers Robot ROS2 LeRobot Vision V4L2 Gstreamer Cooper Metal Cooper Max Cooper Pro Cooper Mini PCIe Cards Cooper Foundry Cooper Developer Platform Cooper Max Cooper Pro Cooper Mini PCIe Cards 27
Page 28
AMBARELLA.COM COPYRIGHT AMBARELLA 2026 AMBARELLA.COM COPYRIGHT AMBARELLA 2026 Trends in Autonomous Driving Software Stacks 28 Radar Vision AD System Data Log End to End Model
Page 29
AMBARELLA.COM COPYRIGHT AMBARELLA 2026 Robotics Learnings from AD SW • One architecture, multiple markets • Common building blocks • Proven scalability • From algorithmic → AI → end-to-end • Faster deployment, lower development risk • High-autonomy ready • Two-layer architecture: • Fast real-time control • LLM-based high-level reasoning • From demo to production • Reusable data pipeline (acquisition, labeling, selection) • Automotive-grade know-how in safety, power efficiency, and hardware deployment 29
Page 30
AMBARELLA.COM COPYRIGHT AMBARELLA 2026AMBARELLA.COM COPYRIGHT AMBARELLA 2025AMBARELLA CONFIDENTIAL av Systems Software Silicon Applications Video Processors Agentic AI Tools Open Source Frameworks 2nd Gen Edge AI SoCs 3rd Gen Edge AI SoCs Ambarella Algorithm Stacks AI Tool Chain Cooper Model Garden es et ViT PCIe CardsCooper Developer Kits Edge AI Servers Edge InfrastructureEdge Endpoints Automotive Modules Security Camera Modules Machine Vision Camera Modules Edge AI Boxes Operating Systems T free SDK and Tool Chain (ISP & AI) Developer Framework ASIL Safety Framework Agentic Framework Cooper Developer Platform Real-Time Framework Functional Safety Platform Embedded RTOS Platform
Page 31
AMBARELLA.COM COPYRIGHT AMBARELLA 2026 PROPRIETARY AND CONFIDENTIAL IoT
Page 32
AMBARELLA.COM COPYRIGHT AMBARELLA 2026 AMBARELLA.COM COPYRIGHT AMBARELLA 2025 IoT Applications ~70% F2025 Revenue Sub-segments: Edge Endpoints and Edge Infrastructure ENTERPRISE SECURITY HOME SECURITY PORTABLE VIDEO OTHER ROBOTICS DEVELOPER AI WORKSTATIONS AI BOX ON-PREMISE AI SERVERS CV2 / CV5 / CV7 N Family Edge Endpoint Edge Infrastructure • Our IoT opportunity is now dominated by AI. IoT endpoints currently represent a vast majority of our revenue. • Expanding breadth of AI applications. Edge AI originated in enterprise security markets, then public and smart home applications, automotive safety and telematics, and has now expanded to multiple portable video applications, with multiple robotic, edge infrastructure and other “green shoot applications” on the horizon. • The 3rd generation AI accelerator embedded in our new SoCs enables advanced AI reasoning and vision AI applications supporting leading GenAI models ranging from 500 million to 70 billion parameters. • Our scalable technology enables us to now extend our reach into the edge infrastructure market. 32
Page 33
AMBARELLA.COM COPYRIGHT AMBARELLA 2025 Edge Infrastructure 33 Feb 2021P ¼ of CV22 AI Jul 2019P ½ of CV22 AI Dec 2018P Baseline (CVflow DNN AI) Aug 2019P 4x CV22 AI 2022P 3.5x CV22 AI 5nm Samsung 20xx = Calendar Year P = Production S = Sample 2026P ~6x CV22 AI 2022S ~150x CV22 AI 2026P ~1.5x CV22 AI C2025S ~27x CV22 AI 2nm Samsung 5nm Samsung10nm Samsung New 2nm SoCs Edge AI SoCs 2026S ~8x CV22 AI 4nm Samsung 3rd gen AI (e.g., Transformer and CNN AI networks) Human Viewing 2nd gen AI (e.g., CNN AI networks) New Scalable SoC Family Unified SDK for portable SW stack development Leading Image Quality Long-time image leader in camera market AI-based ISP Power Efficiency High AI performance with minimal power budget Modern GenAI Support Support for latest LLM/VLM/GenAI models Balanced Architecture Sustainable AI performance with all functions enabled Patented CVflow® IP with low memory & low power
Page 34
AMBARELLA.COM COPYRIGHT AMBARELLA 2026AMBARELLA.COM COPYRIGHT AMBARELLA 2026 34 CV7: Powerful Edge AI 8K Vision SoC With Industry- Leading AI and Multi-Sensor Perception Performance >2.5x AI, CVflow ® 3.0 2x encode ~2x CPU Performance 20% lower power 5nm CV5 CV7 4nm Enterprise Security Sports / WearableDrones Automotive
Page 35
AMBARELLA.COM COPYRIGHT AMBARELLA 2025AMBARELLA.COM COPYRIGHT AMBARELLA 2025 35 Play Keenfinity Sabrina Video IQSight Testimonial Video (Not Available on Replay)
Page 36
AMBARELLA.COM COPYRIGHT AMBARELLA 2026 AMBARELLA.COM COPYRIGHT AMBARELLA 2026 36 Ambarella Families of SoCs: CV2x: CV5x: CV7x: N1x: Ambarella in Robotics – From Perception to Full Autonomy Ambarella combines perception and AI to enable a seamless integration of edge compute solution Ambarella Ecosystem (Partner) Offerings: • Full Camera Solutions • Camera/Stereo Modules • AI Boxes • SoMs CV28m CV25 CV22 CV2 CV75 CV72 CV52 CV5 N1-655 N1 Ambarella Robotics in Action! Aerial Drones – the 1st high volume application: Antigravity 360 degrees drone (CV5)
Page 37
AMBARELLA.COM COPYRIGHT AMBARELLA 2026 AMBARELLA.COM COPYRIGHT AMBARELLA 2025 Auto
Page 38
AMBARELLA.COM COPYRIGHT AMBARELLA 2026 AMBARELLA.COM COPYRIGHT AMBARELLA 2026 38 Auto Applications ~30% of F2025 Revenue Sub segments: Safety/Telematics and Autonomy DRIVE RECORDER E-MIRROR SURROUND IN-CABIN (DMS/OM S) FORWARD ADAS AI TELEMATICS L2+ TO L4 CENTRAL DOMAIN CONTROLLER L2+ to L4 SW 4D Radar SW CV2/CV5/CV7 Series Safety/Telematics/ADAS Autonomy • Our auto SAM is now dominated by AI. Auto Safety/Telematics currently represents a majority of our automotive revenue • Many Auto Safety/Telematics applications now utilize AI inferencing technology, increasing the utility of drive recorders, e-mirrors, surround view, driver and occupant monitoring and forward ADAS. A combination of increasing market penetration and/or share gains are expected to be key factors in our anticipated Safety/Telematics growth • In Auto Autonomy our CV3-AD platform includes AI central domain controllers, L2+ to L4 SW stack IP and 4D Radar SW. Our platform is differentiated by our superior power efficiency, open platform, high performance (ability to support end-to-end AI and Birds-Eye View transformer for early fusion) and our scalable CV3 -AD SoC family CV3 Series
Page 39
Automotive Vision SoCs
Page 40
AMBARELLA.COM COPYRIGHT AMBARELLA 2026AMBARELLA.COM COPYRIGHT AMBARELLA 2026 40 Play Kodiak Video Kodiak Testimonial Video (Not Available on Replay)
Page 41
AMBARELLA.COM COPYRIGHT AMBARELLA 2026 AMBARELLA.COM COPYRIGHT AMBARELLA 2026 DEEP SENSING Static and dynamic environmental perception Ambarella AD Stack & Data Pipeline Camera Images Oculii 4D Imaging Radar Real Time Map Generator Deep Obstacle Sensing DEEP PLANNER Motion forecasting Maneuver and trajectory planning Vehicle control DATA CAPTURE data acquisition system AUTO ANNOTATION data automatic annotation tool DATA SELECTION ODD coverage statistics training data selection Online Software Stacks Offline Data Pipeline 41
Page 42
AMBARELLA.COM COPYRIGHT AMBARELLA 2026 AMBARELLA.COM COPYRIGHT AMBARELLA 2026 PROPRIETARY AND CONFIDENTIAL Go To Market
Page 43
AMBARELLA.COM COPYRIGHT AMBARELLA 2026 Current Go-To-Market Motion 43 Auto OEMs Edge Endpoint Edge Infrastructure Automotive Safety/ADAS Automotive Autonomy OEMs Auto Tier 1’s Sales Rep & Logistics Stocking Distributor Auto OEMs Auto ODMs Camera ODMs OEM Tier 2 Auto Tier 2’s Sales and Support Offices • USA • Taiwan • Shenzhen • Shanghai • Korea • Germany • Japan • Italy Field Sales FAEs Design Centers SDK Development Etc. Etc. Etc. ODM/Tier 1 OEM Tier 2
Page 44
AMBARELLA CONFIDENTIAL av Systems Software Silicon Applications Video Processors Agentic AI Tools Open Source Frameworks 2nd Gen Edge AI SoCs 3rd Gen Edge AI SoCs Ambarella Algorithm Stacks AI Tool Chain Cooper Model Garden es et ViT PCIe CardsCooper Developer Kits Edge AI Servers Edge InfrastructureEdge Endpoints Automotive Modules Security Camera Modules Machine Vision Camera Modules Edge AI Boxes Operating Systems T free SDK and Tool Chain (ISP & AI) Developer Framework ASIL Safety Framework Agentic Framework Cooper Developer Platform Real-Time Framework Functional Safety Platform Embedded RTOS Platform 44
Page 45
AMBARELLA.COM COPYRIGHT AMBARELLA 2026AMBARELLA.COM COPYRIGHT AMBARELLA 2026 45 Scaling through Channels for End-to-End Solutions ISV Software Layer GSI OEM Product Layer Device Design SoC, SDK, AI Acceleration Ambarella Compute Layer AI Models, Analytics, UX Integration Layer Deployment & Orchestration • End-to-end, ready to deploy, catalogue of industry solutions • Backed with fully tested and trained ecosystem of partners, and a community of developers • Scaling through reseller channels by leveraging • Vertical use cases, • Independent Software Vendor (ISV) solutions, • Certified OEM products, • Global Systems Integrator (GSI) services.
Page 46
AMBARELLA.COM COPYRIGHT AMBARELLA 2026AMBARELLA.COM COPYRIGHT AMBARELLA 2026 Launching the Ambarella Developer Zone (developer .ambarella.com) Ambarella Edge AI Software Stack Ambarella Developer Zone is launching as an early access program for AI ecosystem partners to build on Ambarella’s Edge AI stack. This provides them with a collection of optimized models, and blueprints using agentic workflows for building Edge AI applications. Edge AI Applications Agentic Blueprints Model Garden Core SDK APIs Base OS & Frameworks Developer Kits Ambarella Edge AI SoCs Ambarella DevZone Cooper Developer Platform ISV Partners 46
Page 47
AMBARELLA.COM COPYRIGHT AMBARELLA 2026 Ambarella Developer Workflow Any Data | Any Model | Any Modality | Any Edge Any Data Type Any Foundational Model Test Model Deploy Developer Kits Software Stack Real Simulated Build, Optimize & Compile Models Robots Drones IoT Automotive Ambarella AI Kits Cooper Home Kubernetes Open AI Toolchain Agentic AI Tools Cooper SDKCNN VIT LLM/VLM Any Edge Any Modality ToFVision Radar Lidar Audio Port & Train Model Finetuning IR 47
Page 48
AMBARELLA.COM COPYRIGHT AMBARELLA 2026AMBARELLA.COM COPYRIGHT AMBARELLA 2026 Timing Benefits Across the Ecosystem The Ambarella Advantage The Ambarella Ecosystem Advantage • Edge AI inference workloads are exploding • Agentic Systems ready for scale • Customers get an open & developer-enabled ecosystem, proven AI stacks, predictable TCO, faster scalability • Robust Platform for ISVs to anchor workloads • Faster Time-to-Market with Pre-validated designs • 200+ model arch 48
Page 49
AMBARELLA.COM COPYRIGHT AMBARELLA 2026 AMBARELLA.COM COPYRIGHT AMBARELLA 2025 49 Ambarella Growth Trajectory IoT & Auto customer expansion Scaled Channel GTM New Edge AI Infrastructure Customers
Page 50
AMBARELLA.COM COPYRIGHT AMBARELLA 2026 PROPRIETARY AND CONFIDENTIAL AMBARELLA.COM COPYRIGHT AMBARELLA 2026 Scaled Operations Chan Lee Chief Operating Officer
Page 51
AMBARELLA.COM COPYRIGHT AMBARELLA 2026 – 100 200 300 400 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 22 years of Exceptional Silicon Execution 51 Cumulative SOCs Shipped in millions • Exceptional: 40+ production tapeouts – latest in 2 nm – 97+% A1/A0 production, and 70+% A0 production • Technology leader: 14 10/5/4/2 nm production tapeouts,12 SoCs in mass production • Domain expertise: Mature, proven Vision and multimodal AI IP’s and Deep system design expertise • Reliable: High Quality, Efficient execution and Rapid process node migration • Scale and Quality: 400M+ SoCs cumulatively shipped 1st 32nm SoC 1st 28nm SoC 1st 14nm SoC 1st 10nm SoC 1st 5nm SoC 2/4nm T aped out
Page 52
AMBARELLA.COM COPYRIGHT AMBARELLA 2026 Ambarella Supply chain and Operation 52 Wafer Fabrication Texas and South Korea Packaging Taiwan, South Korea and SE Asia Testing Taiwan, South Korea and SE Asia • Strong, decades-long cooperation and partnership • Global geographic diversity and resilience • Flexible and rapid resource allocation 17 years of close partnership with Samsung Foundry Early access to most advanced process and packaging Reliable and Diligent Testing
Page 53
AMBARELLA.COM COPYRIGHT AMBARELLA 2026AMBARELLA.COM COPYRIGHT AMBARELLA 2026 53 Play Samsung Video Samsung Testimonial Video (Not Available on Replay)
Page 54
AMBARELLA.COM COPYRIGHT AMBARELLA 2026 Custom/Semi-custom Design Business for Edge AI customers 54 System OEM’s need for differentiation • Ambarella’s unique system level experience, SDK, IP’s • Successfully taped out a 2nm semi-custom SoC for a leading customer • Exploring additional custom semicustom opportunities 2/4 nm SOC development cost estimated at $100’s of mil Scarce design expertise in advanced process node
Page 55
AMBARELLA.COM COPYRIGHT AMBARELLA 2026 AMBARELLA.COM COPYRIGHT AMBARELLA 2026 Q&A 55