Slide 1: Cautionary Statement
Cautionary Statement This presentation contains forward-looking statements concerning Advanced Micro Devices, Inc. (AMD) such as the features, functionality, performance, availability, timing and expected benefits of AMD’s current products, future products and product roadmaps, which are made pursua
Slide 2: AMD @ CES
i il Syl ~~. AMDd¢dl | together we advance_
Slide 3: Solving the World's Most Important Challenges
Solving the World’s Most Important Challenges Accelerating AI from Silicon to Solutions
Slide 4: Powers the Daily Lives of Billions
Powers the Daily Lives of Billions Cloud Healthcare Industrial Transportation Connectivity PCs Entertainment Science
Slide 5: AI is Everywhere
AI is Everywhere Cloud Healthcare Industrial Transportation Connectivity PCs Entertainment Science
Slide 6: AI is for Everyone
AI is for Everyone Professionals Creators Students Scientists Gamers Developers
Slide 7: 5 Billion People Using AI Daily
Within 5 Years 5 Billion 5B People Using AI Daily Active AI Users 1B+ Active AI Users 1M Active AI Users
Slide 8: From Zetta to Yotta Scale Compute
From Zetta to Yotta Scale Compute 10,000x Increase In AI Compute 1 ZettaFLOP 2022 100 ZettaFLOPS 2025 10+ YottaFLOPS
Slide 9: Unlocking AI's Potential Requires Massive Compute Everywhere
Unlocking AI’s Potential Requires Massive Compute Everywhere Data Center Personal Computing Edge / Physical AI
Slide 10: Cloud Personal Computing
Personal Science & Cloud Healthcare Physical AI Space Computing Education
Slide 11: Cloud is the Fastest Path From Idea to Impact
Cloud is the Fastest Path From Idea to Impact Develop Integrate Deploy Scale
Slide 12: Cloud is the Epicenter for AI Development
Cloud is the Epicenter for AI Development Chatbots & Personal Generative AI Agents & Smart Coding Copilots Assistants Media Automation Search Assistants
Slide 13: Together with our Partners
Together with our Partners AMD is Enabling the AI Era AI Innovators 310.AI AMD Partners
Slide 14: AI Model Innovation Accelerating Compute Demand 4x Every Year
AI Model Innovation and Adoption are Accelerating Compute Demand Training Flops Increasing 4x Every Year GPT-5 Llama 4 GPT-4 DeepSeek-R1 Qwen Kimi K2 GPT-3.5 Stable GPT-3 Diffusion 3
Slide 15: AI Model Innovation and Adoption Accelerating Compute Demand
AI Model Innovation and Adoption are Accelerating Compute Demand Training Flops Increasing Inference Tokens Increased 4x Every Year 100x in the Last 2 Years sA & I ou ic m nt no ge to A GPT-5 Au Llama 4 GPT-4 > 1B Qwen DeepSeek-R1 Kimi K2 AI Users GPT-3.5 Stable ChatGPT GPT-3 Diffusion 3 Launch
Slide 16: The Blueprint for Yotta Scale Infrastructure
The Blueprint for Yotta Scale Infrastructure ESUN Leadership Open Rack Open Fabric Rack Scale Efficiency Turnkey Compute Capability Architecture to Scale AI & Serviceability Solutions
Slide 17: AMD "Helios" AI Rack
AMD “Helios” AI Rack
Slide 18: AMD Instinct MI455X GPUs
AMD Instinct™ MI455X GPUs AMD EPYC™ "Venice" CPU
Slide 19: AMD "Helios"
AMD “Helios” The Open Rack Platform Towards Yotta Scale AI 2.9 Exaflops 31 TB 43 TB/s AI Compute HBM4 Memory Scale Out Bandwidth 2nm / 3nm 4,600 18,000 Advanced Process “Zen 6” CPU Cores GPU Compute Units
Slide 20: AMD Instinct MI300X to MI355X
AMD INSTINCT™ MI355X AMD INSTINCT™ AMD INSTINCT™ MI325X MI300X
Slide 21: MI455X: A Massive Leap in AI Performance
A Massive Leap in AI Performance with AMD INSTINCT™ MI455X MI455X 10x Performance Increase (vs. MI355X)
Slide 22: AMD Data Center Portfolio
AMD Data Center Portfolio Hyperscale AI Enterprise AI Sovereign AI & HPC 72 GPU Rack Scale 8 GPU Drop-In Ready Hybrid Compute “Venice” + MI455X “Venice” + MI440X “Venice-X” + MI430X “Venice” MI400 Series “Vulcano”
Slide 23: Open Ecosystem Drives AI Innovation
Open Ecosystem Drives AI Innovation Open Hardware Open Software Open Ecosystem Scale Up Scale Out Infrastructure Software Software Network Network Stack Ecosystem ESUN
Slide 24: AMD ROCm Platform
AMD ROCm Platform ™ Adopted by the Most Popular Developer Communities AI Inference Model Local Image & RL & Open Framework Engines Hub LLM Video Gen Finetuning Physical AI
Slide 25: Extending the Leadership Roadmap
Extending the Leadership Roadmap AMD Instinct™ AMD Instinct™ MI500 Series AMD Instinct™ MI400 Series AMD Instinct™ MI350 Series AMD Instinct™ MI325X MI300A/X 2023 2024 2025 2026 2027
Slide 26: Next-Gen MI500 Series
Next-Gen MI500 Series Next Big Leap in AI Performance AMD CDNA™ 6 | Advanced 2nm Process | HBM4E >1000x AI Performance in 4 Years AMD INSTINCT MI500 Series AMD INSTINCT MI455X AMD INSTINCT MI355X
Slide 27: AI Experiences are Evolving from Cloud to AI PCs
AI Experiences are Evolving from Cloud to AI PCs Private and Secure Personal to Your Usage Proactively Works for You Performance You Can Depend on
Slide 28: AI PC Experiences Everywhere
AI PC Experiences Everywhere Revolutionary Assisted Immersive Enterprise Personal Content Creation Coding Collaboration Productivity Assistance
Slide 29: AMD is at the Forefront of AI PC Innovation
AMD is at the Forefront of AI PC Innovation First First First x86 NPU Copilot+ 200B Model x86 Processor Capable x86 Processor Ryzen™ 7000 Ryzen™ AI 300 Ryzen™ AI Max Series Series Series
Slide 30: AMD Ryzen AI 400 Series
INTRODUCING AMD Ryzen AI 400 Series ™ Powering Next Generation AI PCs 1.7x Up To 12 / 24 60 AI TOPS Faster “Zen 5” CPU Cores / Threads AMD XDNA™ 2 NPU Content Creation 16 8,533 MT/s 1.3x Up To Faster AMD RDNA™ 3.5 GPU Cores Memory Speed Multitasking VS INTEL CORE ULTRA 9 288V
Slide 31: AMD Ryzen AI 400
AMD Ryzen AI 400™ Available Starting Q1 2026
Slide 32: AMD Ryzen AI Max
AMD Ryzen AI Max ™ The PC Reimagined for Creators, Gamers, and AI Developers Content Creation Gaming AI Development
Slide 33: AMD Ryzen AI Max
AMD Ryzen AI Max ™ The PC Reimagined for Creators, Gamers, and AI Developers 16 / 32 40 “Zen 5” CPU Cores / Threads AMD RDNA™ 3.5 GPU Compute Units 50 AI TOPS 128 GB AMD XDNA™ 2 NPU Unified Memory
Slide 34: AMD Ryzen AI Max+
AMD Ryzen AI Max+ ™ Windows and Linux Support Works with 1000s of Applications vs Apple Macbook Pro M5 vs Nvidia DGX Spark 1.4x Up To 1.7x Up To Faster AI Performance Tokens/sec per Dollar LM Studio: GPT-OSS 120B tokens per second 1.8x 1.8x Up To Up To Faster Tokens/sec per Dollar Content Creation L
Slide 35: AI Development & Creation Unleashed
AI Development & Creation Unleashed Infinite Possibilities with AMD Ryzen™ AI Max
Slide 36: AMD Ryzen AI Halo Processors
INTRODUCING AMD Ryzen AI Halo Processors ™ AI Developer Platform Available Q2 26 Linux and Windows Support Optimized Applications Preloaded Models
Slide 37: AMD Ryzen AI Halo Processors
INTRODUCING AMD Ryzen AI Halo Processors ™ AI Developer Platform Available Q2 26 Full AMD ROCm™ Platform Support Dev-Ready Apps Optimized: LM Studio, ComfyUI, VS Code, and more Optimized Models Included: GPT-OSS, FLUX.2, SDXL, and more Day-0 Support for Leading AI Models
Slide 38: AMD Powers Over 1 Billion Gamers & Creators
AMD Powers Over 1 Billion Gamers & Creators
Slide 39: Advancing Healthcare
Advancing Healthcare Genomic & Precision Medicine & Diagnostics & Medical Devices & Patient Computational Biology Drug Discovery Imaging Robotics Care
Slide 40: Advancing Physical AI
Advancing Physical AI High Performance x86 CPUs | Market-Leading Adaptive SoCs Aerospace Healthcare Industrial Robotics Transportation Smart Cities
Slide 41: Over 20 Years of Robotics Innovation
Over 20 Years of Robotics Innovation Industrial Healthcare Space
Slide 42: Physical AI Demands Real-Time Intelligence and Rapid Action
Physical AI Demands Real-Time Intelligence and Rapid Action Perception Decision Action Sensor Fusion Understands AI Compute that Picks the Deterministic Control that the World in Real-Time Right Answer Fast Moves the System Instantly
Slide 43: The Blueprint for Physical AI Architecture
The Blueprint for Physical AI Architecture Leadership Compute Vision & Sensor Open-Source Architectures Processors Ecosystem
Slide 44: Advancing Space Technology
Advancing Space Technology
Slide 45: Partnering to Further Breakthroughs and Discovery
Partnering to Further Breakthroughs and Discovery Jet Propulsion Laboratory
Slide 46: AMD is the Compute Foundation for Modern Science
Communities Receive Earlier, Development of Cleaner Energy Faster Innovation Lifesaving Vaccines More Accurate Warnings and Sustainable Materials Across Industries Reach Patients Sooner AMD is the Compute Foundation for Modern Science
Slide 47: Evaluating and Forecasting Geohazards and Natural Disasters
2.7 Evaluating and Forecasting Geohazards and Natural Disasters Million Cores Predictive Modeling Under Varying Weather Conditions 380 Petaflops Time to Generate Climate Forecasts Reduced by 85%
Slide 48: Developing Higher Performance Batteries
3.1 Developing Higher Performance Batteries Million Cores Optimizing Industrial Facilities 478 Petaflops Improving Accuracy of Dynamic Studies for CO2 Storage
Slide 49: ORBIT-2 Delivers Hyper-Resolution Earth System Predictions
9 ORBIT-2 Delivers Hyper-Resolution Earth System Predictions Using Scalable AI Million Cores Produces Hyper-Detailed Global Weather Predicting Enabling Actionable Insights 1.4 Exaflops Delivers Accuracy in the Range of ~0.98–0.99 When Compared to Observational Data
Slide 50: Optimized Antibody to Neutralize SARS-CoV-2 Variants
11 Pre-Emptively Optimized Antibody to Neutralize Broad Diversity of SARS-CoV-2 Variants Million Cores Anticipates Viral Evolution to Reduce Need for Constant Redevelopment 1.8 Exaflops Improves Biodefense Preparedness and Antibody Therapy Resilience
Slide 51: A New Age of AI Accelerated Innovation and Discovery
A New Age of AI Accelerated Innovation and Discovery First AI Factory for Science Next-Generation Exascale Supercomputer Multi-Tenant AI Cloud Platform Accelerating the Speed of Scientific Discovery
Slide 52: Advancing Education
Advancing Education
Slide 53: $150M Pledge to Advance U.S. AI Education
YW Pledge to Advance | S] 50M nves U.S. Al Education X tments
Slide 54: Empowering everyone to shape the future of AI
800+ Empowering everyone to shape the future of AI Institutions Worldwide
Slide 55: Empowering everyone to shape the future of AI
150,000+ Empowering everyone to shape the future of AI Students in 2026 Through AMD AI Education Program
Slide 56: Empowering AI Learning
AMD2I « Atm Empowering Al Learning
Slide 57: Winning Team Armtender
Winning Team Armtender Emme McDonald Ruzanna Gaboyan Afia Ava
Slide 58: AI Everywhere for Everyone
AI Everywhere for Everyone Unmatched Technology Portfolio Open Ecosystem Deep Co-Innovation
Slide 59: Disclaimer & Attributions
DISCLAIMER & ATTRIBUTIONS DISCLAIMER: The information contained herein is for informational purposes only and is subject to change without notice. While every precaution has been taken in the preparation of this document, it may contain technical inaccuracies, omissions and typographical errors, and
Slide 60: Endnotes
ENDNOTES • GD-164: Day-0 driver compatibility and feature availability depend on system manufacturer and/or packaged driver version. For the most up-to-date drivers, visit AMD.com. • GD-243: Trillions of Operations per Second (TOPS) for an AMD Ryzen processor is the maximum number of operations per
Slide 61: Endnotes
ENDNOTES • SHOP-26: Testing as of November 2025 by AMD. All tests conducted in LM Studio 0.3.30 (Build 2). Vulkan llama.cpp v 1.57.1 used with Ubuntu 24.0.4.3 and therock-gfx1151- 7.9rc1 for AMD Ryzen™ AI Max+ 128GB. CUDA llama.cpp v1.57.1 used with pre-installed DGX OS (based on Ubuntu 24.04.3) Dri