Slides
Page 1
Cautionary StatementThis 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 pursuant to the Safe Harbor provisions of the Private Securities Litigation Reform Act of 1995. Forward-looking statements are commonly identified by words such as "would," "may," "expects," "believes," "plans," "intends," "projects" and other terms with similar meaning. Investors are cautioned that the forward-looking statements in this presentation are based on current beliefs, assumptions and expectations, speak only as of the date of this presentation and involve risks and uncertainties that could cause actual results to differ materially from current expectations. Such statements are subject to certain known and unknown risks and uncertainties, many of which are difficult to predict and generally beyond AMD's control, that could cause actual results and other future events to differ materially from those expressed in, or implied or projected by, the forward-looking information and statements. Investors are urged to review in detail the risks and uncertainties in AMD’s Securities and Exchange Commission filings, including but not limited to AMD’s most recent reports on Forms 10-K and 10-Q.AMD does not assume, and hereby disclaims, any obligation to update forward-looking statements made in this presentation, except as may be required by law.
Page 2
AMD @ CES
Page 3
Solving the World’s Most Important ChallengesAccelerating AI from Silicon to Solutions
Page 4
Powers the Daily Lives of Billions CloudHealthcareIndustrialTransportationConnectivityPCsEntertainmentScience
Page 5
AI is Everywhere CloudHealthcareIndustrialTransportationConnectivityPCsEntertainmentScience
Page 6
AI is for Everyone ProfessionalsCreatorsStudentsScientistsGamersDevelopers
Page 7
5 BillionPeople Using AI DailyWithin 5 Years 1B+Active AI Users1MActive AI Users 5BActive AI Users
Page 8
10+YottaFLOPS100ZettaFLOPS2025 1ZettaFLOP2022 Yotta Scale ComputeFrom Zetta to10,000xIncrease In AI Compute
Page 9
Unlocking AI’s Potential Requires Massive Compute Everywhere Personal ComputingData Center Edge / Physical AI
Page 10
CloudPersonal ComputingHealthcarePhysical AISpaceScience & Education
Page 11
DevelopIntegrateDeployScale Cloud is the Fastest Path From Idea to Impact
Page 12
Cloud is the Epicenter for AI Development Chatbots & CopilotsPersonal AssistantsGenerative MediaAI Agents & AutomationSmart SearchCoding Assistants
Page 13
Together with our Partners AI Innovators AMD Partners AMD is Enabling the AI Era 310.AI
Page 14
GPT-3.5GPT-4DeepSeek-R1Stable Diffusion 3 GPT-5Llama 4 GPT-3 Training Flops Increasing 4x Every Year QwenKimi K2 AI Model Innovation and Adoption are Accelerating Compute Demand
Page 15
GPT-3.5GPT-4DeepSeek-R1Stable Diffusion 3 GPT-5Llama 4 GPT-3 Training Flops Increasing 4x Every Year QwenKimi K2 AI Model Innovation and Adoption are Accelerating Compute Demand Agentic & Autonomous AI ChatGPT Launch > 1BAI Users Inference Tokens Increased100x in the Last 2 Years
Page 16
The Blueprint for Yotta Scale Infrastructure Rack Scale Efficiency & ServiceabilityOpen Fabricto Scale AIOpen RackArchitectureLeadership Compute Capability ESUN TurnkeySolutions
Page 17
AMD “Helios” AI Rack
Page 19
The Open Rack Platform Towards Yotta Scale AI 43 TB/sScale Out Bandwidth 31 TBHBM4 Memory 4,600 “Zen 6” CPU Cores 18,000GPU Compute Units 2.9 ExaflopsAI Compute 2nm / 3nmAdvanced Process AMD “Helios”
Page 20
AMD INSTINCT™MI355XAMD INSTINCT™MI325XAMD INSTINCT™MI300X
Page 21
A Massive Leap in AIPerformance withMI455X 10x PerformanceIncrease(vs. MI355X) AMD INSTINCT™MI455X
Page 22
“Venice”MI400 Series“Vulcano” 72 GPU Rack Scale“Venice” + MI455X8 GPU Drop-In Ready “Venice” + MI440XHybrid Compute “Venice-X” + MI430X Enterprise AIHyperscale AISovereign AI & HPC PortfolioAMD Data Center
Page 23
Open Ecosystem Drives AI Innovation Open Hardware Open Software Open Ecosystem Scale UpNetwork Scale OutNetwork Infrastructure SoftwareStack Software Ecosystem ESUN
Page 24
Adopted by the Most Popular Developer CommunitiesLocalLLMAIFramework Image & Video GenRL &FinetuningModelHub Open Physical AIInferenceEngines AMD ROCm™ Platform
Page 25
Extending the Leadership Roadmap 2025 AMD Instinct™ MI350 SeriesAMD Instinct™ MI325X 2024 AMD Instinct™MI300A/X 2023 2026 AMD Instinct™MI400 Series 2027 AMD Instinct™MI500 Series
Page 26
AMD INSTINCTMI500Series AMD INSTINCTMI355X AMD INSTINCTMI455X >1000x AI Performance in 4 Years AMD CDNA™ 6 | Advanced 2nm Process | HBM4E Next-Gen MI500 SeriesNext Big Leap in AI Performance
Page 27
Private and Secure Proactively Works for You Performance You Can Depend on Personal to Your Usage AI Experiences are Evolving from Cloud to AI PCs
Page 28
AI PC Experiences Everywhere EnterpriseProductivityAssistedCodingRevolutionaryContent Creation PersonalAssistanceImmersiveCollaboration
Page 29
AMD is at the Forefront of AI PC Innovation Ryzen™ 7000SeriesRyzen™ AI 300Series Firstx86 NPUFirstCopilot+x86 Processor Ryzen™ AI MaxSeries First200B ModelCapable x86 Processor
Page 30
16AMD RDNA™ 3.5 GPU Cores 60 AI TOPSAMD XDNA™ 2 NPU 8,533 MT/s Memory Speed 12 / 24“Zen 5” CPU Cores / Threads 1.3xFaster Multitasking 1.7xFaster Content Creation VS INTEL CORE ULTRA 9 288V Up To Up To AMD Ryzen™AI 400 SeriesPowering Next Generation AI PCsINTRODUCING
Page 31
AMD Ryzen™AI 400Available Starting Q1 2026
Page 32
AMD Ryzen™AI MaxThe PC Reimagined for Creators, Gamers, and AI Developers Content CreationGamingAI Development
Page 33
AMD Ryzen™AI MaxThe PC Reimagined for Creators, Gamers, and AI Developers 40AMD RDNA™ 3.5 GPU Compute Units 50 AI TOPSAMD XDNA™ 2 NPU128 GB Unified Memory 16 / 32“Zen 5” CPU Cores / Threads
Page 34
AMD Ryzen™AI Max+Windows and Linux Support Works with 1000s of Applications 1.4xFaster AI Performancetokens per second 1.8xFaster Content Creation vs Apple Macbook Pro M5 1.7xTokens/sec per DollarLM Studio: GPT-OSS 120B 1.8xTokens/sec per Dollar LM Studio: GLM 4.5 Air vs Nvidia DGX Spark Up To Up To Up To Up To
Page 35
AI Development & Creation UnleashedInfinite Possibilities with AMD Ryzen™ AI Max
Page 36
Linux and Windows Support Preloaded Models Optimized Applications AMD Ryzen™AI Halo ProcessorsAI Developer Platform Available Q2 26INTRODUCING
Page 37
AMD Ryzen™AI Halo ProcessorsAI Developer Platform Available Q2 26INTRODUCING Dev-Ready Apps Optimized:LM Studio, ComfyUI, VS Code, and more Day-0 Support for Leading AI Models Optimized Models Included:GPT-OSS, FLUX.2, SDXL, and more Full AMD ROCm™ Platform Support
Page 38
AMD Powers Over1 Billion Gamers & Creators
Page 39
Diagnostics &ImagingMedical Devices & RoboticsPatient Care Precision Medicine & Drug Discovery Genomic &Computational Biology Advancing Healthcare
Page 40
Advancing Physical AI AerospaceRoboticsHealthcareIndustrialSmart CitiesTransportation High Performance x86 CPUs | Market-Leading Adaptive SoCs
Page 41
Industrial SpaceHealthcare Over 20 Years of Robotics Innovation
Page 42
Decision AI Compute that Picks the Right Answer FastActionDeterministic Control that Moves the System InstantlyPerceptionSensor Fusion Understands the World in Real-Time Physical AIDemands Real-Time Intelligence and Rapid Action
Page 43
Open-SourceEcosystemVision & Sensor ProcessorsLeadership Compute Architectures The Blueprint for Physical AI Architecture
Page 44
Advancing Space Technology
Page 45
Partnering to Further Breakthroughs and Discovery Jet Propulsion Laboratory
Page 46
Faster Innovation Across IndustriesLifesaving VaccinesReach Patients SoonerDevelopment of Cleaner Energyand Sustainable MaterialsCommunities Receive Earlier,More Accurate Warnings AMD is the Compute Foundation for Modern Science
Page 47
Evaluating and Forecasting Geohazards and Natural Disasters Predictive Modeling Under Varying Weather Conditions Time to Generate Climate Forecasts Reduced by 85% 2.7Million Cores380Petaflops
Page 48
Developing Higher Performance Batteries Optimizing Industrial Facilities Improving Accuracy of Dynamic Studies for CO2 Storage 3.1Million Cores478Petaflops
Page 49
ORBIT-2 Delivers Hyper-Resolution Earth System Predictions Using Scalable AI Produces Hyper-Detailed Global Weather Predicting Enabling Actionable Insights Delivers Accuracy in the Range of ~0.98–0.99 When Compared to Observational Data 9Million Cores1.4Exaflops
Page 50
Pre-Emptively Optimized Antibody to Neutralize Broad Diversity of SARS-CoV-2 Variants Anticipates Viral Evolution to Reduce Need for Constant Redevelopment Improves Biodefense Preparedness and Antibody Therapy Resilience 11Million Cores1.8Exaflops
Page 51
First AI Factory for ScienceMulti-Tenant AI Cloud Platform Next-Generation Exascale Supercomputer Accelerating the Speed of Scientific Discovery A New Age of AI Accelerated Innovation and Discovery
Page 52
Advancing Education
Page 53
$150MInvestments
Page 54
Empowering everyoneto shape the future of AI 800+InstitutionsWorldwide
Page 55
Empowering everyoneto shape the future of AI 150,000+Students in 2026 ThroughAMD AI Education Program
Page 56
Empowering AI Learning
Page 57
Emme McDonald Ruzanna GaboyanAfia AvaWinning Team Armtender
Page 58
Open Ecosystem Deep Co-Innovation Unmatched Technology Portfolio AI Everywhere for Everyone
Page 59
DISCLAIMER & ATTRIBUTIONSDISCLAIMER: 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 AMD is under no obligation to update or otherwise correct this information. Advanced Micro Devices, Inc. makes no representations or warranties with respect to the accuracy or completeness of the contents of this document, and assumes no liability of any kind, including the implied warranties of noninfringement, merchantability or fitness for particular purposes, with respect to the operation or use of AMD hardware, software or other products described herein. No license, including implied or arising by estoppel, to any intellectual property rights is granted by this document. Terms and limitations applicable to the purchase or use of AMD products are as set forth in a signed agreement between the parties or in AMD's Standard Terms and Conditions of Sale. GD-18u.© 2025 Advanced Micro Devices, Inc. All rights reserved. AMD, the AMD Arrow logo, AMD CDNA, AMD RDNA, Radeon, Ryzen, Versal, XDNA and combinations thereof are trademarks of Advanced Micro Devices, Inc. Microsoft is a registered trademark of Microsoft Corporation in the US and/or other countries. Other product names used in this publication are for identification purposes only and may be trademarks of their respective owners. Certain AMD technologies may require third-party enablement or activation. Supported features may vary by operating system. Please confirm with the system manufacturer for specific features. No technology or product can be completely secure.
Page 60
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 second that can be executed in an optimal scenario and may not be typical. TOPS may vary based on several factors, including the specific system configuration, AI model, and software version. •GD-250: Based on AMD chip shipment figures, AMD has powered over 1 billion gaming devices with processors and graphics for desktop PCs, notebook PCs, and gaming consoles from 2008 to 2025. •MI350-047B: Based on engineering projections by AMD Performance Labs in September 2025, to estimate the peak theoretical precision performance of seventy-two (72) AMD Instinct™ MI455X GPUs "Helios" AI Rack using MXFP4 dense Matrix datatype vs. an 8xGPU AMD Instinct MI355X platform using the MXFP4 dense Matrix datatype. Results subject to change when products are released in market. •MI500-001: Based on engineering projections by AMD Performance Labs in December 2025, to estimate the peak theoretical precision performance of AMD Instinct™ MI500 Series GPU powered AI Rack vs. an AMD Instinct MI300X platform. Results subject to change when products are released in market. •SHO-42: Testing as of Oct/Nov 2025 by AMD CMPL. The systems tested were configured as follows: Ryzen AI Max+ 395, 55W, Asus Flow Z13, best performance, 8x4GB @8000MHz, AMD Radeon 8060S graphics, VBS=ON, SAM/ReBar=ON INT vs. an Apple M5 10 Core CPU, Apple Macbook Pro 2025 14in, 24GB RAM, Balanced, 10 core GPU, VBS= not supported SAM/Rebar=not supported Int, power=DC. The applications tested include: 7zip, Handbrake, Blender CPU, VRay 6.0, Corona, and LM Studio. System manufacturers may vary configurations yielding different results. Results may vary.
Page 61
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) Driver Version 580.95.05 and CUDA Toolkit Version 13.0 Update 2 for NVIDIA DGX Spark. Flash Attention = ON in all cases. Token/s: sustained performance average of multiple runs with specimen prompt “How long would it take for a ball dropped from 10 meter height to hit the ground?”. Models tested: OpenAI GPT-OSS 120B, OpenAI GPT-OSS 20B, GLM 4.5 Air and DeepSeek R1 Distill Llama 70b. Tokens per second per dollar measured using market pricing of $2500 for the HP Z2 G1a and $4000 for the NVIDIA DGX Spark as of November 2025. AMD Ryzen™ AI Max+ 395 PRO on an HP Z2 Mini G1a with 128GB memory and NVIDIA DGX Spark with 128GB memory. Performance may vary. •SHOP-27: 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. Flash Attention = ON in all cases. MMLU and GPQA scores as-reported from research papers and github repos. Cloud-quality statement from OpenAI “The gpt-oss-120b model achieves near-parity with OpenAI o4-mini on core reasoning benchmarks.” AMD Ryzen™ AI Max+ 395 PRO on an HP Z2 Mini G1a with 128GB memory. 200 billion parameters require 128GB of unified memory. The AMD Ryzen™ AI Max+ was the first x86 processor to launch with 128GB of unified memory. Performance may vary. •GPT-2: Testing done as of November 2025 by AMD to measure multitasking performance by running Procyon Office Suite with Microsoft Teams on “Balanced” power mode. Configuration for AMD Ryzen™ AI 9 HX 470 processor: AMD reference board, Radeon™ 890M integrated graphics, Graphics Driver 25.20.24-251007a, 32GB LPDDR5x-8533 memory, Windows 11 Home. Configuration for Intel Core Ultra 9 288V processor: HP OmniBook Ultra Flip, Arc 140V integrated graphics, Graphics driver 32.0.101.8136, 32GB LPDDR5x-8533 memory, Windows 11 Home. System manufacturers may vary configurations, yielding different results. •GPT-10: Testing done as of December 2025 by AMD to measure content creation performance in the following applications and benchmarks on “Best Performance” power mode: Blender (CPU Classroom), Cinebench nT, Handbrake, PugetBench for Photoshop, PugetBench for DaVinci Resolve Studio, 7zip. Configuration for AMD Ryzen™ AI 9 HX 470 processor (28W): ASUS Zenbook S16, Radeon™ 890M integrated graphics, Graphics Driver 25.20.32-251114n, 32GB LPDDR5x-8533 memory, Windows 11 Pro. Configuration for Intel Core Ultra 9 288V processor (30W): HP OmniBook Ultra Flip, Arc 140V integrated graphics, Graphics driver 32.0.101.7026, 32GB LPDDR5x-8533 memory, Windows 11 Home. System manufacturers may vary configurations, yielding different results.