Slides
Page 2
Matt Ramsay CVP , Financial Strategy and Investor Relations
Page 3
3 3 Cautionary Statement This presentationcontainsforward-looking statements concerning Advanced Micro Devices, Inc. (AMD) such as AMD being positioned to drive accelerated growth and deliver shareholder value; expectations of its financial plans and long-term financial model, including compound annual growth rate, revenue market share, non-GAAP gross margin, non-GAAP operating margin, non-GAAP effective tax rate, non-GAAP FCF margin and non-GAAP earnings per share; expected leadership in high-performance and AI computing; AMD's expected growth and total addressable market; expected customer and market share opportunities; AMD's ability to deliver significant value to its shareholders; and the features, functionality, performance, availability, timing and expected benefits of AMD products, which are madepursuant tothe Safe Harbor provisions of the Private Securities Litigation Reform Act of 1995. Forward-looking statements are commonlyidentifiedby words such as "would," "may," "expects," "believes," "plans," "intends," "projects" and other terms with similarmeaning. Investors are cautioned that the forward-looking statements in this presentation are based on current beliefs,assumptionsand expectations, speak only as of the date of thispresentationand 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 andgenerally beyondAMD'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 herebydisclaims, any obligation to update forward-looking statements made in this presentation, except as may be required by law. Non-GAAP Financial Measures This presentation contains historical non-GAAP financial measures including non-GAAP gross profit and margin, non-GAAP operating profit, and free cash flow. AMD is providing these financial measures because it believes this non-GAAP presentation makes it easier for investorsto compare its operating results for current and historical periodsand alsobecause AMD believes it assists investors in comparing AMD’s performance across reporting periods on a consistent basis by excluding itemsthat it does not believe are indicative of its core operating performance. The non-GAAP financial measures disclosed in this presentation should be viewed in addition to and not as a substitute for or superior toAMD’s reported results prepared in accordance with GAAP and should be read only in conjunction with AMD’s Consolidated Financial Statements prepared in accordance with GAAP. These non-GAAP financialmeasures referenced are reconciled to their most directly comparable GAAP financial measures in the Appendices at the end of this presentation. This presentation also contains forward-looking non-GAAP measuresconcerning AMD’s 2025 financial outlook and long-term financial model such as gross profit or margin, operating income or margin, effective tax rate, free cash flow margin and earnings per share. Projected 2025 non-GAAP gross profit and operating profit exclude $800 million inventory and related charges associated with U.S. export restrictions recognized in Q2’25. For long-term financial model, AMD uses a projected non-GAAP tax rate of 13-15%, which excludes the tax impact of pre-tax non-GAAP adjustments, reflecting currently available information. These forward-looking non-GAAP measures are based on current expectations, assumptions and beliefs that involve numerous risks and uncertainties. Adjustments to arrive at the GAAP financial outlook and long-term financial model typically include stock-based compensation, amortization of acquired intangible assets, income tax provision, and other non-recurring items such as impairment charges and acquisition-related costs. A reconciliation to equivalent GAAP measures is not practicable at this time as the timing and impact of such adjustments are dependent on future events that are typically uncertain or outside of AMD's control. Such events may include unanticipated changes in AMD’s GAAP effective tax rate, unanticipated one-time charges related to asset impairments, unanticipated acquisition-related expenses, unanticipated gains, losses, and impairments, and other unanticipated non-recurring items not reflective of ongoing operations. All statements made in this presentation are based on current expectations as of November 11, 2025, and assumptions and beliefs that involve numerous risks and uncertainties. AMD undertakes no intent or obligation to publicly update or revise its forward-looking statements made in this presentation except as may be required by law.
Page 4
Start Time Agenda Item Presenter 1:00 PM Welcome Matt Ramsay 1:05 PM AMD Vision and Growth Strategy Dr. Lisa Su 1:35 PM AMD Technology Strategy Mark Papermaster 2:05 PM Building on Data Center Leadership Forrest Norrod 2:10 PM Server Leadership Dan McNamara 2:25 PM AI Leadership Across Hardware & Software Vamsi Boppana 2:45 PM Networking & Rack-Scale Systems Forrest Norrod 3:05 PM Break 3:25 PM Client, Graphics & Semi-Custom Momentum Jack Huynh 3:45 PM AMD Embedded Transformation Salil Raje 4:05 PM Accelerating Financial Momentum Jean Hu 4:25 PM Closing Remarks Dr. Lisa Su 4:30 PM Q&A Session 5:15 PM Reception & Demo Showcase AMD Financial Analyst Meeting
Page 5
Lisa Su Chair & Chief Executive Officer
Page 6
2 2 Cautionary Statement This presentation contains forward-looking statements concerning Advanced Micro Devices, Inc. (AMD) including, but not limited to, the features, functionality, performance, availability, timing and expected benefits of future AMD products, including the "Helios" AI rack-scale platform; AMD Instinct MI450 series accelerators; AMD’s projected revenue growth and CAGR across its segments; AMD’s ability to accelerate Data Center growth; AMD's growing semi-custom design wins; AMD's expected benefits of acquisitions and strategic investments; AMD's strong revenue growth and profit expansion; AMD’s Data Center TAM; AMD’s focus areas and strategic pillars; the strategic partnership with OpenAI and the deployment of six-gigawatts of AMD Instinct and timing thereof; the strategic partnership with Oracle and the deployment of 50,000 GPUs and the timing thereof; the expected benefits of AMD AI compute strategic partnerships; AMD’s ability to expand market leadership and accelerating revenue growth; and AMD's ability to deliver compelling shareholder returns, 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. Non-GAAP Financial Measures This presentation contains historical non-GAAP financial measures including non-GAAP gross profit, and non-GAAP operating profit. AMD is providing these financial measures because it believes this non-GAAP presentation makes it easier for investors to compare its operating results for current and historical periods and also because AMD believes it assists investors in comparing AMD’s performance across reporting periods on a consistent basis by excluding items that it does not believe are indicative of its core operating performance. The non-GAAP financial measures disclosed in this presentation should be viewed in addition to and not as a substitute for or superior to AMD’s reported results prepared in accordance with GAAP and should be read only in conjunction with AMD’s Consolidated Financial Statements prepared in accordance with GAAP. These non-GAAP financial measures referenced are reconciled to their most directly comparable GAAP financial measures in the Appendices at the end of this presentation. This presentation also contains forward-looking non-GAAP measures concerning AMD’s 2025 financial outlook and long-term financial model such as gross profit or margin, operating income or margin, and earnings per share. Projected 2025 non-GAAP gross profit and operating profit exclude $800 million inventory and related charges associated with U.S. export restrictions recognized in Q2’25. For long-term financial model, AMD uses a projected non-GAAP tax rate of 13-15%, which excludes the tax impact of pre-tax non-GAAP adjustments, reflecting currently available information. These forward-looking non-GAAP measures are based on current expectations, assumptions and beliefs that involve numerous risks and uncertainties. Adjustments to arrive at the GAAP financial outlook and long-term financial model typically include stock-based compensation, amortization of acquired intangible assets, income tax provision, and other non-recurring items such as impairment charges and acquisition-related costs. A reconciliation to equivalent GAAP measures is not practicable at this time as the timing and impact of such adjustments are dependent on future events that are typically uncertain or outside of AMD's control. Such events may include unanticipated changes in AMD’s GAAP effective tax rate, unanticipated one-time charges related to asset impairments, unanticipated acquisition-related expenses, unanticipated gains, losses, and impairments, and other unanticipated non-recurring items not reflective of ongoing operations. All statements made in this presentation are based on current expectations as of November 11, 2025, and assumptions and beliefs that involve numerous risks and uncertainties. AMD undertakes no intent or obligation to publicly update or revise its forward-looking statements made in this presentation except as may be required by law.
Page 7
High Performance and AI Computing Leadership Solving The World’s Most Important Challenges 3
Page 8
4 4 AMD Computing Powers The World Edge & Intelligent Devices Cloud and Data Center HPC 4 AI Everywhere Manufacturing and Automation Healthcare and Life Sciences Aerospace Communications Automotive and Transportation Science and Discovery Financial Services Consumer and Education Gaming
Page 9
5 5 Our Focus Areas 2022 Strategic Pillars 5 Leading Provider of Custom Silicon & Solutions Expanding Data Center Leadership Software Platforms Advancing Pervasive AI Compute Technology Leadership
Page 10
6 6 Data Center Leadership EPYC Powers the Modern Data Center ▪ 10/10 largest hyperscalers ▪ 10/10 most popular social media platforms ▪ >60% of Fortune® 100 AMD Instinct MI350 Series Fastest Ramping Product in AMD History ▪ 7/10 top AI companies ▪ Numbers 1 and 2 fastest supercomputers ▪ Multi-generation strategic partnerships with OpenAI and OCI Segment Highlights $2B >$16B 2020 2025e Revenue Growth Industry Leaders Run on EPYC CPUs Expanding AMD Instinct GPU Deployments Customer Momentum CAGR: ~52% 2025e includes Q4’25 financial outlook; See Cautionary Statement on Slide 2 See endnotes: 1, EPYC-045D, & EPYC-058 Fortune Magazine and Fortune Media (USA) Corporation are not affiliated with, and do not endorse products or services of Adva nced Micro Devices, Inc.
Page 11
7 7 “Naples” 2017 2018 2019 2020 2021 2022 2023 2024 2025 “Rome”“Milan” “Genoa” “Turin” Accelerating Data Center Growth Record Server CPU Market Share ~40% $10B 2020 ~$34B 2025e Revenue Transformation at Scale Core BusinessData Center 2025e Data Center Revenue >$16B ~47% Total 2020 Data Center Revenue $2B ~17% Total Source: Mercury Research Source: Mercury Research Q2 2025 Server CPU Revenue Share 2025e includes Q4’25 financial outlook. See Cautionary Statement on Slide 2 See endnotes: EPYC-055A
Page 12
8 8 Based on AMD internal projections as of Nov 2025. Actual performance subject to change. AMD Instinct MI325XAMD Instinct MI300X AMD Instinct MI355X AMD Instinct MI450 Series 2025 2026 20242023 Leadership AI Performance with AMD Instinct MI450 Series AI Compute Performance
Page 13
9 9 Client and Gaming Leadership Accelerating Ryzen CPU Share Gains ▪ Record client share of 28% ▪ >50% desktop CPU channel share ▪ 250+ Ryzen AI platforms in market ▪ Deployed by >50% of Fortune® 100 >1 Billion Gaming Devices ▪ 3 generations of leadership consoles ▪ Leadership mainstream gaming GPUs Segment Highlights $8B >$14B 2020 2025e Revenue Growth ~13% CAGR Customer Momentum Broadest Portfolio of AI Consumer & Commercial PCs Growing Fortune 500 Footprint CAGR: ~13% 2025e includes Q4’25 financial outlook. See Cautionary Statement on Slide 2 See endnote(s): GD-250 Fortune Magazine and Fortune Media (USA) Corporation are not affiliated with, and do not endorse products or services of Adva nced Micro Devices, Inc.
Page 14
10 10 Embedded Leadership #1 Adaptive Computing Provider ▪ 10/10 largest Aerospace & Defense ▪ 10/10 largest Automotive ▪ 10/10 largest Medical Imaging ▪ 6/7 largest 6G Wireless equipment ▪ 5/5 largest commercial satellite systems 3rd Straight Year of Record Design Wins ▪ >$50B design wins since Xilinx acquisition Segment Highlights Broad & Diverse Set of 7000+ Customers Customer Momentum $9B $12B $14B >$16B 2022 2023 2024 2025e Design Win Growth CAGR: ~21% Based on AMD internal data as of Nov. 2025
Page 15
11 11 ~$20B Gaming Expected lifetime revenue of design wins >$45B Aerospace & Defense Automotive Data Center Gaming Wireless 2025 Design Wins LTR >125% Growth 2020 Design Wins LTR Expanding Set of Customers Across Multiple Markets Growing Semi-Custom Design Wins See Cautionary Statement on Slide 2
Page 16
12 12 Organic Investments 2020 2025e >$40B ~$60B Acquisitions ~$1B Investments >$100B Invested in Last 5 Years Strategic Investments Driving AI Growth 2025e includes Q4’25 financial outlook. See Cautionary Statement on Slide 2
Page 17
13 13 $1B $4B ~$18B 2016 2020 2025e $4B $10B ~$34B 2016 2020 2025e $43M $2B ~$8B 2016 2020 2025e Gross Profit Revenue Operating Profit CAGR: 34% CAGR: 79%CAGR: 26% Strong Revenue Growth and Profit Expansion1 (1) Non-GAAP financial measures, except for Revenue. See appendices for GAAP to non-GAAP reconciliation 2025e includes Q4’25 financial outlook and excludes $800M inventory and related charges due to the U.S export restrictions. S ee Cautionary Statement on Slide 2 Not drawn to scale; Figures may be impacted by rounding
Page 18
14 14 AI Is Redefining Every Market Massive Infrastructure Buildout ▪ Insatiable demand for AI compute ▪ Unprecedented investments in power ▪ Models growing larger and more diverse ▪ AI driving growing demand for CPU compute AI Everywhere ▪ Extending to edge, end points and PCs ▪ Transformative user experiences ▪ Heterogenous platforms combining CPUs, GPUs, FPGAs and NPUs ▪ Breakthroughs in robotics and physical AI Rapidly Expanding Adoption ▪ Explosion of agentic workloads ▪ Inference growing faster than training ▪ Enterprise-class solutions ▪ Sovereign and nation-scale deployments
Page 19
15 15 Our Largest Growth Opportunity AMD Data Center TAM >$1T 2030 ~$200B 2025 CAGR: >40% Source: AMD Internal Projections
Page 20
16 16 Powering AI and High-Performance Compute Everywhere Strategic Pillars High-Performance, Adaptive and Custom AI Platforms Powering AI Everywhere Cloud, Enterprise and Rack-Scale Solutions Expand Data Center Leadership Production-Grade AMD ROCm Software Stack Open Software Platforms & Developer Enablement CPU, GPU, FPGA, Networking, System-on-Chip, Chiplets & Packaging Extend Compute Technology Leadership
Page 21
17 17 Leadership Product Portfolio Unmatched Technology Leadership GPUs FPGAs, NPUs and Adaptive SoCs SmartNICs and DPUs CPUs Foundational IP Chiplets, SerDes, Advanced Packaging, Infinity Fabric Strategic Pillars
Page 22
18 18 Accelerating Data Center Leadership Strategic Pillars Cloud Optimized Solutions for AI and General-Purpose Compute AI Solutions Rack and Cluster Designs at Data Center Scale Networking Open Solutions with Leadership Scale- Up/Scale-Out Technology Enterprise Broad Portfolio of Enterprise Server and AI Solutions Supercomputing Industry-Leading Performance for Scientific and AI Computing See endnote: EPYC-045D
Page 23
19 19 Industry-Leading Software Stacks Strategic Pillars Unified Software Ecosystem Enterprise-Optimized Solutions Industry’s Premier Open-Source AI Stack Empowering an Expanded Set of Developers Native Support in Largest Open-Source Projects and Day-0 Models
Page 24
20 20 Delivering AI Factories DPUs, UALink & Ultra Ethernet Networking Cluster Level Systems Design AMD ROCm Software GPUs AMD Instinct CPUs AMD EPYC ESUN
Page 25
21 21 Powering AI Everywhere Strategic Pillars Physical AI Semi-Custom Solutions Gaming Powering Emerging AI Experiences AI PCs Edge AI
Page 26
22 22 Frontier AI Development 6GW AMD Instinct Accelerators Starting in 2H’2026 Zetta-Scale Compute 50,000 AMD Instinct MI450 Series GPUs Starting in 2H’2026 Co-Designed Open Infrastructure AMD “Helios” Rack Co-defined with Meta and Previewed at OCP Strategic Partnerships Driving An Inflection in Growth Extending US HPC Leadership Lux: First US AI Factory AMD Instinct MI355X Series Discovery: Flagship AI Supercomputer AMD Instinct MI430X Series
Page 27
23 23 Data Center AI Revenue Opportunity Next 3 – 5 Years >80% Data Center AI Revenue CAGR Broad Customer Traction Multiple Hyperscaler, AI Native & Sovereign Opportunities MI450 Series & “Helios” Rack-Scale Solutions Strong Momentum with Multiple GW-Scale Opportunities Revenue Growth On-Track to Tens of Billions of Dollarsin 2027 From 2025 baseline which includes Q4’25 financial outlook. See Cautionary Statement on Slide 2
Page 28
24 24 >40% Client Revenue Market Share >70% Adaptive Revenue Market Share >50% Server Revenue Market Share >80% Data Center AI Revenue CAGR AMD Expanding Market Leadership Next 3 – 5 Years From 2025 baseline which includes Q4’25 financial outlook. See Cautionary Statement on Slide 2
Page 29
25 25 25 Accelerating Revenue Growth Next 3 – 5 Years AMD Total >35% CAGR Core Business >10% CAGR Data Center Business >60% CAGR From 2025 baseline which includes Q4’25 financial outlook. See Cautionary Statement on Slide 2
Page 30
26 26 The Path to Market Leadership Deliver Compelling Shareholder Returns Financial Analyst Day 2025 Broadest Product Portfolio World-Class Execution Deep Strategic Partnerships Large and Growing >$1T TAM Unmatched Data Center Solutions
Page 31
27 27 Disclaimer & Attribution 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 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 Instinct, EPYC, Pensando, Radeon, ROCm, Ryzen, Versal, Xilinx, and combinations thereof are trademarks of Advanced Micro Devices, Inc. CXL is a registered trademark of Compute Express Link Consortium, Inc. OpenAI is a trademark of OpenAI, Inc. PCIe® is a registered trademark of PCI-SIG Corporation. UCIE is a trademark of Universal Chiplet Interconnect Express, Inc. Ultra Accelerator Link and UALink are trademarks of the UALink Consortium. 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. Fortune content is From Fortune Magazine. © 2025 Fortune Media IP Limited. All rights reserved. Used under license. Fortune is a registered trademark of Fortune Media IP Limited and is used under license.
Page 32
28 28 Endnotes • 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. • Endnote #1: Top 10 Social Media platforms (based on active users) and Top 10 Cloud Native SaaS Services (based on Market Cap). • EPYC-045D - Top500 list as of June, 2025 (https://top500.org/lists/top500/2025/06/). • EPYC-055A - Mercury Research Sell-In Revenue Shipment Estimates, Q2 2025. Revenue share of 41.0%, unit share of 27.3%. • EPYC-058: Top 100 U.S. companies by revenue according to 2025 Fortune 500 list as of June 2, 2025. https://fortune.com/ranking/fortune500/. 'Fortune 100’ refers to the top 20% ranked companies in the 2025 Fortune 500 list, published in June 2025. From Fortune Magazine. ©2025 Fortune Media IP Limited. All rights reserved. Used under license. Fortune and Fortune Media IP Limited are not affiliated with, and do not endorse products or services of Advanced Micro Devices, Inc.
Page 33
29 APPENDIX
Page 34
30 APPENDICES $ in millions, except % (Unaudited) 2016 2020 YTD Q3 2025 GAAP gross profit $ 1,003 $ 4,347 $ 11,575 GAAP gross margin 23% 45% 47% Stock-based compensation 2 6 18 Amortization of acquisition-related intangibles - - 771 Acquisition-related and other costs (1) - - 1 Inventory loss at (recovery from) contract manufacturer (2) - - (67) Loss contingency on legal matter - - 12 Charge related to the sixth amendment to the WSA with GF (3) 340 - - Non-GAAP gross profit $ 1,345 $ 4,353 $ 12,310 Non-GAAP gross margin 31% 45% 51% Inventory and related charges associated with U.S. export restrictions - - 800 Non-GAAP gross profit (as adjusted to exclude inventory and related charges associated with U.S. export restrictions) $ 1,345 $ 4,353 $ 13,110 Non-GAAP gross margin (as adjusted to exclude inventory and related charges associated with U.S. export restrictions) 31% 45% 54% Reconciliation of GAAP to Non-GAAP Gross Profit and Gross Margin (1) Acquisition-related and other costs primarily include transaction costs, purchase price fair value adjustments for inventory, certain compensation charges, and workforce rebalancing charges. (2) Inventory loss at (recovery from) contract manufacturer is related to losses due to an incident at a third-party contract manufacturing facility in Q1’24 and the corresponding recovery. (3) Charge related to the sixth amendment to the Wafer Supply Agreement with GlobalFoundries
Page 35
31 APPENDICES $ in millions, except % (Unaudited) 2016 2020 YTD Q3’25 GAAP operating income (loss) $ (373) $ 1,369 $ 1,942 GAAP operating margin (9)% 14% 8% Stock-based compensation 86 274 1,152 Amortization of acquisition-related intangibles - - 1,697 Acquisition-related and other costs (1) - 14 178 Inventory loss at (recovery from) contract manufacturer (2) - - (67) Loss contingency on legal matter - - 12 Restructuring charges (3) (10) - - Charge related to the sixth amendment to the WSA with GF (4) 340 - - Non-GAAP operating income $ 43 $ 1,657 $ 4,914 Non-GAAP operating margin 1% 17% 20% Inventory and related charges associated with U.S. export restrictions - - 800 Adjusted Non-GAAP operating income (excluding inventory and related charges associated with U.S. export restrictions) $ 43 $ 1,657 $ 5,714 Adjusted Non-GAAP operating margin (exclude inventory and related charges associated with U.S. export restrictions) 1% 17% 23% Reconciliation of GAAP Operating Income (Loss) to Non-GAAP Operating Income (1) Acquisition-related and other costs primarily include transaction costs, purchase price fair value adjustments for inventory, certain compensation charges, and workforce rebalancing charges. (2) Inventory loss at (recovery from) contract manufacturer is related to losses due to an incident at a third-party contract manufacturing facility in Q1’24 and the corresponding recovery. (3) Restructuring charges are related to the 2015 and 2024 Restructuring Plan which comprised of employee severance charges and non-cash asset impairments. (4) Charge related to the sixth amendment to the Wafer Supply Agreement with GlobalFoundries
Page 36
Mark Papermaster EVP & CTO
Page 37
2 2 Cautionary Statement This presentationcontainsforward-looking statements concerning Advanced Micro Devices, Inc. (AMD) such as the features, functionality, performance, availability, timing and expected benefits of AMD products, including "Helios"rack-scale platform; leadership CPU core roadmap; data center GPU architecture roadmap; AMD's ability to accelerate GPU performance gains; high-speed SerDes interconnect; chiplet and packaging leadership; AMD Infinity Fabric roadmap; Gaming GPU architecture roadmap; NPU inference IP Roadmap or client and embedded; AMD’s opportunities in AI; and AMD's focus areas, which are madepursuant tothe Safe Harbor provisions of the Private Securities Litigation Reform Act of 1995. Forward-looking statements are commonlyidentifiedby words such as "would," "may,” "expects," "believes," "plans," "intends," "projects" and other terms with similarmeaning. Investors are cautioned that the forward-looking statements in this presentation are based on current beliefs,assumptionsand expectations, speak only as of the date of thispresentationand 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 andgenerally beyondAMD'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 herebydisclaims, any obligation to update forward-looking statements made in this presentation, except as may be required by law.
Page 38
3 3 Product Portfolio Driving Revenue Growth Common Foundational IP Leadership Core IP CPU DPU FPGA NPU GPU Data Center Client & Gaming Embedded Industry’s Broadest Solution Portfolio Infinity Fabric Interconnect System-on- a-Chip Chiplets & Packaging End-to-End Product Security Open-Source Software
Page 39
4 4 Product Portfolio Driving Revenue Growth Common Foundational IP Leadership Core IP CPU DPU FPGA NPU GPU Infinity Fabric Interconnect System-on- a-Chip Chiplets & Packaging End-to-End Product Security Open-Source Software Data Center Client & Gaming Embedded Semi-Custom Industry’s Broadest Solution Portfolio
Page 40
5 5 Financial Analyst Day • 2015 We Called the Play
Page 41
6 6 We Delivered Industry Leadership “Zen” x86 CPU ▪ Consistent Leadership Performance ▪ Chiplet Enabled Core Density Leadership ▪ High-Performance & Energy Optimized AMD Instinct GPU with Leadership Inference Performance ▪ 2.5D Integrated HBM & 3D Chiplets Drove Memory & Compute Density Leadership ▪ TCO Advantages for Key AI Workloads Breakthrough SoC Approach ▪ Scalable Infinity Fabric Enabled Modular Architecture ▪ First Chiplet Design Methodology
Page 42
7 ▪ Compute Leadership Roadmap from Laptops to Rack-Scale ▪ Driving a Truly Open Software & Hardware AI Ecosystem ▪ SerDes, Optical & Networking Investments ▪ Confidential Computing for Trusted AI Processing We Are Accelerating Optimizing AI Across Portfolio 7
Page 43
8 Leadership CPU Core Roadmap *All roadmaps subject to change. 2022 2026 New AI Data Type Support More AI Pipelines Industry-First 2nm Extension of Leadership “Zen 6” “Zen 6c” New Matrix Engine AI Data Format Expansion Future Node Next-Generation “Zen 7” Wider/Deeper Compute Full 512 Bit AI Vectors Re-optimized Cache Hierarchy 4nm • 3nm New Foundation “Zen 5” “Zen 5c” AVX-512 AI ISA Support Dual Core Design Points Larger L2 Cache 5nm • 4nm Performance & Density “Zen 4” “Zen 4c”
Page 44
9 AMD Instinct : Data Center GPU Architecture Roadmap *All roadmaps subject to change. 2023 2026 Increased HBM3e Capacity & Bandwidth Block Scaled AI Formats Optimized Platform Solutions MI350 Series Gen AI Inferencing & Training Increased HBM3e Capacity & Bandwidth Higher Compute Throughput Platform Compatibility MI325 Gen AI Performance Leadership 8-Stack HBM3 Capacity & Bandwidth 3.5D Chiplet Construction Industry Standard Subsystem MI300 AI-Centric Design Increased HBM4 Capacity & Bandwidth Expanded AI Formats with Higher Throughput Standard-Based Rack-Scale Networking (UALoE, UAL, UEC) MI400 Series Rack-Scale Architecture Next-Gen Compute, Memory & Interconnect MI500 Series
Page 45
10 0x 20x 40x 60x 80x 100x 120x 140x 2x/1 year Trend FP8 FP4 Accelerated GPU Performance Gains 2022 2023 2024 2025 2026 2027 2028 Pre-2025 Trend ▪ 2x Per Two Year Performance Trend Pre-2025 (16b Math) ▪ 2x Per One Year Performance Trend Commences in 2025 ▪ AI-Optimized Low Precision Math (FP8 & FP4) Provide Even Higher Gains FLOP Rates *Based on AMD internal data.
Page 46
11 High-Speed SerDes Interconnect *All roadmaps subject to change. 2022 2026 First Server PCIe® 5.0 10s of Millions in Production 40 Gb/s First 112G Ethernet FPGA Leading Edge PCIe 6.0 PAM4 112 Gb/s Leading Edge PCIe 7.0 PAM4 2026 “Helios” Rack Interconnect 224 Gb/s Rack Transition Point Copper & Optical Solutions Next Gen
Page 47
12 *All roadmaps subject to change. Up to 8-Way GPU Connectivity Coherent Memory for Exascale Computing PCIe® 5.0 3rd Generation Heterogeneous Chiplet CXL 2.0 Memory Expansion Support for Neural Engine 4th Generation Scale Up & Scale Out Leadership Unified System Level Coherency Standard Based Open Ecosystem (UALink , CXL 3.1 & UCIe ) PCIe 6.0 Introducing 5th Generation Continued Scale Next-Gen 2022 2026 AMD Infinity Fabric Roadmap
Page 48
13 MI350 3.5D/CoWoS-S + HBM3e MI100 2.5D + HBM2Multichip 13 30x Compute & Memory Area MI250 2.5D + HBM2e MI400 3.5D/CoWoS-L + HBM4 Chiplet & Packaging Leadership 2017 2026
Page 49
14 Solution Example “Helios” Rack Hardware-Software-Application Co-Optimization ▪ AMD ROCm Inference & Training Scale with Compute & Communication ▪ Day-0 Optimization of Most Popular Models & Frameworks Link at Rack-Scale – Scale Up & Scale Out ▪ AMD Infinity Fabric for CPU to GPU link ▪ UALink Over Ethernet for Scale-Up ▪ UALink for GPU to AI Network Interface Card (NIC) ▪ AI NIC for Scale Out with Ultra Ethernet (UEC)
Page 50
15 15
Page 51
16 Trusted Computing ▪ Top Priority for IT Administrators: Protecting AI Models, Data & Weights ▪ Solution Enables End-to-End Encryption ▪ Confidential Computing Including Memory, Storage & I/O Devices Accelerators SmartNICsStorage Trusted Boundaries Trusted IO TDISP-Based Endpoints TDISP-Based Endpoints TDISP-Based Endpoints
Page 52
17 *All roadmaps subject to change. 2nd Gen AI Dedicated AI Optimized Pipelines Advanced Raytracing Radeon RX 9000 Series Next-Gen AI & Raytracing Next-Gen 1st Gen AI Wave Matrix Multiply Accumulate Optimized Ray Accelerators Radeon RX 7000 Series AI Foundation Dot Product Acceleration Raytracing Radeon RX 6000 Series 2021 2025 Gaming GPU Architecture Roadmap
Page 53
18 NPU Inference IP Roadmap for Client & Embedded *All roadmaps subject to change. 2023 2026 Better Scale & Power Efficiency Next-Generation AI-Optimized FP Formats RyzenAI & Versal Next Block Floating Point Support Ryzen AI & Versal AI Edge Gen 2 Industry’s First x86 AI PC Ryzen AI & Versal AI Edge Series AI Engine 1st Gen AI Engine 2nd Gen AI Engine 3rd Gen
Page 54
19 AI Beyond the Data Center One Software Stack Enables AI from Cloud to Edge Train Once • Infer Everywhere Libraries Compilers & Tools Runtime AI Models & Algorithms AMD Instinct EPYC Radeon AI Ryzen AI Embedded
Page 55
20 Investments for the Future Agentic AI Leadership CPUs to Accelerate Agents Edge AI Explosion Integrated CPU, GPU & NPU Physical AI Leadership Embedded FPGA, CPU & GPU Quantum Computing Hybrid Quantum with AMD Compute Cluster Data Center AI Compute Leadership Rack-Scale GPU & Open Ecosystem
Page 56
21 21 Innovation for the AI Era Product Leadership Broadest IP Portfolio Open Ecosystems Relentless Execution
Page 57
22 22 Disclaimer & Attribution 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 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 Instinct, EPYC, Pensando, Radeon, ROCm, Ryzen, Versal, Xilinx, and combinations thereof are trademarks of Advanced Micro Devices, Inc. CXL is a registered trademark of Compute Express Link Consortium, Inc. OpenAI is a trademark of OpenAI, Inc. PCIe® is a registered trademark of PCI-SIG Corporation. UCIE is a trademark of Universal Chiplet Interconnect Express, Inc. Ultra Accelerator Link and UALink are trademarks of the UALink Consortium. 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 58
Forrest Norrod EVP & GM, Data Center
Page 59
2 2 Cautionary Statement This presentationcontainsforward-looking statements concerning Advanced Micro Devices, Inc. (AMD) such as the features, functionality, performance, availability, timing and expected benefits of AMD products including "Helios" rack-scale platform; AMD server CPU revenue market share; AMD's ability to advance AI infrastructure on an annual cadence; AMD's Data Center AI opportunities; AMD's ability to accelerate Data Center growth; and AMD's Data Center projected revenue growth and CAGR, which are madepursuant tothe Safe Harbor provisions of the Private Securities Litigation Reform Act of 1995. Forward-looking statements are commonlyidentifiedby words such as "would," "may," "expects," "believes," "plans," "intends," "projects" and other terms with similarmeaning. Investors are cautioned that the forward-looking statements in this presentation are based on current beliefs,assumptionsand expectations, speak only as of the date of this presentationand 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 andgenerally beyondAMD'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 herebydisclaims, any obligation to update forward-looking statements made in this presentation, except as may be required by law.
Page 60
3 3 3 Data Center Leadership Our Mission
Page 61
4 Trends That Shaped EPYC ▪ Cloud-scale & virtualized infrastructure ▪ Data explosion ▪ Disaggregated & heterogenous compute
Page 62
5 5 “Turin” 5th Gen “Genoa” 4th Gen “Milan” 3rd Gen “Rome” 2nd Gen “Naples” 1st Gen AMD EPYC Server CPU Leadership Consistent Delivery of Industry-Leading Performance & Efficiency
Page 63
6 6 Server CPU Revenue Market Share 1st Gen “Naples” 2nd Gen “Rome” 3rd Gen “Milan” 4th Gen “Genoa” 5th Gen “Turin” ~40% AMD EPYC Server CPU Leadership Consistent Delivery of Industry-Leading Performance & Efficiency 2017 2018 2019 2020 2021 2022 2023 2024 2025 Source: Mercury Research Q2 2025 Server CPU Revenue Share See endnote(s): EPYC-055A
Page 64
7 7 AI is Redefining the Data Center Compute Diversification Unprecedented Scale AI Factories
Page 65
8 General-Purpose Compute Racks GPU Racks Storage Racks 8
Page 66
9 9 Delivering AI Factories DPUs, UALink & Ultra Ethernet Networking Cluster Level Systems Design AMD ROCm Software GPUs AMD Instinct CPUs AMD EPYC ESUN
Page 67
10 10 Disclaimer & Attribution 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 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 Instinct, EPYC, Pensando, Radeon, ROCm, Ryzen, Versal, Xilinx, and combinations thereof are trademarks of Advanced Micro Devices, Inc. CXL is a registered trademark of Compute Express Link Consortium, Inc. OpenAI is a trademark of OpenAI, Inc. PCIe® is a registered trademark of PCI-SIG Corporation. UCIE is a trademark of Universal Chiplet Interconnect Express, Inc. Ultra Accelerator Link and UALink are trademarks of the UALink Consortium. 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 68
11 11 Endnotes • EPYC-055A: Mercury Research Sell-In Revenue Shipment Estimates, Q2 2025. Revenue share of 41.0%, unit share of 27.3%.
Page 69
Dan McNamara SVP & GM, Compute & Enterprise AI
Page 70
2 2 Cautionary Statement This presentationcontainsforward-looking statements concerning Advanced Micro Devices, Inc. (AMD) such as the features, functionality, performance, availability, timing and expected benefits of AMD products; AMD's ability to gain server market share; AMD EPYC CPU roadmap; AI driving additional CPU demand; AMD's data center CPU revenue growth and CAGR; and AMD's ability to accelerate server leadership,which are madepursuant tothe Safe Harbor provisions of the Private Securities Litigation Reform Act of 1995. Forward-looking statements are commonlyidentifiedby words such as "would," "may," "expects," "believes," "plans," "intends," "projects" and other terms with similarmeaning. Investors are cautioned that the forward-looking statements in this presentation are based on current beliefs,assumptionsand expectations, speak only as of the date of thispresentationand 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 beyondAMD'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 herebydisclaims, any obligation to update forward-looking statements made in this presentation, exceptas may be required by law.
Page 71
3 3 Accelerating Our Path To #1 Server Market Share Relentless Execution Customer Focus Product Leadership
Page 72
4 4 EPYC CPU: The Foundation of Next Generation Compute Optimized Performance & TCO for AI Use Cases *See endnotes 9xx5C-041, 9xx5-261, and 9xx5-262 Enterprise Highest Performance Per Core Modernization & Consolidation Cloud Highest vCPU Density Generational TCO Gain HPC Highest FLOPS Per Node Faster Solutions to Complex Problems
Page 73
5 5 2019 2026 Leadership in Density, Performance & Efficiency Customer Centricity Drives Innovation Roadmap subject to change. Next-Generation “Venice” Next-Gen Leadership Broad AI Use Cases Increased Efficiency Optimized Design Points Cloud Native Edge & Telco 4th Generation “Genoa” | “Bergamo” 3rd Generation “Milan” Core Performance Leadership Enterprise Technical Compute 5th Generation “Turin” Classic & Dense AI Optimization AI Inference & Host Node 2nd Generation “Rome” >2x Thread Density Cloud Computing HPC
Page 74
6 6 General IT Migration Lowest TCO >1.5x Performance/$ Specialty Compute Faster Time to Results >2.7x Simulation Performance Mission Critical Leadership Performance >1.7x VM Performance *See endnote 9xx5C-012, 9xx5C-046 Cloud: Leading Across Applications Enabling Secure Hybrid & Multi-Cloud Deployments
Page 75
7 7 EPYC CPUs Connect Billions of Users Daily 7 *See endnote 1 10/10 Largest Cloud Native SaaS Services 10/10 Top 10 Social Media Platforms >3x Y/Y Public Cloud VM Adoption
Page 76
8 8 *5 Years.See endnotes 9xx5-142, 9xx5TCO-005 Enterprise Ecosystem Centers Around EPYC CPUs Optimized Platforms >170 Platforms Core IT ~80% Lower TCO* Industry Workloads Accelerate Time to Results by 1.8x
Page 77
9 9 Enterprise Adoption Accelerating >2x New Customers YTD Landing New Customers >2x Consumption Growth Expanding FootprintFortune® 100 Adoption >60% Customers See endnote: EPYC-058 Fortune Magazine and Fortune Media (USA) Corporation are not affiliated with, and do not endorse products or services of Advanced Micro Devices, Inc.
Page 78
10 10 Continued Exascale Leadership Optimized AI & HPC Data Types AI to Accelerate Scientific Discovery Double Precision Floating Point Established Exascale HPC Leader Molecular Dynamics Genomics Oil & Gas Training LLMs Extreme Weather Prediction National Security Oak Ridge National Laboratory and the oak leaf symbol are registered trademarks of the U.S. Department of Energy. Use of this mark does not constitute or imply its endorsement, recommendation, or favoring by the United States Government or any agency thereof or its contractors or subcontractors.
Page 79
11 11 AI Requires Leadership CPUs Head Node CPU Unlocks Training & Inference Cluster Performance AI Ready Data Center Modernization for AI Driven Computing Processing & Storage Optimizing Data Pipelines CPU Inference CPU Inference Where Data Lives Agentic Actions Agentic Actions Drive Data Processing
Page 80
12 1212 $19B $19B $20B $26B 2022 2023 2024 2025 2026 2027 2028 2029 AI Will Create an Additional ~$30B of CPU Demand by 2030 AI Inflection Drives Server CPU Consumption Source: AMD Internal Projection Data Center CPU TAM AI-Driven General Compute 2030 ~$60B CAGR: 4% CAGR: ~18%
Page 81
13 13 >1.3x Thread Density >1.7x Performance & Efficiency Next Generation “Venice” Delivering the Next Generation of Performance With “Venice” The World’s Best Gets Even Better *See Endnote VEN-001A. Performance/watt efficiency based on SPECrate®2017_int-base performance projection/processor TDP.
Page 82
14 14 Accelerating Server Leadership Leadership in AI Growth Accelerated Enterprise Adoption Sustained Cloud Expansion Clear Path to >50% Market Share
Page 83
15 15 Disclaimer & Attribution 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 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 Instinct, EPYC, Pensando, Radeon, ROCm, Ryzen, Versal, Xilinx, and combinations thereof are trademarks of Advanced Micro Devices, Inc. CXL is a registered trademark of Compute Express Link Consortium, Inc. OpenAI is a trademark of OpenAI, Inc. PCIe® is a registered trademark of PCI-SIG Corporation. UCIE is a trademark of Universal Chiplet Interconnect Express, Inc. Ultra Accelerator Link and UALink are trademarks of the UALink Consortium. 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. Fortune content is From Fortune Magazine. © 2025 Fortune Media IP Limited. All rights reserved. Used under license. Fortune is a registered trademark of Fortune Media IP Limited and is used under license.
Page 84
16 16 Endnotes • Endnote #1: Top 10 Social Media platforms (based on active users) and Top 10 Cloud Native SaaS Services (based on Market Cap) • 9xx5c-012: Testing by AMD Performance Labs as of 3/12/2025. H4D highmem-192 vCPU instance with Local SSD provides an average ~2.70% performance uplift versus H3 standard 88 vCPU instance running industry standard benchmarks on single-instance including Ansys® Fluent®, StarCCM+ , OpenFOAM®, Altair® Radioss®, Ansys® LS-DYNA®, Ansys® Mechanical , WRF®, and GROMACS. H4D and H3 are both SMT off VMs. Local SSD is not supported on H3. Average (mean) of 3 runs for each benchmark on each VM type: H3H4DUplift Fluent aircraft_wing_14mSolver Rating (Higher is Better)1752.303495.001.99 Fluent exhaust_system_33mSolver Rating (Higher is Better)644.871208.671.87 StarCCM+ hlmach10Avg Elapsed Time (Lower is Better)1.520.532.88 StarCCM+ lemans_segregatedAvg Elapsed Time (Lower is Better)7.692.852.70 OpenFOAM 1004040Elapsed Time (Lower is better)997.99484.962.06 OpenFOAM 1084646Elapsed Time (Lower is better)1409.63615.062.29 OpenFOAM 1305252Elapsed Time (Lower is better)2163.55955.322.26 Radioss neonElapsed Time (Lower is better)1583.52437.813.62 Radiosst 10mElapsed Time (Lower is better)8389.523082.642.72 LSDYNA 3carsElapsed Time (Lower is better)732.33193.333.79 LSDYNA car2carElapsed Time (Lower is better)6495.002541.672.56 LSDYNA odb10mElapsed Time (Lower is better)60246.6722692.332.65 Mechanical GearBoxSolver Rating (Higher is Better)361.92905.982.50 Mechanical SemiSubmersibleSolver Rating (Higher is Better)378.51887.682.35 WRF conus2.5kmElapsed Time (Lower is better)3.351.113.03 GROMACS benchPEPNS per day (Higher is Better)1.123.763.35 GROMACS lignocellulose-rfNS per day (Higher is Better)6.8922.323.24 Average (mean) of all workload benchmark uplifts: 2.70 Cloud performance results presented are based on the test date in the configuration. Results may vary due to changes to the underlying configuration, and other conditions such as the placement of the VM and its resources, optimizations by the cloud service provider, accessed cloud regions, co-tenants, and the types of other workloads exercised at the same time on the system. • 9xx5c-041: Testing by AMD Performance Labs as of 2/21/2025. C4D-standard-16 score comparison to C3D-standard-16 estimated SPECrate®2017_int_base benchmark scores Performance differences (normalized to C3D): C4D 16 vCPU vs C3D 16 vCPU Perf (est) Perf/$ (est) OpEx Savings SPECrate®2017 Integer 1.271.24~19.6% On-demand hourly pricing from https://cloud.google.com/compute/vm-instance-pricing (us-central1) as of 8/1/2025: C4D: $0.740, C3D $0.726 Cloud performance results presented are based on the test date in the configuration. Results may vary due to changes to the underlying configuration, and other conditions such as the placement of the VM and its resources, optimizations by the cloud service provider, accessed cloud regions, co-tenants, and the types of other workloads exercised at the same time on the system. SPEC®, SPECrate® and SPECint® are registered trademarks of Standard Performance Evaluation Corporation. Learn more at spec.org.
Page 85
17 17 Endnotes • 9xx5c-046: Testing by AMD Performance Labs as of 10/25/2025. M8a.4xlarge score comparison to M8i-4xlarge running the following benchmarks: - FFmpeg v6.1.1 (average of 2x encode and 2x transcode) - Redis 7.2 (average of get and set) - NGINX 1.1.9-2, WRK 4.2 - MySQL HammerDB TPROC-C 4.2 - SQL Server (TPC_E derivative nd TPC_H) Performance differences (normalized to M8a): M8A.4XL 16 vCPU vs M8i.4XL 16 vCPU Perf Perf/$ OpEx Saving FFmpeg Average 1.90 1.66 40.0% Redis Average 2.02 1.75 42.0% NGINX 2.14 1.86 46.0% MySQL TPROC-C 1.53 1.33 25.0% SQL Server TPC_E derivative 1.35 1.18 15% SQL Server TPC_H 1.47 1.28 22% Top 6 Workload Benchmarks 1.73 1.51 .31.7. 0% On-demand hourly pricing from https://aws.amazon.com/ec2/pricing/on-demand/ (us-east) as of 10/25/2025: M8A.4XL: $0.973, M8i.4XL $0.846 Cloud performance results presented are based on the test date in the configuration. Results may vary due to changes to the underlying configuration, and other conditions such as the placement of the VM and its resources, optimizations by the cloud service provider, accessed cloud regions, co-tenants, and the types of other workloads exercised at the same time on the system. Cloud OpEx savings calculated from relative runtime and instance cost ratio. The HammerDB TPROC-C for MySQL workload is an open-source workload derived from TPC-Benchmark Standard, and as such is not comparable to published TPC-C results, as the results do not comply with the TPC-C Benchmark Standard .. TPC, TPC Benchmark, TPC-C are trademarks of the Transaction Processing Performance Council. The SQL Server TPC-E derivative is derived from TPC-E Benchmark standard and as such is not comparable to published TPC-E results as the results do not compy wih the TPC-E Benchmark standard. • 9xx5-142: AMD testing as of 04/02/2025. The detailed results show the average uplift of the performance metric (ns/day) of this benchmark for a 2P 192-Core AMD EPYC 9965 powered reference system compared to a 2P Intel® Xeon® 6980P powered production system running select tests on Open-Source GROMACS 2023.1. Uplifts for the performance metric normalized to the Intel® Xeon® 6980P follow for each benchmark: * gmx_water1536K_PME: ~1.76x * benchPEP: ~1.83x System Configurations CPU: 2P Intel® Xeon® 6980P (256 total cores) Memory: 24x 96 GB DDR5-6400 Storage: SOLIDIGM SBFPF2BU076T 7.68 TB NVMe Platform and BIOS: BIOS Options: SMT=OFF, SNC=3, HPC Workload Profile OS: rhel 9.4 5.14.0-427.16.1.el9_4.x86_64 Kernel Options: BOOT_IMAGE=(hd1,gpt2)/vmlinuz-5.14.0-427.16.1.el9_4.x86_64 crashkernel=1G-4G:192M,4G-64G:256M,64G-:512M rhgb mitigations=off tsc=nowatchdog nmi_watchdog=0 intel_pstate=disable processor.max_cstate=1 intel_idle.max_cstate=0 iommu=pt Runtime Options: cpupower idle-set -d 2 cpupower frequency-set -g performance echo 3 > /proc/sys/vm/drop_caches echo 0 > /proc/sys/kernel/nmi_watchdog echo 0 > /proc/sys/kernel/numa_balancing echo 0 > /proc/sys/kernel/randomize_va_space echo 'always' > /sys/kernel/mm/transparent_hugepage/enabled echo 'always' > /sys/kernel/mm/transparent_hugepage/defrag CPU: 2P 192-Core AMD EPYC 9965 (384 total cores) Memory: 24x 64 GB DDR5-6400 Storage: SAMSUNG MZWLO3T8HCLS-00A07 3.84 TB NVMe Platform and BIOS: RVOT1000C BIOS Options: SMT=Off NPS=4 Power Determinism Mode OS: rhel 9.4 5.14.0- 427.16.1.el9_4.x86_64 Kernel Options: amd_iommu=on iommu=pt mitigations=off tsc=nowatchdog nmi_watchdog=0 Runtime Options: cpupower idle-set -d 2 cpupower frequency-set -g performance echo 3 > /proc/sys/vm/drop_caches echo 0 > /proc/sys/kernel/nmi_watchdog echo 0 > /proc/sys/kernel/numa_balancing echo 0 > /proc/sys/kernel/randomize_va_space echo 'always' > /sys/kernel/mm/transparent_hugepage/enabled echo 'always' > /sys/kernel/mm/transparent_hugepage/defrag Results may vary based on system configurations, software versions, and BIOS settings.
Page 86
18 18 Endnotes • 9xx5-261: The AMD EPYC 9965 has the highest theoretical Floating Operations per Node (FLOPS per node) in any 1P or 2P server. This is calculated by: 2.25 Ghz Base Frequency, 192 Cores, 32 CPU instructions per cycle, and 1 or 2 CPUs per node. For: 2.25 x 192 x 32 x 1 = 13824 GFLOPS OR 2.25 x 192 x 32 x 2 = 27648 GFLOPS For reference, the Intel Xeon 6980P: 2.0 GHz Base Frequency. 128 Cores, 32 CPU instructions per cycle, and 1 or 2 CPUs per node. For: 2 x 128 x 32 x 1 = 8192 GFLOPS OR 2 x 128x 32 x 2 = 16384 GFLOPS MLN-058B: Based on SPECrate®2017_fp_base on 07/06/2021, a server powered by two 8c AMD EPYC 72F3 CPU has scored 247, http://spec.org/cpu2017/results/res2021q3/cpu2017-20210621-27506.html with a per core score of 15.4 which is a higher per core floating point base performance score than any currently posted in any SPEC.org publication. SPEC®, SPECrate® and SPEC CPU® are registered trademarks of the Standard Performance Evaluation Corporation. See www.spec.org for more information. • 9xx5-262: AMD EPYC 9175F has the highest 1P and 2P SPECrate®2017_int_base / Core as of November 04, 2025. 2P AMD EPYC 9175F (618 SPECrate®2017_int_base, 16 cores), 19.313 SPECrate®2017_int_base / Core, https://www.spec.org/cpu2017/results/res2025q2/cpu2017-20250602- 48622.html 1P AMD EPYC 9175F (303 SPECrate®2017_int_base, 16 cores), 18.938 SPECrate®2017_int_base / Core, https://www.spec.org/cpu2017/results/res2025q2/cpu2017-20250324-47116.html Highest Intel Xeon 1P and 2P results for Reference: 2P Intel Xeon 6507P (216 SPECrate®2017_int_base, 8 cores), 13.500 SPECrate®2017_int_base C / Core, https://www.spec.org/cpu2017/results/res2025q4/cpu2017-20250922- 49721.html 1P Intel Xeon 6349P (87.1 SPECrate®2017_int_base, 6 cores), 14.517 SPECrate®2017_int_base / Core, https://www.spec.org/cpu2017/results/res2025q2/cpu2017-20250518-48122.html SPEC®, SPEC CPU®, and SPECrate® are registered trademarks of the Standard Performance Evaluation Corporation. See www.spec.org for more information. • 9xx5TCO-005: This scenario contains many assumptions and estimates and, while based on AMD internal research and best approximations, should be considered an example for information purposes only, and not used as a basis for decision making over actual testing. The AMD Server & Greenhouse Gas Emissions TCO (total cost of ownership) Estimator Tool - version 1.3, compares the selected AMD EPYC and Intel® Xeon® CPU based server solutions required to deliver a TOTAL_PERFORMANCE of 391000 units of SPECrate®2017_int_base performance as of November 21, 2024. This estimation compares upgrading from a legacy 2P Intel Xeon 28 core Platinum_8280 based server with a score of 391 (https://spec.org/cpu2017/results/res2020q3/cpu2017- 20200915-23984.pdf ) versus 2P EPYC 9965 (192C) powered server with a score of 3100 (https://spec.org/cpu2017/results/res2024q4/cpu2017-20241004- 44979.pdf) and compared to a 2P Intel Xeon 64 core Platinum_8592+ (64C) based server with a SPECrate2017_int_base score of 1130, https://spec.org/cpu2017/results/res2023q4/cpu2017-20231127-40064.pdf. For additional details, see https://www.amd.com/en/claims/epyc.html#q=9xx5TCO- 005. SPEC®, SPEC CPU®, and SPECpower® are registered trademarks of the Standard Performance Evaluation Corporation. See www.spec.org for more information. Intel CPU specifications at https://ark.intel.com/. FOR USE ONLY WHEN ENVIRONMENTAL DATA IS CITED: Environmental impact estimates made leveraging this data, using the Country / Region specific electricity factors from Country Specific Electricity Factors - 2024, and the United States Environmental Protection Agency Greenhouse Gas Equivalencies Calculator.
Page 87
19 19 Endnotes • VEN-001A: SPECrate®2017_int_base comparison based on AMD internal estimates for top of stack 2P 6th Gen EPYC CPU and 5th Gen EPYC measurements as of 10/30/2025. Preliminary performance estimates based on AMD engineering projections or measurements as of 10/30/2025 and subject to change. • EPYC-058: Top 100 U.S. companies by revenue according to 2025 Fortune 500® list as of June 2, 2025. https://fortune.com/ranking/fortune500/. 'Fortune 100’ refers to the top 20% ranked companies in the 2025 Fortune 500 list, published in June 2025. From Fortune Magazine. ©2025 Fortune Media IP Limited. All rights reserved. Used under license. Fortune and Fortune Media IP Limited are not affiliated with, and do not endorse products or services of Advanced Micro Devices, Inc.
Page 88
Vamsi Boppana SVP , Artificial Intelligence Group
Page 89
2 Cautionary Statement This presentationcontainsforward-looking statements concerning Advanced Micro Devices, Inc. (AMD) such as the features, functionality, performance, availability, timing and expected benefits of AMD products including AMD Instinct MI400 series accelerators and AMD Instinct MI500 series accelerators; AMD's ability to deliver leadership roadmap on an annual cadence; the strategic partnership with OpenAI and the deployment of six gigawatts of AMD Instinct GPUs and the timing thereof; the strategic partnership with Oracle and the deployment of 50,000 GPUs and the timing thereof; the expected benefits of AMDAI compute strategic partnerships; and AMD's ability to advance AI leadership,which are made pursuant tothe Safe Harbor provisions of the Private Securities Litigation Reform Act of 1995. Forward-looking statements are commonlyidentifiedby words such as "would," "may," "expects," "believes," "plans," "intends," "projects" and other terms with similarmeaning. Investors are cautioned that the forward-looking statements in this presentation are based on current beliefs,assumptionsand expectations, speak only as of the date of thispresentationand 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 andgenerally 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 herebydisclaims, any obligation to update forward-looking statements made in this presentation, except as may be required by law.
Page 90
3 3 Entering Our Next Phase of Growth
Page 91
4 4 The Journey is Accelerating AI Performance on Annual Cadence Open-Source Software Ecosystem Adoption Rapidly Growing Partners & Customers Expanded Routes to Market
Page 92
5 5 Rapid Expansion of Partners & Customers 7 of Top 10 AI Companies Have Deployed AMD Instinct at Scale GTM Partners Customers Software Ecosystem
Page 93
AMD Instinct MI350 Series ▪ Next Gen AI Compute • FP4 | FP6 | FP8 ▪ Leadership Memory • HBM3E ▪ Industry Standard Design • OCP AMD Instinct MI350X AMD Instinct MI355X *See endnote MI350-006 6
Page 94
7 7 1x 1.1x 1.1x GPT-OSS-120b Grok 2.5 Hunyuan Video Inference Performance B200 MI355X Training Performance B200 MI355X See endnotes MI350-062, 063, 064, 065, 066, & 067. 1.1x 1.1x 1x Llama3 8B Mixtral 8x7B Llama2 70B BF16 Pre-Training FP8 • BF16 Pre-Training FP8 Fine-Tuning AMD Instinct MI350 with ROCm Delivers Performance Leadership
Page 95
8 8 0 100 200 300User Interactivity (tokens/sec/user) Cost per Million Tokens (USD) Up to 10x Inference Cost-per-Token Reduction Gen-on-Gen AMD Instinct MI350: Generational Efficiency Gains 3 - 10x GPT-OSS-120B, 1K/8K Sequence Length Based on SemiAnalysis’ InferenceMax, AMD data (10/27) AMD Instinct MI325X AMD Instinct MI355X
Page 96
9 9 Growing Workload Deployments on AMD Instinct GPUs Q4’25 Chatbot Bio Medical Ranking RAGImage Generation World ModelShopping AdsRecommendation Coding Video GenerationConversational AI LLM Pre-Training Reinforcement LearningVision Fine-Tuning Multi-ModalDiffusion Training Inference Expanding Customer Workloads Q1’24 Inference
Page 97
10 10 Enabling Open Innovation at Scale 10
Page 98
11 11 11 Relentless Focus on Developers Developer-First Approach to Enablement Expanded Support for Training Richer Out-of-the-Box Experience Accelerated Inference Capabilities Day-0 Support for Leading Models Accelerated Release Cadence Deepening Ecosystem Partnerships Industry Benchmarks
Page 99
12 12 Top 10 AI Open-Source GPU Projects* Natively Supported 2M Hugging Face Models Supported 10x Increase in ROCm Software Downloads Y/Y AMD ROCm Momentum *Top OSS projects = Pytorch, TensorFlow, JAX, Triton, Hugging Face Transformers, Unsloth, Ollama, ComfyUI, vLLM, SGLang12
Page 100
13 1313 Expanding the Developer Community Growing Learning & Compute Access for AI Communities Scalable Developer Cloud for Instant GPU Access Early Support for Research Labs / Start-Ups 100,000+ Learners Through AMD AI Education Program Active Engagement Through Developer Events
Page 101
14 14 Abstraction Enabling Higher Productivity Open Source Driving Rapid Innovation Advancing AMD ROCm Strategy
Page 102
15 15 Abstraction Enabling Higher Productivity Open Source Driving Rapid Innovation AI Assist Automating GPU Programming Advancing AMD ROCm Strategy
Page 103
16 16 Leadership Roadmap on Annual Cadence Accelerating Customer Adoption Roadmap subject to change. MI300X 2023 MI325X 2024 MI350 Series 2025 MI400 Series 2026
Page 104
17 17 AMD Instinct MI450 Series Most Advanced AI Accelerator Advanced Process Technology Leadership IP Chiplet Architecture 3.5D Packaging Co-Designed Hardware & Software 40 PF 20 PF FP4 • FP8 Flops 432 GB 19.6 TB/s HBM4 Memory Scale Up Bandwidth 3.6 TB/s Scale Out Bandwidth 300 GB/s Source: AMD specs based on engineering projections, results subject to change.
Page 105
18 18 Scale Out Bandwidth Scale Up Bandwidth FP4 | FP8 Flops (Avg.) Memory Bandwidth Memory Capacity 1.5x ~1x ~1x 1x 1.5x Rack-Scale Performance Leadership With MI450 Series Nvidia Vera Rubin AMD Instinct MI450 Series Source: AMD specs based on engineering projections, results subject to change MI450-002.
Page 106
19 19 AMD Instinct MI400 Series Portfolio Leadership Across AI & Scientific Computing At Scale AI Training & Inference AI Compute Scale Out Performance HBM4 Memory AMD Instinct MI455X Sovereign AI & HPC Hybrid Compute Hardware-Based FP64 HBM4 Memory AMD Instinct MI430X
Page 107
20 20 Frontier AI Development 6GW AMD Instinct Accelerators Starting in 2H’2026 Zetta-Scale Compute 50,000 AMD Instinct MI450 Series GPUs Starting in 2H’2026 Co-Designed Open Infrastructure AMD “Helios” Rack Co-defined with Meta and Previewed at OCP Strategic Partnerships Driving The Next Generation of AI Compute Extending US HPC Leadership Lux: First US AI Factory AMD Instinct MI355X Series Discovery: Flagship AI Supercomputer AMD Instinct MI430X Series
Page 108
21 21 Roadmap subject to change. MI300X 2023 MI325X 2024 MI350 Series 2025 MI500 Series 2027 MI400 Series 2026 Accelerating Customer Adoption Leadership Roadmap on Annual Cadence
Page 109
22 22 AMD Instinct MI500 Series AI Compute Performance AMD Instinct MI325X AMD Instinct MI355X AMD Instinct MI450 Series 2025 2027 20242023 Next Big Leap in AI Performance With MI500 Series 2026 Based on AMD internal projections as of Nov. 2025. AMD Instinct MI300X
Page 110
23 23 Advancing AI Leadership Deep Partnerships at Scale Building the Future of Compute Together with AI Leaders Leadership Roadmap Delivering Transformative AI Performance Open Innovation Ecosystem Driving Adoption Through Open Platforms
Page 111
24 24 Disclaimer & Attribution 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 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 Instinct, EPYC, Pensando, Radeon, ROCm, Ryzen, Versal, Xilinx, and combinations thereof are trademarks of Advanced Micro Devices, Inc. CXL is a registered trademark of Compute Express Link Consortium, Inc. OpenAI is a trademark of OpenAI, Inc. PCIe® is a registered trademark of PCI-SIG Corporation. UCIE is a trademark of Universal Chiplet Interconnect Express, Inc. Ultra Accelerator Link and UALink are trademarks of the UALink Consortium. 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 112
25 25 Endnotes • MI350-006: Calculations by AMD Performance Labs in May 2025, based on the published memory capacity specifications of AMD Instinct MI350X / MI355X OAM GPU vs. an NVIDIA Hopper H200 GPU and Server manufacturers may vary configurations, yielding different results. • MI350-062: Based on testing by AMD Performance Labs on Oct 24, 2025, using a Grok 2.5 xAI model to measure the average end-to-end latency improvement on an AMD Instinct 8x GPU MI355X platform vs. an Nvidia 8x GPU B200 platform. End-to-end latency measured during an inference performance cycle, with an 8K input length and 1k output length. Server manufacturers may vary configurations, yielding different results. Performance may vary based on use of latest drivers and optimizations. AMD system configuration: Dual Core AMD EPYC 9575F 64-core processor , 8x AMD Instinct MI355X GPU platform, AMD ROCm 7.0.0 software, SGLang 0.5.4, PyTorch 2.9.0, Ubuntu® 22.04. NVIDIA system configuration: 2x Intel Xeon 6960P processors, 8x Nvidia B200 (NVLink 192G, 1000W) GPU platform, Nvidia driver 570.133.20, and Ubuntu 22.04. • MI350-063: Based on testing by AMD Performance Labs on Oct 23, 2025, using an OpenAI GPT-OSS-120b model to measure the average throughput per second (TPS) of an 8x GPU AMD Instinct MI355X platform vs. an 8x GPU Nvidia B200 platform TPS measured during an inference performance cycle with an 8K input length and 1k output length. Server manufacturers may vary configurations, yielding different results. Performance may vary based on use of latest drivers and optimizations. AMD system configuration: One (1) dual core AMD EPYC 9575F 64-core processor, 8x GPU AMD Instinct MI355X platform, System Bios: 1.4a, ROCm 7.0.2, amdgpu version: 6.14.14 (2218231), Ubuntu® 22.04.5 LTS. NVIDIA system configuration: 2x Intel Xeon 6960P processors, 8x Nvidia B200 (NVLink 192G, 1000W) GPU platform, Nvidia driver 570.133.20, and Ubuntu 22.04. • MI350-064: Based on testing by AMD Performance Labs on Oct 27, 2025, measuring the Llama3 70B model’s pretraining throughput (tokens/second/GPU) on an AMD Instinct 8x GPU MI355X platform running Primus Torchtitan vs. an NVIDIA 8x GPU B200 platform, running NVIDIA Nemo. Each config tested with BF16 data precision and a maximum sequence length of 8192 tokens and a per-GPU batch size of 8. Server manufacturers may vary configurations, yielding different results. Performance may vary based on use of latest drivers and optimizations. AMD system configuration: Dual Core AMD EPYC 9575F 64-core processor, 8x AMD Instinct MI355X GPU platform, System BIOS 1.4a, 4 NUMA nodes per socket, Host OS Ubuntu 22.04.5 LTS with Linux kernel 5.15.160-generic, Host GPU driver AMD ROCm 7.0.1 + amdgpu 6.14.12, PyTorch deep learning framework 2.9.0, and AMD ROCm 7.0.1 software. NVIDIA system configuration: 2x Intel Xeon 6960P processors, 8x Nvidia B200 (NVLink 179GB, 1000W) GPU platform, System Bios 1.0, 3 NUMAnodes per socket, Host OS Ubuntu 22.04.5 LTS with Linux kernel 5.15.0-156-generic, Host GPU driver 580.82.07, PyTorch deep learning framework 2.8.0, and CUDA 13.0 software (MI350-064)
Page 113
26 26 Endnotes • MI350-065: Based on testing by AMD Performance Labs on Oct 27, 2025, measuring the Mixtral 8x7B model’s pretraining throughput (tokens/second/GPU) on an AMD Instinct 8x GPU MI355X platform running Primus Megatron-LM, vs. an NVIDIA 8x GPU B200 platform, running NVIDIA NeMo. Both configs tested with BF16 precision dataype, a maximum sequence length of 8192 tokens, and a per-GPU batch size of 8. Server manufacturers may vary configurations, yielding different results. Performance may vary based on use of latest drivers and optimizations. AMD system configuration: Dual Core AMD EPYC 9575F 64-core processor, AMD Instinct 8x GPU MI355X platform, System BIOS 1.4a, 4 NUMA nodes per socket, Host OS Ubuntu 22.04.5 LTS with Linux kernel 5.15.160- generic, Host GPU driver ROCm 7.0.1 + amdgpu 6.14.12, PyTorch deep learning framework 2.9.0, and AMD ROCm 7.0.1 software. NVIDIA system configuration: 2x Intel Xeon 6960P processors, Nvidia 8x GPU B200 (NVLink 179GB, 1000W) GPU platform, System Bios 1.0, NUMA 3 nodes per socket, Host OS Ubuntu 22.04.5 LTS with Linux kernel 5.15.0-156-generic, Host GPU driver 580.82.07, PyTorch deep learning framework 2.8.0, and CUDA 13.0 software. • MI350-066: Based on testing by AMD Performance Labs on Oct 27, 2025, measuring the Mixtral 8x7B model’s pretraining throughput (tokens/second/GPU) on an AMD Instinct 8x GPU MI355X platform running Primus Megatron-LM, vs. an NVIDIA 8x GPU B200 platform, running NVIDIA NeMo. Both configs tested with BF16 precision dataype, a maximum sequence length of 8192 tokens, and a per-GPU batch size of 8. Server manufacturers may vary configurations, yielding different results. Performance may vary based on use of latest drivers and optimizations. AMD system configuration: Dual Core AMD EPYC 9575F 64-core processor, AMD Instinct 8x GPU MI355X platform, System BIOS 1.4a, 4 NUMA nodes per socket, Host OS Ubuntu 22.04.5 LTS with Linux kernel 5.15.160- generic, Host GPU driver ROCm 7.0.1 + amdgpu 6.14.12, PyTorch deep learning framework 2.9.0, and AMD ROCm 7.0.1 software. NVIDIA system configuration: 2x Intel Xeon 6960P processors, Nvidia 8x GPU B200 (NVLink 179GB, 1000W) GPU platform, System Bios 1.0, NUMA 3 nodes per socket, Host OS Ubuntu 22.04.5 LTS with Linux kernel 5.15.0-156-generic, Host GPU driver 580.82.07, PyTorch deep learning framework 2.8.0, and CUDA 13.0 software. • MI350-067: Based on time-to-complete testing by AMD as of 10/17/2025 on an 8x GPU AMD Instinct MI355X platform for overall GPU-normalized training throughput for fine-tuning, using the Llama2-70B model with LoRA (FP8 precision). AMD test results were compared to the respective published performance results of an 8x GPU Nvidia B200 Platform (FP8). Server manufacturers may vary configurations, yielding different results. Performance may vary based on the use of the latest drivers and optimizations. AMD system configuration: Supermicro smci355-ccs-aus-n05-17: 2x AMD EPYC 9575F Processors (2 sockets, 64 cores per socket, 2 threads per core), 8x AMD Instinct MI355X (256GB, 2TB host memory) GPUs, BIOS RP 700D, SFO, PMFW 86.42.150, ROCm 7.0.0, Ubuntu 22.04.5 LTS, PyTorch 2.8.0++git64359f59. Docker image: rocm/mlperf:llama2_70b_training_5.1_2025-10-09-01-36-33 vs. Nvidia B200 8xGPU platform Results submitted here - https://mlcommons.org/benchmarks/training
Page 114
27 27 Endnotes • MI450-002: Preliminary performance comparison for AMD Instinct MI450 AI Rack Versus Nvidia Vera Rubin Rack. AMD Instinct specifications based on AMD engineering projections as of 11.10.2025 and are subject to change. Nvidia Vera Rubin specifications are from published data by Nvidia and may also be subject to change. AMD Rack vs. Nvidia Rack, Memory Capacity (HBM4) - 1.5x, Memory Bandwidth - 1x, FP4|FP8 FLOPS (avg) - 1x, Scale up bandwidth - 1x, Scale out bandwidth - 1.5x.
Page 115
Forrest Norrod EVP & GM, Data Center
Page 116
2 2 Cautionary Statement This presentationcontainsforward-looking statements concerning Advanced Micro Devices, Inc. (AMD) such as the features, functionality, performance, availability, timing and expected benefits of AMD products including "Helios" rack-scale platform; AMD server CPU revenue market share; AMD's ability to advance AI infrastructure on an annual cadence; AMD's Data Center AI opportunities; AMD's ability to accelerate Data Center growth; and AMD's Data Center projected revenue growth and CAGR, which are madepursuant tothe Safe Harbor provisions of the Private Securities Litigation Reform Act of 1995. Forward-looking statements are commonlyidentifiedby words such as "would," "may," "expects," "believes," "plans," "intends," "projects" and other terms with similarmeaning. Investors are cautioned that the forward-looking statements in this presentation are based on current beliefs,assumptionsand expectations, speak only as of the date of this presentationand 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 andgenerally beyondAMD'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 herebydisclaims, any obligation to update forward-looking statements made in this presentation, except as may be required by law.
Page 117
3 3 DPUs, UALink & Ultra Ethernet Networking Cluster Level Systems Design AMD ROCm Software GPUs AMD Instinct CPUs AMD EPYC ESUN
Page 118
4 4 Networking Drives Performance Front-End Scale Up Scale Across Scale Out
Page 119
5 5 ESUN AMD Networking: Open, Scalable, Adaptive
Page 120
6 Front-End Network: Connecting to Users, Storage & Applications ▪ Accelerate AI & cloud workloads ▪ Offload networking, storage & security from CPU / GPU ▪ Secure workloads with firewall, line-rate encryption & customer isolation 6
Page 121
7 AMD DPUs Deliver Network Innovation at the Speed of AI 7 ▪ Concurrent services offload ▪ Fully programmable P4 engines ▪ Line rate performance
Page 122
8 Back-End Networks: Scaling GPUs From Chiplets to Giga-Clusters ▪ Scale Up: Tightly couples hundreds of GPUs ▪ Scale Out: Linking racks of 100,000+ GPUs ▪ Scale Across: Federating data centers for giga-scale 8
Page 123
9 ▪ Low latency communication ▪ High bandwidth interconnects ▪ Scalable topology & reliability Scaling Up: Connecting Hundreds of GPUs to Work As One 9
Page 124
10 10 Scaling Up with “Helios” Taking an Open Approach ▪ 3.6TB/s of bandwidth with MI450 Series at rack-scale ▪ 72 GPUs with 260 TB/s aggregate bandwidth ▪ Shared load / store pod memory, acting as a single GPU ▪ Resilient communication over six parallel network planes ESUNEthernet
Page 125
11 Scaling Out: Connecting Hundreds of Thousands of GPUs to Work as One ▪ Fast, Reliable Network Performance UEC-ready protocol accelerates AI workloads Improves RAS vs RoCEv2 ▪ Accelerate AI Job Completion Boosts workload performance Improves compute resource utilization ▪ Lower Network TCO Multi-plane support & intelligent packet distribution reduce network costs 11 See Endnotes: PEN-016, PEN-019, PEN-020, PEN-018
Page 126
12 12 Scaling Across: Delivering Giga- Scale by Connecting Data Centers ▪ Intelligent Traffic Management Automatic rate adjustment for optimal performance over long distances ▪ Dynamic Load Balancing Adaptive path adjustment to optimize utilization across non-uniform links ▪ Peak Performance with Open Systems Approach Maintain line-rate communication between distributed data centers
Page 127
13 Scaling Out & Across Evolving Ethernet With Open Next-Generation AI NICs 13 *Roadmap subject to change AMD Pensando Pollara 400 ▪ 400 Gbps ▪ MI3XX Broadcom Thor 2 ▪ 400 Gbps AMD Pensando “Vulcano” ▪ 800 Gbps ▪ MI4XX Broadcom Thor Ultra ▪ 800 Gbps
Page 128
14 Available starting Q3’26 Optimized AI Rack Solution “Helios”
Page 129
15 15 *Roadmap subject to change AMD EPYC “Venice” AMD Instinct MI400 Series AMD Pensando “Vulcano” “Helios” Next-Gen AI Rack 2026 2027 Advancing AI Infrastructure on an Annual Cadence AMD EPYC “Verano” AMD Instinct MI500 Series AMD Pensando “Vulcano”
Page 130
16 16 DPUs, UALink & Ultra Ethernet Networking Cluster Level Systems Design AMD ROCm Software GPUs AMD Instinct CPUs AMD EPYC ESUN Data Center Leadership
Page 131
17 17 2025 includes Q4’25 financial outlook Data Center Leadership Next 3 – 5 Years Annual Data Center Revenue >$100B Data Center Revenue >60% CAGR
Page 132
18 18 Disclaimer & Attribution 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 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 Instinct, EPYC, Pensando, Radeon, ROCm, Ryzen, Versal, Xilinx, and combinations thereof are trademarks of Advanced Micro Devices, Inc. CXL is a registered trademark of Compute Express Link Consortium, Inc. OpenAI is a trademark of OpenAI, Inc. PCIe® is a registered trademark of PCI-SIG Corporation. UCIE is a trademark of Universal Chiplet Interconnect Express, Inc. Ultra Accelerator Link and UALink are trademarks of the UALink Consortium. 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 133
19 19 Endnotes • PEN-016: Testing conducted by AMD Performance Labs as of [28th April 2025] on the [AMD Pensando Pollara 400 AI NIC ], on a production system comprising of: 2 Nodes of 8xMI300X AMD GPUs (16 GPUs): Broadcom Tomahawk-4 based leaf switch (64x400G) from MICAS network; CLOS Topology; AMD Pensando Pollara AI. NIC – 16 NICs; CPU Model in each of the 2 nodes - Dual socket 5th gen Intel® Xeon® 8568 - 48 core CPU with PCIe® Gen-5 BIOS version 1.3.6 ; Mitigation – Off (default)System profile setting - Performance (default) SMT- enabled (default); Operating System Ubuntu 22.04.5 LTS, Kernel 5.15.0-139-generic. Following operation were measured: AllreduceAverage 25% for All-Reduce operations with 4QP and using UEC ready RDMA vs the RoCEv2 for multiple different message size samples (512MB, 1GB, 2GB, 4GB, 8GB, 16GB) . The results are based on the average at least 8 test runs. • PEN-018: AMD comparison and pricing as of July 6, 2025, for network fabric costs to support 128,000 GPUs. Comparison of a Pollara NIC with multiplane fabric and packet spray on an 800G Tomahawk 5–based multiplane design versus a generic fat-tree fabric built on fully scheduled, big-buffer (Jericho3/Ramon3) 800G switching platforms. The generic system is assumed to use a competitive NIC, with NIC costs considered comparable. The Pollara-based design is estimated to deliver up to 58% network switching cost savings by enabling the use of more cost-effective Tomahawk 5–based switching in a multiplane architecture. AMD comparison and pricing as of 4/23/2025 of a Tomahawk 5 system with Pensando Pollara NIC featuring exclusive multiplane fabric and packet spray versus a generic big-buffer 800G switching platform; the generic system woud employ a competitive NIC, costs of NICs are assumed to be comparable. Deploying Pollara with multi-fabric support and packet spray, allows customers to build cost-effective multiplane network fabrics, instead of a fat-tree design using less network switches to deliver the same amount of network bandwidth across the fabric, and dramatically reducing both switch platform cost, and cost associated with cables, optics. Fat-Tree Big Buffer Fully Scheduled Network (Leaf/Spine/Core) estimated Cost: $1.22B• 3556 leaf (Jericho3-AI) units at $104,998 each = $373M• 1557 spine/core (Ramon3) units at $147,998 each = $247M• 128K AOC-10m cables at $1059 each = $136M• 568,889 QDD-SR4-400G transceivers at $819 each = $466M• Total (Switching & Optics) = $1.22B• Naddod Tomahawk5 800G Multiplane Fabric Network estimated Cost: $511M• 3,000 Leaf and Spine Units (Naddod N9600-640C) at $26,999 each = $81M• 384K (QDD-SR4-400G) transceivers at $819 each=$313M• 64K Switch (OSFP-2x400G-DR4) transceivers for NIC connections at $759 each = $48M• 256K MPO Cables at $26 each = $6.6M• 2K Optical Shuffle box, modules and internal cables at $30K per rack =$60M• Total (Switching & Optics) = $511MPrices subject to change. Comparison for specific network configurations only, and may not be representative of all possible network configurations and comparisons. • PEN-019: Testing conducted by AMD Performance Labs as of [15 September 2025] on the AMD Pensando Pollara AI NIC, on a test system comprising of SMC- 300X server for GPU-GPU communication: 2x AMD Pensando Pollara AI NIC, 2P AMD EPYC 9454 48-Core -2P Processor, 8x AMD Instinct MI300X GPU, Ubuntu 22.04.5 LTS, kernel 5.15.0-139-generic, ROCm 6.4.1.0-83-69b59e5. Testing running Llama-3.1-8B, Model Configuration: SEQ_LEN=2048, TP=1, PP=1, CP=1,FP8=1, MBS=10, GBS = 5120. Iteration = 2, No. of paths/QP : 128. Results may vary based on factors including but not limited to system configuration and software settings.
Page 134
20 20 Endnotes • PEN-020: Testing conducted by AMD Performance Labs as of [15 September 2025] on the AMD Pensando Pollara AI NIC running Llama 3.1-405B @ 64 global batch size (GBS) with 8K sequence length, on a test system comprising of 8 node SMC-300X server for GPU-GPU communication using 2x AMD Pensando Pollara AI NIC or 2x Nvidia CX-7, 2P AMD EPYC 9454 48-Core 2P - Processor, 8x AMD Instinct MI300X GPUs, Ubuntu 22.04.5 LTS, kernel 5.15.0-139-generic, ROCm 6.4.1.0-83-69b59e5. Following operation are part of gateway function. Configuration : Num layers=4, Data Type=BF16, DCN - TP=1, PP=1, SP=1, DP=1, FSDP=-1, ICI - TP=1, PP=1, SP=1, DP=1, FSDP=8. AINIC container: jax-private:rocm6.4.0-jax0.5.0-py3.10.12-tedev2.1-20250801_training. Results may vary based on factors including but not limited to system configuration and software settings.
Page 135
Jack Huynh SVP & GM, Computing & Graphics
Page 136
2 Cautionary Statement This presentationcontainsforward-looking statements concerning Advanced Micro Devices, Inc. (AMD) such as the features, functionality, performance, availability, timing and expected benefits of AMD products; AMD’s client and gaming segment revenue growth, outpacing market growth; AMD's client business momentum, including revenue growth and client PC revenue share growth; AMD Ryzen AI roadmap; AMD's AI PC compute trajectory;AMD's ability to accelerate market leadership; AMD's client and gaming segment revenue growth, outpacing market rate through 2030; AMD's client business revenue market share growth; and AMD being positioned to obtain 40% client revenue share, which are madepursuant tothe Safe Harbor provisions of the Private Securities Litigation Reform Act of 1995. Forward-looking statements are commonlyidentifiedby words such as "would," "may," "expects," "believes," "plans," "intends," "projects" and other terms with similarmeaning. Investors are cautioned that the forward-looking statements in this presentation are based on current beliefs,assumptionsand expectations, speak only as of the date of thispresentationand 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 andgenerally beyondAMD'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 herebydisclaims, any obligation to update forward-looking statements made in this presentation, except as may be required by law.
Page 137
3 3 From Momentum to Market Leadership Leading the Gaming & AI PC Era
Page 138
4 4 $9.6B > $14B 2024 2025e Client & Gaming Revenue >$14B & ~50% Y/Y Growth in 2025 Outpacing Market Growth Client Gaming 2025e includes Q4’25 financial outlook
Page 139
5 5 Enterprise The World’s Best Workstation Platform 2x Platform Growth across HP , Lenovo, Dell & Asus Consumer The World’s Best & Top Selling Gaming CPUs The Most Powerful AI PC Processor in the Market Client Momentum Leadership Products Accelerating Growth Across All Segments See endnotes: GNR-21, FRG-01, SHO-06, SHP-03 Source: AMD Internal Projection
Page 140
6 6 “We are very happy with AMD since we are seeing better processor performance & better battery life compared to non-AMD solutions." CW5 Shawn E. Petermann Command Chief Warrant Officer, U.S. Army NETCOM
Page 141
7 7 1x 1.1x ~1.5x 2023 2024 2025e 15% 20% ~28% 2023 2024 2025e $5B $7B >$10B 2023 2024 2025e Client Revenue Client Processor ASP Client PC Revenue Share Client Momentum 2025e revenue includes Q4’25 financial outlook; Source: AMD Internal Projection
Page 142
8 8 Gaming Momentum Won Consecutive Console Generations >250M Units Shipped Across Multiple Generations of Xbox & PlayStation Consoles The Best Mainstream Gaming GPUs Graphics New Handhelds with Steam Deck, ROG Xbox Ally & Legion Go Handhelds See endnotes: RX-1174, RX-1184, & RX-1226
Page 143
9 9 Powering >1 Billion AI End-User Devices AMD internal projection
Page 144
10 10 We’re Infusing AI Into Everything We Do From Client Devices to Broader Compute Solutions
Page 145
11 11 The Ryzen AI PC Journey is Accelerating Over 250 AI PC Platforms With 2.5x Growth in 1 Year AI Will Transform the PC as We Know It Local Context Makes AI More Personal Models Adapt to Your Micro-Interactions Private Data Access Creates Deeper Insight
Page 146
12 12 MMLU = Measuring Massive Multitask Language Understanding *5 shot, post-trained. ** Reported leaderboard score. Small Models Are Getting Incredibly Capable & Efficient Running Faster, With Larger Context Windows & Fewer Parameters GPT-OSS Model Size Context Window MMLU ~128K tokens 20 billion ~85.3%** August 2025 Llama 3 Model Size Context Window MMLU ~8K tokens 70 billion ~84.0%* April 2024
Page 147
13 Leading the Industry in AI PC Compute Accelerating the Performance Trajectory 202520242023 20262022 2027 AI Compute Performance >10x AI Performance Increase “Hawk” “Strix” / “Krackan” “Phoenix” “Gorgon” Projection based on AMD internal analysis and data as of Nov. 2025. Roadmap subject to change. “Medusa”
Page 148
14 Accelerating With an AI-First Mindset Client Gaming +
Page 149
15 Edge AI AI Beyond the Endpoint Client Gaming + +
Page 150
16 16 Expanding AI Portfolio Exponential Growth of Real-Time & Private Data Creates New Demand for Edge AI Compute 16 Edge AI Laptop Desktop Workstation
Page 151
17 17 Unified ROCm Ecosystem from Endpoint to Cloud Dense Multi-GPU Edge Inference Deployment Deepening R&D Investment to Attack Edge Computing 24B 128B 500B Ryzen AI Ryzen AI Max Radeon AI Local Model Parameter Size Limit Based on AMD internal data and GPU specifications as of Nov. 2025 Building the Next Generation of Edge AI
Page 152
18 Client Revenue Market Share ~28% >40% Target Share 2025e 2026 2027 2028 2029 2030 Growing >3x Market Rate Through 2030 Client & Gaming Revenue 2025e Next 3-5 Years Accelerating Market Leadership 2025e revenue includes Q4’25 financial outlook
Page 153
19 19 Accelerating to expected >40% Client Revenue Share Unprecedented Consecutive Generations of Console Wins Attacking Edge AI as the Next Growth Frontier From Momentum to Market Leadership Leading the Gaming & AI PC Era
Page 155
21 21 Disclaimer & Attribution 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 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 Instinct, EPYC, Pensando, Radeon, ROCm, Ryzen, Versal, Xilinx, and combinations thereof are trademarks of Advanced Micro Devices, Inc. CXL is a registered trademark of Compute Express Link Consortium, Inc. OpenAI is a trademark of OpenAI, Inc. PCIe® is a registered trademark of PCI-SIG Corporation. UCIE is a trademark of Universal Chiplet Interconnect Express, Inc. Ultra Accelerator Link and UALink are trademarks of the UALink Consortium. 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 156
22 22 Endnotes • GNR-21: Testing as of October 2024 by AMD Performance Labs on test systems configured as follows: AMD Ryzen 7 7800X3D & 9800X3D system: GIGABYTE X670E AORUS MASTER, Balanced, 2x16GB DDR5-6000, Radeon RX 7900 XTX, VBS=On, SAM=On, KRACKENX63 (September 27, 2024); Intel Core i9- 14900K system: MSI MEG Z790 ACE MAX (MS-7D86), Balanced, 2x16GB DDR5-6000, Radeon RX 7900 XTX, VBS=On, SAM=On, KRAKENX63 (September 11, 2024) {profile=MSI Performance} on the following games: Ashes Of The Singularity: Escalation, Assassins Creed Mirage, Assassins Creed Valhalla, Avatar: Frontiers Of Pandora, Baldurs Gate 3, Black Myth: Wukong, Borderlands 3, Counter-Strike 2, CyberPunk 2077, Deus Ex: Mankind Divided, Dirt 5, DOTA 2, F1 2023, F1 2024, Far Cry 6, Final Fantasy 14 Dawntrail, Forza Horizon 5, Ghost Recon Breakpoint, Guardians Of The Galaxy, Hitman 3, Hogwarts Legacy, Horizon Zero Dawn, League of Legends, Metro Exodus, Metro Exodus Enhanced Edition, Middle Earth Shadow of War, Rainbow 6 Siege, Riftbreaker, Shadow Of The Tomb Raider, Spider Man Remastered, Starfield, Strange Brigade, The Callisto Protocol, Tiny Tinas Wonderlands, Total War Warhammer 3, Warhammer Dawn Of War 3, Watch Dogs Legion, World of Tanks encore, Wolfenstein Youngblood. System manufacturers may vary configurations, yielding different results. • FRG-01: Testing by AMD as of March 2025 on the following systems (best performance mode enabled): (1) an MSI Raider A18 notebook equipped with the AMD Ryzen 9 9955HX3D (75W), 64GB RAM, RTX 4090 GPU, and VBS ON versus (2) Acer Predator equipped with the Intel Core i9 14900HX (55W), 32GB RAM, RTX 4090 GPU, and VBS ON.Performance uplift based on the average of results from the following benchmarks: PC Mark 10, POV-Ray, Procyon Office productivity, 7-Zip, Phoronix Encode, LM Studio, Handbrake, Blender, Geekbench, Cinebench 2024, and V-Ray. System manufacturers may vary configurations, yielding different results. • SHO-06: Testing as of Dec 2024 using the following benchmark scores compared to Intel Core Ultra 9 288V and Qualcomm Snapdragon X Elite X1E-84-100. Cinebench 2024 nT, 3Dmark Wildlife Extreme, and Blender.. Next gen AI PC defined as a Windows PC with a processor that includes a NPU with at least 40 TOPS. Configuration for AMD Ryzen AI Max+ 395 processor: AMD reference board, Radeon 8060S graphics, 32GB RAM, 1TB SSD, VBS=ON, Windows 11. Configuration for Qualcomm Snapdragon X Elite X1E-84-100 processor: Samsung Galaxybook, Adreno Graphics, 16GB RAM, Microsoft Windows 11. Configuration for Intel Core Ultra 9 288V: ASUS Zenbook X 14, Intel Arc Graphics, 32GB RAM, 1TBSSD, Microsoft Windows 11 Home. Laptop manufacturers may vary configurations yielding different results. • SHP-03: V-Ray Benchmark 6 (CPU, ksamples), Keyshot Viewer 2024.2 benchmark, Corona Render (Rays/sec) benchmark 10, PugetBench Premiere Pro and After Effects, SPECapc Autodesk Maya 2024, SPECapc PTC Creo, Autodesk Revit, Unreal Engine 5.5 Compilation, Chromium Compilation 133.0.6868.0 and MATLAB benchmarks to compare the performance of the AMD Ryzen Threadripper PRO 9995WX processor in a reference system configured with 8x 64GB DDR5 memory, Nvidia RTX PRO 6000 Blackwell graphics, 1TB SSD, Win 11 vs. a similarly configured 7995WX, and a similarly configured BOXX workstation with the Intel® Xeon® W9-3595X processor. Workstation manufacturers may vary configurations, yielding different results. Results may vary.
Page 157
23 23 Endnotes • RX-1174: Testing done by AMD performance labs February 2025, on a test system configured with Ryzen 7 9800X3D CPU, 32 GB DDR5-6000 Memory, Windows 11 Pro and Radeon RX 9070 XT (Driver 25.3.1) vs. a similarly configured system with an RTX 5070 Ti (Driver 572.47) comparing gaming performance per dollar at 4K in the following applications: Cyberpunk 2077 (DX12, Ultra), Cyberpunk 2077 (DX12, RT Ultra), Assassin's Creed Mirage (DX12, Ultra High), F1 24 (DX12, Ultra High), F1 24 (DX12, Ultra High RT), Starfield (DX12, Ultra), Far Cry 6 (DX12, Ultra), Far Cry 6 (DX12, Ultra RT), Forza Horizon 5 (DX12, Extreme), Forza Horizon 5 (DX12, RT Extreme), Watch Dogs Legion (DX12, Ultra), Watch Dogs Legion (DX12, RT Ultra), Horizon Forbidden West (DX12, Maxed), Horizon Zero Dawn Remastered (DX12, Maxed), God of War: Ragnarok (DX12, Ultra), Call of Duty: Black Ops 6 (DX12, Extreme), DOOM Eternal (Vulkan, Ultra Nightmare), DOOM Eternal (Vulkan, Ultra Nightmare RT), Total War: Warhammer 3 (DX11, Ultra), Dying Light 2 (DX12, High), Dying Light 2 (DX12, High Raytracing), Alan Wake 2 (DX12, High), Alan Wake 2 (DX12, High w/Med RT), Avatar: Frontiers of Pandora (DX12, Ultra), Hitman 3 (DX12, Ultra), Hitman 3 (DX12, Ultra RT), The Witcher 3 (DX12, Ultra+), The Witcher 3 (DX12, RT Ultra), Metro Exodus Enhanced Edition (DX12, Extreme), Black Myth: Wukong (DX12, Cinematic), Black Myth: Wukong (DX12, Cinematic RT) Baldur's Gate 3 (DX11, Ultra), Ghost of Tsushima (DX12, Very High), Star Wars Outlaws (DX12, Ultra RT), Warhammer 40,000: Space Marine 2 (DX12, Ultra), Control (DX12, High), Control (DX12, High RT), Dragon Age: The Veilguard (DX12, Ultra), Dragon Age: The Veilguard (DX12, Ultra RT), Resident Evil 4 (DX12, Max), Resident Evil 4 (DX12, Max RT), Marvel’s Spider-Man 2 (DX12, Maxed), Marvel’s Spiderman 2 (DX12, Maxed RT), Microsoft Flight Simulator 2024 (DX12 Ultra), The Last of Us: Part 1 (DX12, Ultra), S.T.A.L.K.E.R. 2: Heart of Chornobyl (DX12, Epic), Final Fantasy XVI Demo (DX12, Ultra). Testing conducted on Radeon RX 9070 XT with latest game builds as of February 5, 2025 (Marvel’s Spider-Man 2, Microsoft Flight Simulator 2024, The Last of Us: Part 1, and Forza Horizon 5 using latest builds as of February 14th, 2025). Testing conducted on RTX 5070 Ti with latest game builds as of February 20, 2025. Performance per Dollar calculations using official launch SEP pricing as of Feb 25, 2025. System manufacturers may vary configurations, yielding different results.
Page 158
24 24 Endnotes • RX-1184: RX-1184: Testing done by AMD performance labs February 2025, on a test system configured with Ryzen 7 9800X3D CPU, 32 GB DDR5-6000 Memory, Windows 11 Pro and Radeon RX 9070 XT Partner Board (Asrock Taichi RX 9070 XT rated at 340W running on Driver 25.3.1) vs. a similarly configured system with an RTX 5070 Ti (Driver 572.47) comparing gaming performance at 4K in the following applications: Cyberpunk 2077 (DX12, Ultra), Cyberpunk 2077 (DX12, RT Ultra), Assassin's Creed Mirage (DX12, Ultra High), F1 24 (DX12, Ultra High), F1 24 (DX12, Ultra High RT), Starfield (DX12, Ultra), Far Cry 6 (DX12, Ultra), Far Cry 6 (DX12, Ultra RT), Forza Horizon 5 (DX12, Extreme), Forza Horizon 5 (DX12, RT Extreme), Watch Dogs Legion (DX12, Ultra), Watch Dogs Legion (DX12, RT Ultra), Horizon Forbidden West (DX12, Maxed), Horizon Zero Dawn Remastered (DX12, Maxed), God of War: Ragnarok (DX12, Ultra), Call of Duty: Black Ops 6 (DX12, Extreme), DOOM Eternal (Vulkan, Ultra Nightmare), DOOM Eternal (Vulkan, Ultra Nightmare RT), Total War: Warhammer 3 (DX11, Ultra), Dying Light 2 (DX12, High), Dying Light 2 (DX12, High Raytracing), Alan Wake 2 (DX12, High), Alan Wake 2 (DX12, High w/Med RT), Avatar: Frontiers of Pandora (DX12, Ultra), Hitman 3 (DX12, Ultra), Hitman 3 (DX12, Ultra RT), The Witcher 3 (DX12, Ultra+), The Witcher 3 (DX12, RT Ultra), Metro Exodus Enhanced Edition (DX12, Extreme), Black Myth: Wukong (DX12, Cinematic), Black Myth: Wukong (DX12, Cinematic RT) Baldur's Gate 3 (DX11, Ultra), Ghost of Tsushima (DX12, Very High), Star Wars Outlaws (DX12, Ultra RT), Warhammer 40,000: Space Marine 2 (DX12, Ultra), Control (DX12, High), Control (DX12, High RT), Dragon Age: The Veilguard (DX12, Ultra), Dragon Age: The Veilguard (DX12, Ultra RT), Resident Evil 4 (DX12, Max), Resident Evil 4 (DX12, Max RT), Marvel’s Spider-Man 2 (DX12, Maxed), Marvel’s Spiderman 2 (DX12, Maxed RT), Microsoft Flight Simulator 2024 (DX12 Ultra), The Last of Us: Part 1 (DX12, Ultra), S.T.A.L.K.E.R. 2: Heart of Chornobyl (DX12, Epic), Final Fantasy XVI Demo (DX12, Ultra). Testing conducted on Asrock Taichi RX 9070 XT (340W) using the latest game builds as of February 25, 2025. Testing conducted on RTX 5070 Ti with latest game builds as of February 20, 2025. System manufacturers may vary configurations, yielding different results • RX-1226: Testing done by AMD performance labs June 2025, on a test system configured with Ryzen 7 9800X3D CPU, 32 GB DDR5-6000 Memory, Windows 11 Pro, B650E Motherboard, and Radeon RX 9060 XT 16 GB (Driver 25.6.2) vs a similarly configured system with a GeForce RTX 5060 Ti 8 GB (Driver 576.52) comparing gaming performance in the following games: Cyberpunk 2077 (DX12, Ultra), Dragon Age: The Veilguard (DX12, Ultra), Call of Duty: Black Ops 6 (DX12, Extreme), F1 25 (DX12, Ultra High RT), Marvel’s Spider-Man 2 (DX12, Maxed RT), Monster Hunter Wilds (DX12, High RT). System manufacturers may vary configurations, yielding different results.
Page 159
Salil Raje SVP & GM, Adaptive & Embedded Computing
Page 160
2 2 Cautionary Statement This presentationcontainsforward-looking statements concerning Advanced Micro Devices, Inc. (AMD) such as the features, functionality, performance, availability, timing and expected benefits of AMD products; AMD's design win momentum; TAM expansion; AMD’s embedded revenue growth, outpacing market growth; AMD’s FPGA market share growth; silicon TAM expansion by 2035; and AMD being uniquely positioned to lead the physical AI inflection, which are madepursuant tothe Safe Harbor provisions of the Private Securities Litigation Reform Act of 1995. Forward-looking statements are commonlyidentified by words such as "would," "may," "expects," "believes," "plans," "intends," "projects and other terms with similarmeaning. Investors are cautioned that the forward-looking statements in this presentation are based on current beliefs,assumptionsand expectations, speak only as of the date of thispresentationand 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 andgenerally beyondAMD'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 herebydisclaims, any obligation to update forward-looking statements made in this presentation, except as may be required by law.
Page 161
3 3 From Acquisition to Acceleration AMD Embedded Transformation 3
Page 162
4 From Leadership FPGAs Adaptive Computing Broad Portfolio Coverage Best-in-Class Tools & IP Classic FPGA • $8B TAM Source: AMD Internal Projection
Page 163
Adaptive Computing Broad Portfolio Coverage Best-in-Class Tools & IP Classic FPGA • $8B TAM 5 To Leadership Embedded, With >$30B TAM Source: AMD Internal Projection Embedded x86 Leveraging AMD x86 Portfolio Targeting High-Value Markets Semi-Custom Silicon Strategic High-Value Customers Broad IP Portfolio & 3.5D Packaging Physical AI Broad AMD Portfolio Advantage Investing in Developer Experience AMD Embedded Today Synergistic TAM Expansion • >$22B TAM
Page 164
6 Investment Primarily in High-End FPGAs Limited Embedded x86 Portfolio Broad FPGA Portfolio from Cost-Optimized to Advanced Purpose-Built Embedded x86 Portfolio Ryzen V3000 Ryzen 9000 Ryzen 7000 Ryzen 5000 EPYC 9004 EPYC 9005 EPYC 7000 EPYC 4005 AMD Embedded Today Pre-Acquisition Expanding Embedded Portfolio Leveraging AMD IP
Page 165
7 7 ▪ Maintaining & extending adaptive leadership ▪ Delivering focused, purpose-built x86 embedded portfolio ▪ Offering end-to-end solution leveraging sales synergy Data Center Aerospace & Defense Test & Measurement Automotive >$36B Winning in Embedded Compute Design Wins Since 2022 Source: AMD Internal Projection Industrial Wired & Wireless
Page 166
8 8 Expanding Into Semi-Custom Silicon Customer Chiplet Integration Heterogeneous & Customizable Platform Supports 3rd Party / Customer IP 3.5D Advanced Packaging Industry Leader in 3.5D Packaging with Proven Track Record for High Quality Industry's Broadest IP Portfolio FPGA, x86 CPU, ARM SoC, GPU, AIE, SerDes, RF Technology
Page 167
9 9 Automotive ▪ Demonstrated safety & reliability capability ▪ Leadership scalar compute & AI into a single device Data Center ▪ Future-ready roadmap for security ▪ FPGA fabric offering customer differentiation Aerospace & Defense ▪ Physical security domain expertise ▪ Advanced packaging capabilities Wireless ▪ Leadership RF IP integrated with customer IP ▪ Technology roadmap partnerships for future generations Secured Semi-Custom Wins Since 2022 Data Center Wireless Aerospace & Defense Automotive ~$15B Source: AMD Internal Projection Winning in Semi-Custom Silicon
Page 168
10 2019 2020 2021 2022 2023 2024 2025e Semi-Custom Silicon & Embedded x86 Driving Major Expansion x86 & Semi-Custom Adaptive Compute AMD Embedded Today Pre-Acquisition Tracking to Exceed $16B in 2025 Unprecedented Design Win Momentum 2025e includes Q4‘25 financial outlook Source: AMD Internal Projection CAGR: 8% CAGR: 21%
Page 169
11 11 2025e 2026 2027 2028 2029 Accelerating to 3x with TAM Expansion Growing 2x Market Rate Through 2030 FPGA Market Share 2025 2030 ~55% >70% Target Share 2030 2025e includes Q4‘25 financial outlook Source: AMD Internal Projection Revenue Trajectory: Outpacing Market Growth x86 & Semi-CustomAdaptive Compute
Page 170
12 1212 As AI expands from the cloud to the physical world, billions of intelligent devices will shape the next wave of embedded computing Ready to Lead in Physical AI >$200B Silicon TAM by 2035 Source: AMD Internal Projection
Page 171
13 Motor Drive & Predictive Maintenance LiDAR – Real-Time Sensor Fusion Vision AI – Smart Camera Perception – Multiview Camera Real-Time Industrial Networking Robot Controller & Generative AI Digital Twin 13 AMD Powers Intelligent Machines
Page 172
14 14 From Perception to Action: AMD Powers Physical AI Decision Inference Compute Low Latency FPGA, GPU, NPU, Semi-Custom AI ASIC Embedded x86, ARM Action Real-Time Control Deterministic Response Perception Sensor Fusion Multi-Stream Processing AMD Delivers Inference Perception, Real-Time Control, Deterministic Response & Functional Safety Hardened IP (Including Semi-Custom), FPGA
Page 173
15 15 From Perception to Action: AMD Powers Physical AI Decision Inference Compute Low Latency FPGA, GPU, NPU, Semi-Custom AI ASIC Embedded x86, ARM Action Real-Time Control Deterministic Response Perception Sensor Fusion Multi-Stream Processing Hardened IP (Including Semi-Custom), FPGA AMD Delivers Inference Perception, Real-Time Control, Deterministic Response & Functional Safety Cost-Optimized Adaptive SoC High-Performance Adaptive SoC Embedded CPU Fully-Integrated Future Platform Broad IP Portfolio. Optimized Physical AI Solutions.
Page 174
16 16 Winning Multi-Generational Semi-Custom Silicon Positioned to Lead the Physical AI Inflection Record Design Wins Across a Broad Portfolio From Acquisition to Acceleration AMD Embedded Transformation
Page 176
18 18 Disclaimer & Attribution 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 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 Instinct, EPYC, Pensando, Radeon, ROCm, Ryzen, Versal, Xilinx, and combinations thereof are trademarks of Advanced Micro Devices, Inc. CXL is a registered trademark of Compute Express Link Consortium, Inc. OpenAI is a trademark of OpenAI, Inc. PCIe® is a registered trademark of PCI-SIG Corporation. UCIE is a trademark of Universal Chiplet Interconnect Express, Inc. Ultra Accelerator Link and UALink are trademarks of the UALink Consortium. 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 177
1 Jean Hu EVP , CFO & Treasurer
Page 178
2 2 Cautionary Statement This presentation contains forward-looking statements concerning Advanced Micro Devices, Inc. (AMD) including, but not limited to, AMD driving AI leadership and value creation; AMD's strong revenue growth and profit expansion; AMD's expected benefits of acquisitions and strategic investments; AMD’s Data Center revenue growth; AMD’s financial and operating performance; AMD’s Data Center TAM; AMD’s ability to expand data center and AI opportunities; AMD’s long-term revenue growth outlook, including potential drivers; AMD’s long-term financial outlook and model, including revenue growth, non-GAAP gross margin, non-GAAP operating margin, non-GAAP effective tax rate, non-GAAP FCF margin, non-GAAP EPS; AMD's disciplined investment strategy; AMD’s capital allocation priorities; AMD’s strong liquidity; AMD's ability to maintain a strong investment grade rating; AMD's ability to drive shareholder returns; and AMD’s ability to accelerate financial momentum, which are madepursuant tothe Safe Harbor provisions of the Private Securities Litigation Reform Act of 1995. Forward-looking statements are commonlyidentifiedby words such as "would," "may," "expects," "believes," "plans," "intends," "projects" and other terms with similarmeaning. Investors are cautioned that the forward-looking statements in this presentation are based on current beliefs,assumptionsand expectations, speak only as of the date of this presentationand 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 andgenerally beyondAMD'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 herebydisclaims, any obligation to update forward-looking statements made in this presentation, except as may be required by law. Non-GAAP Financial Measures This presentation contains historical non-GAAP financial measures including non-GAAP gross profit and margin, non-GAAP operating profit, and free cash flow. AMD is providing these financial measures because it believes this non-GAAP presentation makes it easier for investorsto compare its operating results for current and historical periodsand alsobecause AMD believes it assists investors in comparing AMD’s performance across reporting periods on a consistent basis by excluding itemsthat it does not believe are indicative of its core operating performance. The non-GAAP financial measures disclosed in this presentation should be viewed in addition to and not as a substitute for or superior toAMD’s reported results prepared in accordance with GAAP and should be read only in conjunction with AMD’s Consolidated Financial Statements prepared in accordance with GAAP. These non-GAAP financialmeasures referenced are reconciled to their most directly comparable GAAP financial measures in the Appendices at the end of this presentation. This presentation also contains forward-looking non-GAAP measuresconcerning AMD’s 2025 financial outlook and long-term financial model such as gross profit or margin, operating income or margin, effective tax rate, free cash flow margin and earnings per share. Projected 2025 non-GAAP gross profit and operating profit exclude $800 million inventory and related charges associated with U.S. export restrictions recognized in Q2’25. For long-term financial model, AMD uses a projected non-GAAP tax rate of 13-15%, which excludes the tax impact of pre-tax non-GAAP adjustments, reflecting currently available information. These forward-looking non-GAAP measures are based on current expectations, assumptions and beliefs that involve numerous risks and uncertainties. Adjustments to arrive at the GAAP financial outlook and long-term financial model typically include stock-based compensation, amortization of acquired intangible assets, income tax provision, and other non- recurring items such as impairment charges and acquisition-related costs. A reconciliation to equivalent GAAP measures is not practicable at this time as the timing and impact of such adjustments are dependent on future events that are typically uncertain or outside of AMD's control. Such events may include unanticipated changes in AMD’s GAAP effective tax rate, unanticipated one-time charges related to asset impairments, unanticipated acquisition-related expenses, unanticipated gains, losses, and impairments, and other unanticipated non-recurring items not reflective of ongoing operations. All statements made in this presentation are based on current expectations as of November 11, 2025, and assumptions and beliefs that involve numerous risks and uncertainties. AMD undertakes no intent or obligation to publicly update or revise its forward-looking statements made in this presentation except as may be required by law.
Page 179
3 3 Key Financial Priorities Delivering Compelling Profitability Accelerating Revenue Growth Capital Allocation for AI Leadership Driving AI Leadership & Value Creation
Page 180
4 $1B $4B ~$18B 2016 2020 2025e $4B $10B ~$34B 2016 2020 2025e $43M $2B ~$8B 2016 2020 2025e Gross Profit & GM% Revenue Operating Profit CAGR: 34% (1) Non-GAAP financial measures, except for Revenue. See appendices for GAAP to non-GAAP reconciliation 2025e includes Q4’25 financial outlook and excludes $800M inventory and related charges due to the U.S export restrictions. See Cautionary Statement on Slide 2 Not drawn to scale; Figures may be impacted by rounding Strong Revenue Growth & Profit Expansion 1 ~54% 45% 31% CAGR: 26% CAGR: 79%
Page 181
5 Organic Investments 2020 2025e >$40B ~$60B Acquisitions ~$1B Investments Strategic Investments Driving AI Growth >$100B Invested in Last 5 Years 2025e includes Q4’25 financial outlook.See Cautionary Statement on Slide 2
Page 182
6 $10B 2020 ~$34B 2025e Data Center Business CAGR ~52%Data Center Core Business 2025e includes Q4’25 financial outlook. See Cautionary Statement on Slide 2 Not drawn to scale ~47% Total 2025e Data Center Revenue >$16B ~17% Total 2020 Data Center Revenue $2B Data Center Driving AMD Revenue Growth Trajectory
Page 183
7 Gross Profit Revenue Operating Income (1) Non-GAAP financial measures, except for Revenue. See appendices for GAAP to non-GAAP reconciliation 2025e includes Q4’25 financial outlook and excludes $800M inventory and related charges due to the U.S export restrictions. See Cautionary Statement on Slide 2 Not drawn to scale; Figures may be impacted by rounding $14B ~$18B 2024 2025e $26B ~$34B 2024 2025e $6B ~$8B 2024 2025e 33% 33%32% Margin Expansion While InvestingRevenue Growth Acceleration Operating Leverage Expansion 2025: An Acceleration of Financial Momentum 1
Page 184
8 Our Largest Growth Opportunity AMD Data Center TAM >$1T 2030 ~$200B 2025 CAGR: >40% Source: AMD Internal Projections
Page 185
10 End Market Historical Performance 2020 – 2025e CAGR* Growth Expectations Next 3 – 5 Years Data Center ~52% >60% Core Business ~10% >10% AMD ~21% >35% Client & Gaming, Embedded * Calculated using 2020 pro forma net revenue as if the acquisition of Xilinx had been consummated as of the beginning of 2020. 2025e includes Q4’25 financial outlook; See Cautionary Statement on Slide 2 Accelerating Revenue Growth
Page 186
11 Long Term Financial Model 1 >35% CAGR Revenue Growth ▪ Data Center Momentum ▪ Share Gains 55-58% Gross Margin ▪ Richer Product Mix ▪ Cost Improvements >35% Operating Margin ▪ Scale & Operating Leverage 13-15% Effective Tax Rate ▪ Optimized Tax Structure ~25% FCF Margin ▪ Significant Cash Generation (1) Non-GAAP financial measures except for Revenue. 2025 includes Q4’25 financial outlook. See Cautionary Statement on Slide 2 The Next 3-5 Years
Page 187
12 Scale & Volume Operational leverage through scale Optimization Yield & test time improvement Design for cost & quality focus Mix Product leadership Product & customer mix Gross Margin1 ~54% 55% - 58% 2025e OptimizationScale & Volume Mix Target (1) Non-GAAP financial measure; 2025e includes Q4’25 financial outlook. See Cautionary Statement on Slide 2 Not drawn to scale Long-term model timeframe is 3 – 5 years Key Drivers Driving Gross Margin Expansion
Page 188
13 Disciplined Investment Strategy R&D ▪ Aggressively invest in AI & technology ▪ Leverage AI to drive R&D efficiency SG&A ▪ Invest in go-to-market & ecosystems ▪ Drive G&A efficiency through scale & AI Operating Leverage Through Scale (1) Non-GAAP financial measure See appendices for GAAP to non-GAAP reconciliation. 2025e includes Q4’25 financial outlook and excludes $800M inventory and related charges due to the U.S export restrictions. See Cautionary Statement on Slide 2 Not drawn to scale 17% ~24% >35% 2020 2025e Long-Term Operating Margin1 Delivering Compelling Profitability at Scale
Page 189
14 Long Term Financial Model1 (1) Non-GAAP financial measures except for Revenue; 2025 includes Q4’25 financial outlook. See Cautionary Statement on Slide 2 The Next 3-5 Years ~25% FCF Margin ▪ Significant cash generation >$20 EPS ▪ Financial model leverage 13-15% Effective Tax Rate ▪ Optimized tax structure >35% Operating Margin ▪ Scale & operating leverage 56-58% Gross Margin ▪ Richer product mix ▪ Cost improvements >35% CAGR Revenue Growth ▪ Data Center momentum ▪ Share gains ~25% FCF Margin ▪ Significant Cash Generation > $20 EPS ▪ Financial Model Leverage 13-15% Effective Tax Rate ▪ Optimized Tax Structure >35% Operating Margin ▪ Scale & Operating Leverage 55-58% Gross Margin ▪ Richer Product Mix ▪ Cost Improvements >35% CAGR Revenue Growth *Based on 2025 Outlook ▪ Data Center Momentum ▪ Share Gains
Page 190
15 Capital Structure $2B ~$5B 2024 2025e ▪ Liquidity: $10B ($7B Cash2 + $3B Revolver) ▪ A1/A Investment Grade Rating ~$8.6B 2021 2022 2023 2024 2025e Strong Free Cash Flow1 & Balance Sheet Cash Returns to Shareholders Since 2021 Cumulative Share Repurchase (1) Non-GAAP financial measure. See appendices for GAAP to non-GAAP reconciliation (2) Cash refers to cash, cash equivalents and short-term investments 2025e includes Q4’25 financial outlook. See Cautionary Statement on Slide 2 Not drawn to scale; Figures may be impacted by rounding
Page 191
16 ▪ Technology, products, go-to-market ▪ Infrastructure and talent Invest in the Business ▪ Strategic M&Ato drive AI leadership ▪ AI ecosystem investments M&A ▪ Incremental opportunistic buybacks with baseline of anti-dilutive ▪ $9.4B share repurchase authorization remaining Shareholder Returns ▪ Maintain strong balance sheet ▪ Strong investment grade rating & financial flexibility Balance Sheet Capital Allocation: Invest in AI Leadership
Page 192
17 17 Accelerating Financial Momentum >35% Revenue CAGR >35% Operating Margin1 Targeting >$1T Market Opportunity Capital Allocation For AI Leadership A Clear Path to >$20 EPS 1 (1) Non-GAAP financial measures. 2025 includes Q4’25 financial outlook; See Cautionary Statement on Slide 2
Page 193
18 18 Disclaimer & Attribution 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 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 Instinct, EPYC, Pensando, Radeon, ROCm, Ryzen, Versal, Xilinx, and combinations thereof are trademarks of Advanced Micro Devices, Inc. CXL is a registered trademark of Compute Express Link Consortium, Inc. OpenAI is a trademark of OpenAI, Inc. PCIe® is a registered trademark of PCI-SIG Corporation. UCIE is a trademark of Universal Chiplet Interconnect Express, Inc. Ultra Accelerator Link and UALink are trademarks of the UALink Consortium. 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 194
19 APPENDIX
Page 195
20 APPENDICES $ in millions, except % (Unaudited) 2016 2020 2024 YTD Q3 2025 GAAP gross profit $ 1,003 $ 4,347 $ 12,725 $ 11,575 GAAP gross margin 23% 45% 49% 47% Stock-based compensation 2 6 22 18 Amortization of acquisition-related intangibles - - 946 771 Acquisition-related and other costs (1) - - 1 1 Inventory loss at (recovery from) contract manufacturer (2) - - 65 (67) Loss contingency on legal matter - - - 12 Charge related to the sixth amendment to the WSA with GF (3) 340 - - - Non-GAAP gross profit $ 1,345 $ 4,353 $ 13,759 $ 12,310 Non-GAAP gross margin 31% 45% 53% 51% Inventory and related charges associated with U.S. export restrictions - - - 800 Non-GAAP gross profit (as adjusted to exclude inventory and related charges associated with U.S. export restrictions) $ 1,345 $ 4,353 $ 13,759 $ 13,110 Non-GAAP gross margin (as adjusted to exclude inventory and related charges associated with U.S. export restrictions) 31% 45% 53% 54% Reconciliation of GAAP to Non-GAAP Gross Profit and Gross Margin (1) Acquisition-related and other costs primarily include transaction costs, purchase price fair value adjustments for inventory, certain compensation charges, and workforce rebalancing charges. (2) Inventory loss at (recovery from) contract manufacturer is related to losses due to an incident at a third-party contract manufacturing facility in Q1’24 and the corresponding recovery. (3) Charge related to the sixth amendment to the Wafer Supply Agreement with GlobalFoundries
Page 196
21 APPENDICES $ in millions, except % (Unaudited) 2016 2020 2024 YTD Q3’25 GAAP operating income (loss) $ (373) $ 1,369 $ 1,900 $ 1,942 GAAP operating margin (9)% 14% 7% 8% Stock-based compensation 86 274 1,407 1,152 Amortization of acquisition-related intangibles - - 2,394 1,697 Acquisition-related and other costs (1) - 14 186 178 Inventory loss at (recovery from) contract manufacturer (2) - - 65 (67) Loss contingency on legal matter - - - 12 Restructuring charges (3) (10) - 186 - Charge related to the sixth amendment to the WSA with GF (4) 340 - - - Non-GAAP operating income $ 43 $ 1,657 $ 6,138 $ 4,914 Non-GAAP operating margin 1% 17% 24% 20% Inventory and related charges associated with U.S. export restrictions - - - 800 Adjusted Non-GAAP operating income (excluding inventory and related charges associated with U.S. export restrictions) $ 43 $ 1,657 $ 6,138 $ 5,714 Adjusted Non-GAAP operating margin (exclude inventory and related charges associated with U.S. export restrictions) 1% 17% 24% 23% Reconciliation of GAAP Operating Income (Loss) to Non-GAAP Operating Income (1) Acquisition-related and other costs primarily include transaction costs, purchase price fair value adjustments for inventory, certain compensation charges, and workforce rebalancing charges. (2) Inventory loss at (recovery from) contract manufacturer is related to losses due to an incident at a third-party contract manufacturing facility in Q1’24 and the corresponding recovery. (3) Restructuring charges are related to the 2015 and 2024 Restructuring Plan which comprised of employee severance charges and non-cash asset impairments. (4) Charge related to the sixth amendment to the Wafer Supply Agreement with GlobalFoundries
Page 197
22 APPENDICES Reconciliation of GAAP Net Cash Provided by Operating Activities of Continuing Operations to Free Cash Flow $ in millions, except % (Unaudited) 2024 GAAP net cash provided by operating activities of continuing operations $ 3,041 Operating cash flow margin % from continuing operations 12% Purchases of property and equipment (636) Free cash flow $ 2,405 Free cash flow margin % 9%
Page 198
Lisa Su Chair & Chief Executive Officer
Page 199
2 2 Cautionary Statement This presentation contains forward-looking statements concerning Advanced Micro Devices, Inc. (AMD) including, but not limited to, the features, functionality, performance, availability, timing and expected benefits of future AMD products, including the "Helios" AI rack-scale platform; AMD Instinct MI450 series accelerators; AMD’s projected revenue growth and CAGR across its segments; AMD’s ability to accelerate Data Center growth; AMD's growing semi-custom design wins; AMD's expected benefits of acquisitions and strategic investments; AMD's strong revenue growth and profit expansion; AMD’s Data Center TAM; AMD’s focus areas and strategic pillars; the strategic partnership with OpenAI and the deployment of six-gigawatts of AMD Instinct and timing thereof; the strategic partnership with Oracle and the deployment of 50,000 GPUs and the timing thereof; the expected benefits of AMD AI compute strategic partnerships; AMD’s ability to expand market leadership and accelerating revenue growth; and AMD's ability to deliver compelling shareholder returns, 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. Non-GAAP Financial Measures This presentation contains forward-looking non-GAAP measures concerning AMD’s long-term financial model such as non-GAAP operating margin and non-GAAP earnings per share. AMD is providing these financial measures because it believes this non-GAAP presentation makes it easier for investors to compare its operating results for current and historical periods and also because AMD believes it assists investors in comparing AMD’s performance across reporting periods on a consistent basis by excluding items that it does not believe are indicative of its core operating performance. The non-GAAP financial measures disclosed in this presentation should be viewed in addition to and not as a substitute for or superior to AMD’s reported results prepared in accordance with GAAP and should be read only in conjunction with AMD’s Consolidated Financial Statements prepared in accordance with GAAP. For long-term financial model, AMD uses a projected non-GAAP tax rate of 13-15%, which excludes the tax impact of pre-tax non-GAAP adjustments, reflecting currently available information. These forward-looking non-GAAP measures are based on current expectations, assumptions and beliefs that involve numerous risks and uncertainties. Adjustments to arrive at the GAAP financial outlook and long-term financial model typically include stock-based compensation, amortization of acquired intangible assets, income tax provision, and other non-recurring items such as impairment charges and acquisition-related costs. A reconciliation to equivalent GAAP measures is not practicable at this time as the timing and impact of such adjustments are dependent on future events that are typically uncertain or outside of AMD's control. Such events may include unanticipated changes in AMD’s GAAP effective tax rate, unanticipated one-time charges related to asset impairments, unanticipated acquisition-related expenses, unanticipated gains, losses, and impairments, and other unanticipated non-recurring items not reflective of ongoing operations. All statements made in this presentation are based on current expectations as of November 11, 2025, and assumptions and beliefs that involve numerous risks and uncertainties. AMD undertakes no intent or obligation to publicly update or revise its forward-looking statements made in this presentation except as may be required by law..
Page 200
3 3 The Path to Market Leadership Financial Analyst Day 2025 Broadest Product Portfolio World-Class Execution Deep Strategic Partnerships Large and Growing >$1T TAM Unmatched Data Center Solutions Best-in-Class Growth 1 >35% Revenue CAGR >35% Operating Margin >$20 Annual EPS (1) Non-GAAP financial measures, except for Revenue. See Cautionary Statement on Slide 2 From 2025 baseline which includes Q4’25 financial outlook and excludes $800M inventory and related charges due to the U.S exp ort restrictions.
Page 201
4 4 Disclaimer & Attribution 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 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 Instinct, EPYC, Pensando, Radeon, ROCm, Ryzen, Versal, Xilinx, and combinations thereof are trademarks of Advanced Micro Devices, Inc. CXL is a registered trademark of Compute Express Link Consortium, Inc. OpenAI is a trademark of OpenAI, Inc. PCIe® is a registered trademark of PCI-SIG Corporation. UCIE is a trademark of Universal Chiplet Interconnect Express, Inc. Ultra Accelerator Link and UALink are trademarks of the UALink Consortium. 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.