#1FINANCIAL ANALYST DAY 2022 together we advance_ Advancing Our Journey Dr. Lisa Su, Chair and CEO AMD#2Cautionary Statement This presentation contains forward-looking statements concerning Advanced Micro Devices, Inc. (AMD) including, but not limited to, the expected benefits of AMD's acquisition of Xilinx, Inc., including new markets and customer opportunities; the expected benefits of AMD's acquisition of Pensando Systems; TAM over the next five years and AMD's opportunity; AMD's strategic pillars; AMD accelerating financial momentum; AMD's long-term financial priorities and model, including potential drivers; and AMD being positioned for growth, 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. NON-GAAP FINANCIAL MEASURES In this presentation, in addition to historical GAAP financial results, AMD has provided historical non-GAAP financial measures including non-GAAP gross margin, non-GAAP operating margin, non-GAAP earnings per share and free cash flow. AMD uses a normalized tax rate in its computation of the non-GAAP income tax provision to provide better consistency across the reporting periods. For fiscal year 2019, 2020 and 2021, AMD used a non-GAAP tax rate of 3%, 3%, and 15%, respectively, which excludes the tax impact of pre-tax non-GAAP adjustments. 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 contains forward-looking non-GAAP measures concerning AMD's long-term financial model such as gross margin, operating expenses, operating margin and free cash flow margin. 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 typically include stock-based compensation, amortization of acquired intangible assets, income tax provision, equity income in investee, and other non-recurring items such as loss on debt redemption/conversion, impairment charges and acquisition-related costs. The timing and impact of such adjustments are dependent on future events that are typically uncertain or outside of AMD's control; therefore, reconciliation to equivalent GAAP measures is not practicable at this time. All statements made in this presentation are based on current expectations as of June 9, 2022. AMD undertakes no intent or obligation to publicly update or revise any statement in this presentation whether as a result of new information, future events or otherwise, except to the extent that disclosure may be required by law.#3A HIGH PERFORMANCE AND ADAPTIVE COMPUTING: CREATING SOLUTIONS TO THE WORLD'S TOUGHEST CHALLENGES.#4SINCE 2020 Successful Strategy Execution Excellence Technology Leadership Hypergrowth and Scale Unmatched Opportunity#5AMD COMPUTING NOW POWERS THE DAILY LIVES OF BILLIONS Cloud, Enterprise and HPC 5G & Comms Infrastructure Artificial Intelligence Adaptable Intelligent Systems Gaming, Simulation & Visualization Smarter Client Devices & Edge#6lum ZEN ● OUR JOURNEY TO LEADERSHIP AND SCALE THE STRATEGY IN 2020 AMD RDNA 2 Industry-Leading IP P REEE Advanced Technology Data Center Leadership PC/Gaming Solutions#7BEST-IN-CLASS ORGANIC GROWTH DATA CENTER LEADERSHIP Technology Leadership roadmap: performance, performance-per-watt, and TCO ■ AMD EPYC™ 7003 Series "Milan" Highest performance general purpose computing AMD EPYC™ 7003 Series with 3D V-Cache™ Highest performance for technical computing and data base workloads AMD INSTINCT™ MI200 Series Fastest HPC and Al Data Center Accelerators FINAN PANTERNE ********** Customer Expansion ▪ The world's 10 largest hyperscalers aws - Alibaba Cloud Bai du E IBM Cloud ORACLE Microsoft Azure Ford ▪ #1 on Top500, Green500 and HPL-AI Supercomputers Meta Goldman Sachs ▪ Strong traction across Enterprise customers, with 2X revenue YoY mastercard Google Cloud Tencent Cloud salesforce WELLS FARGO#8BEST-IN-CLASS ORGANIC GROWTH PC LEADERSHIP Technology Sustained processor performance and efficiency leadership from premium laptops to desktops and workstations ■ ■ AMD Ryzen™ 6000 Series Mobile APUS World's Best NB Performance and Battery Life AMD Ryzen™ 6000 Series Mobile PRO Uncompromising Performance, Battery Life, and Manageability with Modern Security Features AMD Ryzen™ 7 5800X3D World's Fastest Gaming Processor with 3D V-Cache™ AMD RYZEN 6000 SERIES Hullabal ****** 19 20 RYZEN # AMD PRO OEM Platform Expansion Commercial Consumer Gaming 185 117 Premium Designs 68 2020 261 204 Premium Designs 57 2021 294 249 Premium Designs 45 2022#9BEST-IN-CLASS ORGANIC GROWTH GAMING LEADERSHIP Technology Strong RDNA gaming graphics roadmap and leadership in combining GPU, CPU and software for next-gen experiences AMD Radeon™ RX 6000 Series Desktop and laptop performance-per-watt leadership with AMD RDNA 2 ▪ AMD Advantage™ Gaming solutions provider combining leadership CPUs, GPUs, software and system design AMD FSR and RSR Industry-leading upscaling solutions RADEON RYZEN AMD RADEON AMD ADVANTAGE AMD FidelityFX Super Resolution 2.0 Customer Momentum World's #1 Game Console Provider PlayStation 5, Xbox Series S|X, Steam Deck Over 50 PC Gaming Laptops#10OUR RESULTS BEST-IN-CLASS REVENUE GROWTH $6.7B 2019 56% CAGR $9.8B 2020 $16.4B 2021#11OUR RESULTS BEST-IN-CLASS REVENUE GROWTH Data Center & Embedded 95% CAGR $1.0B 2019 $1.8B 2020 $3.9B 2021 $3.2B 2019 PCs 46% CAGR $5.2B 2020 $6.9B 2021 $2.5B 2019 Gaming 50% CAGR $2.7B 2020 $5.6B 2021#12OUR RESULTS STRENGTHENED REVENUE MIX ~15% Data Center & Embedded 2019 Revenue Mix $6.7B PCs & Gaming VS. 2021 Revenue Mix $16.4B Data Center & Embedded ~25% Data Center & Embedded#13OUR RESULTS OUTSTANDING FINANCIAL PERFORMANCE Gross Margin 43% 2019 45% 48% 2020 2021 Operating Margin 12% 2019 17% 25% 2020 2021 Earnings per Share $0.64 2019 $1.29 2020 Non-GAAP financial measures. See appendices for GAAP to non-GAAP reconciliation. $2.79 2021 Free Cash Flow $0.3B 2019 $0.8B 2020 $3.2B 2021#14— XILINX. ☐☐ TRANSFORMATIONAL ACQUISITION XILINX Technology and Product Leadership Large and Diversified Markets New Customer Opportunities#15TRANSFORMATIONAL ACQUISITION: XILINX TECHNOLOGY AND PRODUCT LEADERSHIP Product Leadership #1 in FPGAs #1 in Adaptive SoCs Comprehensive Software Suite with Vitis. Vitis Al and Vivado Leadership IP Scalable Al Engine (AIE) High-Speed SERDES 2.5D/3D Packaging Wireless, Networking, Security, Video Solutions SmartNICs and Networking 5G Radio Platforms Automotive Healthcare, Industrial, Test and Measurement#16TRANSFORMATIONAL ACQUISITION: XILINX NEW MARKETS AND CUSTOMERS Diversified Markets Wired and Wireless 5G Communications Automotive, Aerospace/Defense Industrial. Healthcare. Vision. Test & Measurement, Emulation Deep Customer Partnerships 6 of the Top 7 5G Wireless Equipment Manufacturers Designed in the Top 10 Car Manufacturers 200M+ Autos in Production Deployed in the Top 10 Aerospace & Defense Companies #1 in Industrial loT Vision & Medical Imaging#17AMD INSTINCT 7FX A RYZEN EMBEDORD TRANSFORMATIONAL ACQUISITION AMD + XILINX OPPORTUNITIES ALVEO. ED wwwwww Al from Cloud to Edge to Endpoint T ANGKAT SMALLEO AMD EPYC Data Center and Communications Automotive Embedded Building Leadership Solutions for Large and Diversified Markets#18ACQUISITION PENSANDO Adds Industry-leading DPU Technology to Data Center Portfolio Comprehensive Software Stack Accelerates Networking, Security, Storage Deployed at Scale across Leading Cloud and Enterprise Customers World-class Team with Deep Networking, Software, and Systems Knowledge PENSANDO 08-0005-01 2PX2-PR11-4S TXV44W01 1844 TAIWAN N6U869.LN 036#19TAM AT 2020 FINANCIAL ANALYST DAY THE OPPORTUNITY Gaming $12B -$79B TAM Based on AMD Internal Data Data Center $35B PCs $32B#20INITIAL AMD + XILINX TAM THE OPPORTUNITY Gaming $16B PCs $40B $135B TAM Based on AMD Internal Data Data Center $50B Embedded $29B#21THE NEXT FIVE YEARS OUR ENORMOUS OPPORTUNITY Data Center $125B Gaming $37B PCs $50B -$300B TAM Based on AMD Internal Data Embedded $33B Automotive $27B Communications $32B#22COMPUTING MARKET Data Center and Cloud Insatiable Performance Demands THE NEXT FIVE YEARS Workload Optimized Compute/Networking Security from Core to Edge Efficiency and Sustainability Focus D D TRANSFORMATION 0 Explosion of Al 0 0 Al Workloads Proliferating Dominating the Data Center Expanding to Edge and Endpoint Increasingly Large Models PCs & Gaming Hybrid Work: Collaboration, Battery Life, Security Billions of Gamers Gaming Anywhere Metaverse#23Compute Technology Leadership THE NEXT FIVE YEARS STRATEGIC PILLARS Expanding Data Center Leadership Advancing Pervasive AI 25 requirettstras window. Vue required" require("./directives Vue.store store: # Localization Vue.use(vuexI18n.plugin, store, t on TranslationNotFound (locale, key) ( return "Key 's(key)' not found for locale > 111 Software Platforms & Developer Enablement Leading Provider of Custom Silicon and Solutions#24AMO RYZEN ****** STRATEGIC PILLARS UNMATCHED COMPUTE TECHNOLOGY AMD RYZEN HO CPUs AMD EPYC 200 AMD INSTINCT GPUs EVERSAL AMD XDNA Al Engine FPGA Fabric COLO ALVEO. KINTEX FPGAs and Adaptive SoCs Leadership Technology and Product Portfolio PENSANDO 08-0005-01 2PX2-PR11-48 TXV44W01 1844 TAIWAN NOU869.LN 038 SmartNICs and DPUS#25STRATEGIC PILLARS EXPANDING DATA CENTER LEADERSHIP General Purpose Technical Computing Computing and HPC "Zen" CPUs AMD 3D V-Cache™ Technology AMD Instinct™ GPUs Cloud Native Computing 20 Cloud-Optimized CPUs Machine Learning and Al AMD Instinct™ GPUs Versal™ Al Broad AIE Deployment Networking DPUS Adaptive SoCs Alveo™ SmartNICs#26STRATEGIC PILLARS ADVANCING PERVASIVE AI Training and Inference Across Cloud, Edge, and Endpoint Leadership Roadmap with AMD Instinct™ Bikin ******** AI Broad Deployment of Al Engine Across AMD BENEATER GLASS B egie Actor ucta Pall cec cleare Setarian iubend sictas UPROPERTY ptes AMD Unified Al Software Stack#27difiedFn = funct. //mark the element as t.appeared = true; STRATEGIC PILLARS SOFTWARE PLATFORMS $. inArray(check is this supposed to upor f (settings.one) { 8) $.fn.appe ~5,000 Software Engineers bind('scroll', check AMD CPU Tools and Compilers AMD Software Adrenalin Edition AMD Software: Adrenalin Edition¹ TM AMD VITIS VIVADO AMD Vitis™ and Vivado™ Ecosystem AMD AMD ROCM VITIS AMD ZenDNN Unification of Al Software#28STRATEGIC PILLARS LEADING PROVIDER OF CUSTOM SILICON & SOLUTIONS www GPU Leadership Process, Packaging, and Chiplet Solutions AIE CPU 3rd Party Broadest High Performance IP Portfolio Custom-Ready Chiplet Platform for 3rd Party and Customer IP Enabling the Highest-Performance Custom Computing Hyperscale, 5G and Automotive Opportunities#29AMD together we advance_#30LEADERSHIP COMPUTING ACROSS DATA CENTER WORKLOADS An AMD together we advance_data centers POWERING THE MOST ENERGY EFFICIENT SUPERCOMPUTERS ON EARTH AMD together we advance_sustainable computing ADVANCING ARTIFICIAL INTELLIGENCE AMD together we advance_Al Customers and Partners ADVANCING INTERPLANETARY SCIENCE AMD together we advance_aerospace POWERING CLOUD SERVICES FOR BILLIONS AMD together we advance_cloud computing Sadaf Usmani Assistant General Counsel WORKING ON CUTTING EDGE PROJECTS FOR THE DATA CENTER Jason Jacome Senior Manager Design Engineering FTO Enablement AMD together we advance MOTIVATING AND COACHING OTHERS TO ACHIEVE THEIR GOALS AMD together we advance Mario Andrade Quality St. Program Manager SETTING WORLD-CLASS STANDARDS FOR QUALITY PROCESSES Karmen Clark HR Business Partner Manager AMD together we advance ADVANCING WOMEN IN GLOBAL SALES Employees AMD together we advance.#31r THE JOURNEY AHEAD FINANCIAL MOMENTUM Participate In High Growth & Diverse Markets Expand Gross Margins Drive Increased Profitability and Free Cash Flow Disciplined Capital Allocation Continue to Deliver Strong Shareholder Returns#32STRONG AND COMPELLING LONG-TERM FINANCIAL MODEL" Revenue Growth Based on 2021 Proforma (AMD + Xilinx) Gross Margin Operating Expenses Operating Margin FCF Margin ~20% CAGR >57% ~23% -24% ~mid-30s% >25% 1,2 > > Continued Roadmap Execution Market Share Momentum Richer Product Mix Investing in Growth Expanding Margins and Profitability Significant Cash Generation (1) Non-GAAP financial measures except for Revenue; includes 2022 Financial guidance. See Cautionary Statement on Slide 2. (2) Long-term model timeframe is ~3 - 4 years#33DELIVERING SUSTAINED GROWTH World-Class Execution and Focus Unmatched Technology Leadership Deep Customer Relationships Diverse and Growing Markets Driving Compelling Shareholder Returns#34APPENDICES#35RECONCILIATION OF GAAP TO NON-GAAP FINANCIAL MEASURES (Millions, except percentages) GAAP Gross Margin GAAP Gross Margin % Stock-based compensation Non-GAAP Gross Profit Non-GAAP Gross Margin % GAAP Operating Income GAAP Operating Margin % Stock-based compensation Acquisition-related costs Loss contingency on legal matter Non-GAAP Operating Income Non-GAAP Operating Margin % $ $ SA SA 2019 2,868 43% 6 2,874 43% 631 9% 197 12 840 12% $ $ $ 2020 4,347 45% 6 4,353 45% 1,369 14% 274 14 1,657 17% $ $ $ 2021 7,929 48% 5 7,934 48% 3,648 22% 379 42 4,069 25%#36RECONCILIATION OF GAAP TO NON-GAAP FINANCIAL MEASURES (Millions, except per share data) GAAP Net Income / earnings per share Loss on debt redemption/conversion (Gains) losses on equity investments, net Non-cash interest expense related to convertible debt Stock-based compensation Equity income in investee Acquisition-related costs Loss contingency on legal matter Release of valuation allowance on deferred tax assets Income tax provision Non-GAAP Net Income / earnings per share Shares used and net income adjustment in earnings per share calculation Shares used in per share calculation (GAAP) (1) Interest expense add back to GAAP net income (1) Shares used in per share calculation (Non-GAAP) Interest expense add back to Non-GAAP net income $ 2019 341 176 22 197 12 8 $ 756 $ $ 0.30 0.15 0.02 0.16 0.01 0.64 1,120 $ 1,209 $ 16 I 2020 $ 2,490 54 6 274 (5) 14 (1,301) 43 $ 1,575 $ $ 2.06 0.04 0.22 0.01 (1.07) 0.03 1.29 1,207 $ 1 1,228 $ 4 $ 3,162 7 (56) $ 2021 379 (6) 42 . . (93) 3,435 $ 2.57 (0.04) 0.31 0.03 (0.08) $ 2.79 1,229 $ 1,229 $ (1) For the year ended December 28, 2019, GAAP diluted EPS calculation did not include shares related to the assumed conversion of the Company's 2026 Convertible Notes and the associated interest expense add-back to net income under the "if converted" method as their inclusion would have been anti-dilutive. For the year ended December 26, 2020, GAAP diluted EPS calculations include 3 million shares related to the assumed conversion of the Company's 2026 Convertible Notes and the associated $1 million interest expense add-back to net income under the "if converted" method. For the year ended December 26, 2020 and December 28, 2019, Non-GAAP diluted EPS calculations include 24 million and 89 million shares, respectively, related to the assumed conversion of the Company's 2026 Convertible Notes and the associated $4 million and $16 million interest expense, respectively, add-back to net income under the "if converted" method.#37RECONCILIATION OF GAAP TO NON-GAAP FINANCIAL MEASURES Shares (millions) (¹) Basic Shares Dilutive impacts from: Employee Equity Grants (2) 75 million share Warrant (3) Diluted Shares (without 2026 Convertible Notes) 2026 Convertible Notes - GAAP (4) Diluted Shares (with 2026 Convertible Notes) - GAAP 2026 Convertible Notes - Non-GAAP (4) Diluted Shares (with 2026 Convertible Notes) - Non-GAAP 2019 Actual 1,091 22 7 1,120 1,120 89 1,209 2020 Actual 1,184 20 I 1,204 3 1,207 24 1,228 2021 Actual 1,213 16 1,229 1,229 1,229 based on the Treasury Stock method and is dependent upon the average stock price during the period. The FY'19, FY'20, and FY'21 average stock price The table above provides actual share count for FY'19, FY'20, and FY'21 that may be used when calculating GAAP and non-GAAP diluted earnings per share for respective periods. (1) Share counts are weighted average shares. (2) The dilutive impact of employee equity grants was $29.77, $65.05,and $100.04, respectively. (3) The dilutive impact of the warrant to purchase 75 million shares (Warrant) granted in 2016 to a wholly owned subsidiary of Mubadala Investment Company PJSC, West Coast Hitech L.P. (WCH), in consideration for limited waiver and rights under the sixth amendment to our Wafer Supply Agreement with GLOBALFOUNDRIES Inc. is based on the Treasury Stock method and is dependent upon the average stock price during the period. The Warrant was exercised, and common stock shares were issued on February 13, 2019. Therefore, FY'19 includes the dilutive impact through the date of exercise. (4) The dilutive impact from the 2.125% Convertible Senior Notes due 2026 (2026 Convertible Notes) is based on the If-Converted method, where the interest costs associated with the 2026 Convertible Notes are added back to the Net Income and the shares underlying the 2026 Convertible Notes are assumed to be converted and are added to the share count. The impact from the 2026 Convertible Notes, if dilutive, is included in diluted EPS calculation. For the GAAP computation, the add-back to net income includes cash and non-cash interest expense, while only the cash interest expense is added back to the non-GAAP net income. The dilutive shares associated with the 2026 Convertible Notes reflect the weighted average shares subject to conversion during each period. During FY'19, FY'20, and FY'21, 69 million, 28 million, and 3 million shares of the Company's common stock were issued to convert $554 million, $225 million, and $25 million of the outstanding 2026 Convertible Notes, respectively.#38RECONCILIATION OF GAAP TO NON-GAAP FINANCIAL MEASURES (MILLIONS) GAAP net cash provided by operating activities Operating cash flow margin % Purchases of property and equipment Free cash flow Free cash flow margin % GA 2019 493 7% (217) 276 4% GA 2020 1,071 11% (294) 777 8% 2021 3,521 21% (301) 3,220 20%#39FINANCIAL ANALYST DAY 2022 together we advance_ Technology Leadership AMD Mark Papermaster Chief Technology Officer and Executive Vice President, Technology and Engineering#40CAUTIONARY STATEMENT This presentation contains forward-looking statements concerning Advanced Micro Devices, Inc. (AMD) including, but not limited to, the timing, availability, features, functionality and expected benefits of AMD's products; AMD's R&D investment; and AMD's CPU core roadmap, 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.#4190nm 45nm 65nm 22nm 32nm INNOVATION KEEPS AMD ON HIGH PERFORMANCE PACE 14nm 10/7nm 5nm Moore's Law Scaling Semiconductor Industry 2004 2007 2010 2013 2016 2019 2022 Moore's Law is Slowing Chiplet Integration Enables Generational Performance Gains, Keeps AMD on Historic Pace Source: AMD AMD RDNA AMD XDNA 준 AMD CDNA ZEN Heterogeneous Computing Driving Leadership Performance#42AMD BROADEST HIGH PERFORMANCE IP PORTFOLIO AMOU RYZEN CPUs AMD EPYC DNA o AMD INSTINCT GPUs VERSAL AMD XDNA and Al Engines CAN ALVEO SP AV KINTEX FPGAs and Adaptive SoCs PENSANDO 06-0005-01 2PX2-PR11-4S TXV44W01 1844 TAIWAN NOUB69.LN 030 SmartNICs and DPUS#43INVESTED FOR LEADERSHIP 7,000 Engineers $1.2B R&D Investment 2017 10,000 $2.0B 2020 12,000+ $2.8B 2021 16,000+ Engineers > $4.0B R&D Investment 2022#44HIGH PERFORMANCE CPU CORE ROADMAP 7nm “Zen 2” 2019 5nm | 4nm 7nm | 6nm “Zen 4” “Zen 4” “Zen 4c” 3D V-Cache ఉదరం “Zen 3” “Zen 3" 3D V-Cache™ All roadmaps are subject to change. “Zen 5” 4nm | 3nm “Zen 5” 3D V-Cache “Zen Sc” 2024#45LEADERSHIP HIGH PERFORMANCE EFFICIENCY 45% 45% 78% Better Perf/Watt vs. Competitor Less Area vs. Competitor Lower Power vs. Competitor Node Core + L2 Area 8C 16T Power at 3.9G 16C 32T Perf/Watt See endnotes Z4-006. AMD "Zen 3" Core TSMC 7nm 4.11 mm² 43.61W 1.78 Intel Alder Lake Golden Cove Core Intel 7nm Process 7.46 mm² 76.7W 1.0#46L ¯¯ ZEN 4 WORLD'S FIRST HIGH PERFORMANCE x86 5nm CPU For servers, notebooks and desktops Significant generational performance-per-watt and frequency improvement ▪ 8-10% instructions per clock increase ▪ >15% single thread performance gain Up to 125% memory bandwidth per core ISA extensions for Al and AVX-512 Performance estimates subject to change, see endnotes Z4-001, RPL-150 and Z4-002.#47HIGHER PERFORMANCE WITH LOWER POWER CONSUMPTION Performance-Per-Watt Gains Zen 3 >25% Zen 4 Overall Performance Improvement Zen 3 "Zen" Desktop 16C 32T Performance, Cinebench NT See endnotes Z4-003 and 24-004. >35% Zen 4#48Г ZEN 5 Optimized for scale across workloads COMING IN 2024 NEW GROUNDS-UP MICROARCHITECTURE ■ ■ ■ Enhanced performance and efficiency Re-pipelined front end and wide issue Integrated Al and Machine Learning optimizations#49LEADERSHIP HIGH-PERFORMANCE COMPUTE PLATFORM x AMD Infinity Architecture Decade long investment in computing and modular fabric architecture + Chiplet Leadership Heterogeneous compute engines and memory configurations in 2.5D and 3D chiplet packaging Leadership Compute Platform Enabled by AMD Infinity Architecture modularity and leading chiplet portfolio#50SCALING HORIZONTALLY AND VERTICALLY HBM Scalable Solution = Lithographically Defined Substrate GPU 2.5D Elevated Fanout Bridge Standard Flip Chip Process EFB Lower Complexity Bumping, Assembly Process Standard Substrates Lower Cost 3D Cache Die CPU Die > 3X Interconnect Energy Efficiency Compared to Micro Bump 3D See endnotes EPYC-26 and EPYC-27. -- 3D Cache Metal Stack 3D Hybrid Bonding ‒‒‒‒ CPU Die Metal Stack >15X Interconnect Density Compared to Micro Bump 3D#51AMD CHIPLET AND PACKAGING LEADERSHIP AMD 2.5D HBM XILINX. 2015 1.0.1... World's Highest Capacity FPGA in 20nm 2015 Multichip Module 2017 Chiplets 2019 VIRTEX World's Highest Capacity FPGA in 16nm 2019 AMD 3D V-Cache™ MI200 2.5D EFB 2021 VERSAL (PREMUN Versal™ Premium ====⠀⠀⠀⠀⠀⠀⠀⠀⠀‒‒‒‒‒‒‒‒‒‒‒ 2021 2022 MUNG VERSAL Versal 2.5D 2022 3D/2.5D HYBRID COMPUTE FUTURE#52RADEONICE AMD 8 EPYC 2017 1st Gen AMD EPYC AMD INFINITY ARCHITECTURE AMD MODULAR STRATEGY FOUNDATION RADEONFECT PCIe 3.0 CPU connectivity ▪ Common fabric and methodology Scalable data and control ■ 2nd Gen RADEON CTRCT ✔KADEEM. ✔ *www. AMD EPYC AMD EPYC ✔ RADE ENCORE 2019 PCIe® 4.0 4/8-way GPU connectivity CPU multi-node chiplet Advanced chiplet design enablement AMD EPYC AMD EPYC S 3rd Gen Up to 8-way GPU connectivity ▪ Coherent memory between EPYC™ CPUs and Instinct™ GPUs ▪ Exascale CPU and GPU compute 2021#53"ZEN" CPU AMD RDNA GPU AMD 3D V-CACHE 8 AIE FPGA AMD CDNA GPU Unmatched High Performance Configurability INTRODUCING 4TH GEN AMD INFINITY ARCHITECTURE ▪ Enables 2.5D and 3D chiplet integration with unified system level coherency ▪ Extensions for Xilinx and 3rd party IP ▪ CXL™M 2.0-based memory for disaggregation TM Extensible architecture for CXL 3.0 and UCle standards#54CHIPLET PLATFORM LEADERSHIP AMD is the industry leader in chiplet product deployment with >40 chiplet standard products in manufacturing Multi-node solutions with wide range of IP combinations ▪ Low latency, coherent and power efficient memory and cache Xilinx adaptable compute chiplets Custom-ready heterogeneous platform for 3rd party and customer IP chiplets. GPU AIE CPU 3rd Party#55OUR PATH FORWARD RELENTLESS INNOVATION ▪ Consistent delivery of leadership CPU Industry's broadest high performance IP portfolio with leadership chiplet flexibility ▪ No let up in innovation to power the future 140#56ENDNOTES Z4-006: Testing by AMD Performance Labs as of May 22, 2022. Performance per watt evaluated by dividing Cinebench R23 nT score by wall power of otherwise identical systems. Alder Lake system: Core i9-12900KS, ROG Maximus Z690 Hero (BIOS 0702), 2x16GB DDR5-5200 at 392W. Zen 3 system: Ryzen 9 5950X, ROG Crosshair VIII Hero, 2x8GB DDR4-3600. All systems configured with Radeon RX 6950XT (driver 22.10 Prime), Windows 11 build 22000.593, Samsung 980 Pro 1TB SSD, Asetek 280MM liquid cooler at 203W. Results may vary based on system configuration and other variables. Z4-001: IPC uplift based on the average of estimated/published 2017 SPECint® and 2017 SPECfp scores and internal estimates/testing on Cinebench R23 1T and Geekbench 51T.for "Zen4" and "Zen 3" processors. Ⓡ RPL-001: Testing as of May 5, 2022, by AMD Performance Labs. Single-thread performance evaluated with Cinebench R23 1T. AMD Ryzen 9 5950X System: ASUS ROG Crosshair VIII Hero X570, 2x8 DDR4-3600C16. AMD Ryzen 7000 Series: AMD Reference X670 Motherboard, 16-core pre-production processor sample, 2x16GB DDR5- 6000CL30. All systems configured with Radeon™ RX 6950XT GPU (driver: 22.10 Prime), Windows 11 Build 22000.593, Samsung 980 Pro 1TB SSD, Asetek 280MM liquid cooler. Results may vary. Z4-002: Preliminary specification for 96-core Zen 4 are 12 channels of DDR5 memory compared to 64-core EPYC™ 7763 processor with 64-cores and 8 channels of DDR4 memory. Calculation is per core using 64-cores per processor. Z4-003: Testing as of May 31, 2022, by AMD Performance Labs. Power measured at CPU socket only (Watts), CPU performance ("points") measured with Cinebench R23 nT. AMD Ryzen 9 5950X System: AMD Reference X570 Motherboard, 2x8 DDR4-3200. AMD Ryzen 7000 Series: AMD Reference X670 Motherboard, Ryzen 7000 Series 16-core pre-production processor sample, 2x16GB DDR5-5200. All systems configured with Radeon™ RX 6950XT GPU (driver: 22.10 Prime), Windows 11 Build 22000.593, Samsung 980 Pro 1TB SSD, Asetek 280MM liquid cooler. Results may vary when final products are released in market. Z4-004: Testing as of May 5, 2022, by AMD Performance Labs. Single-thread performance evaluated with Cinebench R23 1T. AMD Ryzen 9 5950X System: ASUS ROG Crosshair VIII Hero X570, 2x8 DDR4-3600C16. AMD Ryzen 7000 Series: AMD Reference X670 Motherboard, Ryzen 7000 Series 16-core pre-production processor sample, 2x16GB DDR5-6000CL30. All systems configured with Radeon™ RX 6950XT GPU (driver: 22.10 Prime), Windows 11 Build 22000.593, Samsung 980 Pro 1TB SSD, Asetek 280MM liquid cooler. Results may vary when final products are released in market. EPYC-026: Based on calculated areal density and based on bump pitch between AMD hybrid bond AMD 3D V-Cache stacked technology compared to AMD 2D chiplet technology and Intel 3D stacked micro-bump technology. EPYC-027: Based on AMD internal simulations and published Intel data on "Foveros" technology specifications.#57FINANCIAL ANALYST DAY 2022 together we advance_ Driving GPU Leadership David Wang Senior Vice President, Engineering, Radeon Technologies Group AMD#58Cautionary Statement This presentation contains forward-looking statements concerning Advanced Micro Devices, Inc. (AMD) including, but not limited to, AMD's GPU momentum and strategy; AMD's gaming GPU and compute GPU architecture roadmaps; and the timing, availability, features, functionality and expected benefits of AMD's products, 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.#59THE RIVE DENERGY ENERGY Packard Enterprise AMD FRONTIER Supercomputer Frontier LUMI Adastra AMD GPU MOMENTUM INSTINCT AMDA INSTINE Data Center AWS Google Cloud Microsoft Azure RADEON PC LLC Radeon™ RX 6000 Series Radeon™ W6000 Series Ryzen™ 6000 Series 'Y B A Console PlayStation 5 Xbox Series X S Steam Deck Embedded Magic Leap Tesla 5G SAMSUNG Exynos Mobile Samsung#60GPU TECHNOLOGY STRATEGY DELIVERING PERFORMANCE AND EFFICIENCY LEADERSHIP Architecture Domain-Specific Architecture Optimizations SMILE Technology Advanced Process and Packaging Technologies Efficiency Leadership Performance- Per-Watt Roadmap Ecosystem Open-Source Software, Including Al#61AMD RDNA™ 2 ARCHITECTURE Designed for enthusiast gaming Innovation High-Speed Compute Units, Raytracing Cores, AMD Infinity Cache™, DX12® Ultimate Performance Delivered 2X Performance and More Than 50% Performance-Per-Watt Gains vs. AMD RDNA™ Scalability Powers Mobile, Console. Ryzen™ APU, Radeon™ GPU, and Cloud See endnotes RX-549 and RX-558#62AMD RDNAT 2 LEADERSHIP GAMING PERFORMANCE AND EFFICIENCY Up to 1.27X Performance-Per-Watt Nvidia RTX 3090 Ti Up to 1.14X Performance-Per-Areal RadeonTM RX 6950 XT Average of 20 games, See endnote RX-785 AS A FOR STRAUME & PROM PERBE SEKET A DEORE U PA BERNAMEY&BOAR 11 JID WERKE TO 7. ALAM ANDANT KAUN PRARA#63闺 Vigorous Testing AMD Software Adrenalin Edition Community Engagement Regular Updates Focus on Quality and Stability GREAT HARDWARE NEEDS GREAT SOFTWARE 15% Year-Over-Year Uplift Day-0 Driver Support Continuous Performance Improvements Performance Uplift Comparing AMD Software: Adrenalin 22.3.1 vs. 20.12.2. See endnote RS-462 AMD RADEON Anti-Lag AMD AMD AMD RADEON Image Sharpening AMD FreeSync FidelityFX Link RADEON Boost AMD AMD RADEON Super Resolution Immersive Experiences#64AMD THE JOURNEY RDNA 3 CONTINUES Projected Performance/Watt Uplift RDNA 2 >50%* RDNA 3 Performance-per-watt uplift through: 5nm Process Advanced Chiplet Packaging Rearchitected Compute Unit Optimized Graphics Pipeline Next-Gen AMD Infinity Cache Based on preliminary internal engineering estimates. Actual results subject to change. TM RADEON#65Advanced Chiplet Packaging Leadership Performance and Scalability AT THE FOREFRONT OF GRAPHICS INNOVATION LOENG 454 End-to-End Power Optimization System Level Energy Efficiency Hybrid Rendering Real-Time Immersive Experiences 24 Next-Gen Multimedia Enhanced Video and Display Capabilities#66AMD GAMING GPU ARCHITECTURE ROADMAP DRIVING PERFORMANCE AND EFFICIENCY LEADERSHIP 2019 7nm AMD RDNA AMD RADEON RX 5000 SERIES 7nm AMD RDNA 2 AMD RADEON RX 6000 SERIES 5nm AMD RDNA 3 "Navi 3x" All roadmaps are subject to change. Advanced Node AMD RDNA 4 "Navi 4x" 2024#67AMD CDNA™M 2 ARCHITECTURE Exascale-class technology for HPC/AI TM Innovation High-Performance Dual Engines in MCM, 3rd Gen AMD Infinity Architecture, Ultra-Wide HBM Interface Performance ~4X Higher FP64 (HPC) and ~2X Mixed Precision (AI) Peak Performance than AMD CDNA™ Open Ecosystem Broaden ROCm™ Open Software Platform, Including Al See endnotes M1200-05 and MI200-01 TM#68AMD CDNAT 2 LEADERSHIP PERFORMANCE AND EFFICIENCY AMD Instinct™ MI250X vs. Nvidia A100 ~2.8X HPL ~1.9X HPL-AI Performance ~2.0X HPL ~1.5X HPL-AI Performance-Per-Watt See Endnotes M1200-26B and M1200-58 WA te word oor tet te dink die Seun van die **** *** *** **** ***** *** **** **** **** 000 170 ****** ***** DE HAR FOR men den van ******************************************************* the ***************** LENI ROHV *********************** GE WAR AMDA INSTINCT SON MAN SIN F BAND B#69AMD ROCM OPEN SOFTWARE PLATFORM FOR GPU COMPUTE ▪ Unlocked GPU Performance to Accelerate Computational Tasks ▪ Optimized for HPC and Al Workloads at Scale Open Source Enabling Collaboration, Innovation, and Differentiation Al Framework Support Programming Models Libraries Compilers & Tools Drivers & Runtime Deployment Tools ONNX Runtime OpenMP API BLAS SOLVER RAND ALUTION Compiler Profiler FFT PyTorch HIP API SPARSE TensorFlow OpenCL™ MIGraphX MIVisionX THRUST MIOpen Tracer Debugger hipify ROCm Validation Suite ROCm Data Center Tool PRIM RCCL TENSILE Red Hat, CentOS, SLES & Ubuntu Device Drivers and Run-Time ROCM SMI#70FROM HPC TO AI ROCMT JOURNEY TO ECOSYSTEM ENABLEMENT AMD ROCm 4 HPC and Exascale Optimized HPC Performance Enabled AMD CDNATM GPUs Hardened for Scale 2021 AMD ROCm 5 Expanding to Al Optimized Training & Inference Performance Enabling AMD CDNA™ & AMD RDNA™ GPUs SDK for Development & Deployment 2022#71INNOVATING AI THROUGH DEEP PARTNERSHIPS Microsoft Azure Accelerating ROCm Performance on Al "We're also deepening our investment in the open-source PyTorch framework, working with the PyTorch Core team and AMD both to optimize the performance and developer experience for customers running PyTorch on Azure, and to ensure that developers' PyTorch projects work great on AMD hardware." Kevin Scott Executive Vice President and CTO, Microsoft PyTorch Optimizing ROCm for PyTorch "We are excited to partner with AMD to grow our PyTorch support on ROCm, enabling the vibrant PyTorch community to adopt the latest generation of AMD Instinct GPUs quicker than ever with great performance for major Al use cases running on PyTorch." Soumith Chintala Co-Creator and Lead of PyTorch, Meta Al LANDING AI Developing Data Centric Al for Instinct™ GPUs "We are excited to partner with AMD to leverage AMD MI200 GPUs and the ROCm stack to port and optimize LandingLens, our GPU-optimized computer vision software application. The power of AMD GPUs and the maturity of ROCm will help Landing Al continue to deliver acceleration of high-resolution computer vision models with the purpose of providing better insights into manufacturing defects." Andrew Ng CEO of Landing Al and Adjunct Professor, Stanford University#72AMD CDNA 3 Al Performance/Watt Uplift >5X AMD CDNA 2 AMD CDNA 3 THE JOURNEY CONTINUES Expected performance-per-watt uplift through: 5nm Process and 3D Chiplet Packaging Next-Gen AMD Infinity Cache™ 4th Gen Infinity Architecture. Unified Memory APU Architecture New Math Formats See endnote M1300-04. Preliminary data and projections, subject to change HBM GPU CPU INFINITY CACHE НЕМ#73UNIFIED MEMORY APU ARCHITECTURE BENEFITS AMD CDNA™ 2 Coherent Memory Architecture Simplifies Programming Low Overhead 3rd Gen Infinity Interconnect Industry Standard Modular Design CPU CPU Memory (DRAM) GPU GPU Memory (HBM) *AMD Projections AMD CDNA™ 3 Unified Memory APU Architecture Eliminates Redundant Memory Copies High-Efficiency 4th Gen AMD Infinity Architecture. Low TCO with Unified Memory APU Package Next-Gen AMD Instinct™ APU ↑ Unified Memory (HBM)#74AMD COMPUTE GPU ARCHITECTURE ROADMAP DRIVING PERFORMANCE AND EFFICIENCY LEADERSHIP 2020 7nm AMD CDNA AMD INSTINCT MI100 6nm AMD CDNA 2 AMD INSTINCT MI200 SERIES All roadmaps are subject to change. 5nm AMD CDNA 3 "M1300" 2023#75OUR PATH FORWARD GPU TECHNOLOGY FOCUS Leadership AMD RDNAT / AMD CDNAT Architecture Roadmaps ▪ Advanced Process and Packaging Technologies ▪ Consistent Execution of Performance-Per-Watt Roadmap Expanding Open Software Ecosystems Including Al REE TO#76DISCLAIMER 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-18 Ⓒ2022 Advanced Micro Devices, Inc. All rights reserved. AMD, the AMD Arrow logo, AMD CDNA, AMD Instinct, AMD RDNA, Infinity Cache, and combinations thereof are trademarks of Advanced Micro Devices, Inc. Other product names used in this publication are for identification purposes only and may be trademarks of their respective owners.#77ENDNOTES RX-549 - Testing done by AMD performance labs 10/16/20, using Assassins Creed Odyssey (DX11, Ultra), Battlefield V (DX12, Ultra), Borderlands 3 (DX12, Ultra), Control (DX12, High), Death Stranding (DX12 Ultra), Division 2 (DX12, Ultra), F1 2020 (DX12, Ultra), Far Cry 5 (DX11, Ultra), Gears of War 5 (DX12, Ultra), Hitman 2 (DX12, Ultra), Horizon Zero Dawn (DX12, Ultra), Metro Exodus (DX12, Ultra), Resident Evil 3 (DX12, Ultra), Shadow of the Tomb Raider (DX12, Highest), Strange Brigade (DX12, Ultra), Total War Three Kingdoms (DX11, Ultra), Witcher 3 (DX11, Ultra no HairWorks) at 4K. System comprised of a Radeon RX 6800 XT with AMD Radeon Graphics driver 27.20.12031.1000 and an Radeon RX 5700 XT with AMD Radeon Graphics driver 26.20.13001.9005. Performance may vary. RX-558 - Testing done by AMD performance labs October 20 2020 on RX 6900 XT and RX 5700 XT (20.45-201013n driver), AMD Ryzen 9 5900X (3.70GHz) CPU, 16GB DDR4-3200MHz, Engineering AM4 motherboard, Win10 Pro 64. The following games were tested at 4k at max settings: Battlefield V DX11, Doom Eternal Vulkan, Forza DX12, Resident Evil 3 DX11, Shadow of the Tomb Raider DX12. Performance may vary. RX-558RS-462 - Testing conducted by AMD as of February 16, 2022, on a test system configured with a Ryzen 5 5600X CPU, 16GB DDR4, Radeon RX 6800 XT GPU, and Windows 10 Pro, with AMD Software: Adrenalin 22.3.1 vs. 20.12.2 at 4K, Max settings. Performance may vary. Games tested: Age of Empires 4, Borderlands 3, Call of Duty: Vanguard, Cyberpunk 2077, F1 2021, Far Cry 6, Forza Horizon 4, Guardians of the Galaxy, Hitman 3, The Medium, Metro Exodus Enhanced Edition, Myst, RDR2, Resident Evil Village, and PUBG. RX-785 Testing done by AMD performance labs May 31, 2022, on (11) AMD Radeon™ RX 6000 Series graphics cards, using systems configured with Ryzen™ 9 5900X and Ryzen™ 5 5600X CPUs, each with 16GB DDR4-3600MHz and AMD Smart Access Memory enabled, Win 10 Pro versus similarly configured systems with (11) Nvidia GeForce RTX 3000 Series, GeForce GTX 1650 and GTX 1050 Ti GPUs, each with ReBAR enabled. Performance tested across 20 games at 4K, 1440P and 1080P resolutions, at intended settings for each of the (22) AMD and NVIDIA GPUs. Performance per watt and per die size calculated using the total board power (TBP) and die sizes of the (22) individual AMD and NVIDIA GPUs over the average FPS scores. Performance may vary. RS-462 - Testing conducted by AMD as of February 16, 2022, on a test system configured with a Ryzen 5 5600X CPU, 16GB DDR4, Radeon RX 6800 XT GPU, and Windows 10 Pro, with AMD Software: Adrenalin 22.3.1 vs. 20.12.2 at 4K, Max settings. Performance may vary. Games tested: Age of Empires 4, Borderlands 3, Call of Duty: Vanguard, Cyberpunk 2077, F1 2021, Far Cry 6, Forza Horizon 4, Guardians of the Galaxy, Hitman 3, The Medium, Metro Exodus Enhanced Edition, Myst, RDR2, Resident Evil Village, and PUBG. GD-164 - Day-0 driver compatibility and feature availability depend on system manufacturer and/or packaged driver version. For the most up-to-date drivers, visit AMD.com MI200-05 - Measurements conducted by AMD Performance Labs as of Sep 10, 2021 on the AMD Instinct™ MI250X accelerator designed with AMD CDNA™ 2 6nm FinFET process technology with 1,700 MHz engine clock resulted in 47.9 TFLOPS peak double precision (FP64) floating-point, 383.0 TFLOPS peak Bfloat16 format (BF16) floating-point performance. The results calculated for AMD Instinct™ MI100 GPU designed with AMD CDNA 7nm FinFET process technology with 1,502 MHz engine clock resulted in 11.54 TFLOPS peak double precision (FP64) floating-point, 92.28 TFLOPS peak Bfloat16 format (BF16) performance.#78ENDNOTES MI200-01 - World's fastest data center GPU is the AMD Instinct™ MI250X. Calculations conducted by AMD Performance Labs as of Sep 15, 2021, for the AMD Instinct™ MI250X (128GB HBM2e OAM module) accelerator at 1,700 MHz peak boost engine clock resulted in 95.7 TFLOPS peak theoretical double precision (FP64 Matrix), 47.9 TFLOPS peak theoretical double precision (FP64), 95.7 TFLOPS peak theoretical single precision matrix (FP32 Matrix), 47.9 TFLOPS peak theoretical single precision (FP32), 383.0 TFLOPS peak theoretical half precision (FP16), and 383.0 TFLOPS peak theoretical Bfloat16 format precision (BF16) floating-point performance. Calculations conducted by AMD Performance Labs as of Sep 18, 2020 for the AMD Instinct™ MI100 (32GB HBM2 PCIe® card) accelerator at 1,502 MHz peak boost engine clock resulted in 11.54 TFLOPS peak theoretical double precision (FP64), 46.1 TFLOPS peak theoretical single precision matrix (FP32), 23.1 TFLOPS peak theoretical single precision (FP32), 184.6 TFLOPS peak theoretical half precision (FP16) floating-point performance. Published results on the NVidia Ampere A100 (80GB) GPU accelerator, boost engine clock of 1410 MHz, resulted in 19.5 TFLOPS peak double precision tensor cores (FP64 Tensor Core), 9.7 TFLOPS peak double precision (FP64). 19.5 TFLOPS peak single precision (FP32), 78 TFLOPS peak half precision (FP16), 312 TFLOPS peak half precision (FP16 Tensor Flow), 39 TFLOPS peak Bfloat 16 (BF16), 312 TFLOPS peak Bfloat16 format precision (BF16 Tensor Flow), theoretical floating-point performance. The TF32 data format is not IEEE compliant and not included in this comparison. https://www.nvidia.com/content/dam/en-zz/Solutions/Data-Center/nvidia-ampere-architecture-whitepaper.pdf, page 15, Table 1. MI200-26B - Testing Conducted by AMD performance lab as of 10/14/2021, on a single socket Optimized 3rd Gen AMD EPYC™ CPU (64) server, with 1x AMD Instinct™ MI250X OAM (128 GB HBM2e, 560W) GPU with AMD Infinity Fabric™ technology using benchmark HPL v2.3, plus AMD optimizations to HPL that are not yet upstream. vs. Nvidia DGX dual socket AMD EPYC 7742 (64C) @2.25GHz CPU server with 1x NVIDIA A100 SXM 80GB (400W) using benchmark HPL Nvidia container image 21.4-HPL Information on HPL: https://www.netlib.org/benchmark/hpl/Nvidia. HPL Container Detail: https://ngc.nvidia.com/catalog/containers/nvidia:hpc-benchmarks. Server manufacturers may vary configurations, yielding different results. Performance may vary based on use of latest drivers and optimizations. MI200-58 - Testing Conducted by AMD performance lab as of 5/25/2022, on a dual socket AMD EPYC™ 7200 Series CPUs (64C) server, with 8x AMD Instinct™ MI250X OAM (128 GB HBM2e, 500W) GPU with AMD Infinity Fabric™ technology using benchmark HPL-AI compiled with HIP version 5.1.20531.cacfa990, AMD clang version 14.0.0, OpenMPI 4.1.2. vs. dual socket AMD EPYC 7002 (64C) Series CPU (64) server with 8x NVIDIA A100 SXM 80GB (400W) using benchmark HPL-AI with CUDA 11.6. HPL-AI container 21.4-hpl Information on HPL-AI: https://hpl-ai.org/ AMD HPL-AI container detail: https://github.com/ROCmSoftwarePlatform/hpl-ai.git rev bae3342 Nvidia HPL-AI Container Detail: https://catalog.ngc.nvidia.com/orgs/nvidia/containers/hpc-benchmarks Nvidia container image 21.4-HPL Server manufacturers may vary configurations, yielding different results. Performance may vary based on use of latest drivers and optimizations. M1300-004 - Measurements by AMD Performance Labs June 4, 2022. MI250X (560W) FP16 (306.4 estimated delivered TFLOPS based on 80% of peak theoretical floating-point performance). M1300 FP8 performance based on preliminary estimates and expectations. M1300 TDP power based on preliminary projections. Actual results based on production silicon may vary.#79FINANCIAL ANALYST DAY 2022 together we advance_ Server CPU Leadership Dan McNamara Senior Vice President and General Manager, Server Business Unit AMD#80CAUTIONARY STATEMENT This presentation contains forward-looking statements concerning Advanced Micro Devices, Inc. (AMD) including, but not limited to, the timing, availability, features, functionality and expected benefits of AMD's adaptive computing products; AMD's momentum and TAM; AMD's silicon roadmap; and AMD's new revenue opportunities and path forward, 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.#81OUR JOURNEY OF RELENTLESS SERVER EXECUTION 1st Gen EPYC™ CPU IRRINE BISSISESS *ILE**INTER ITE 111111111111 1||||||||||| ************ ********* "Naples" Storage National Labs 2017 2nd Gen EPYC CPU ⠀⠀ HH 1113 BERRIRIBUR BEIRERIRINI ⠀⠀⠀⠀ (ELEFEITO000 OHINOOO000 "Rome" ********* High-End Enterprises Supercomputer Leading laaS, Search 2019 ## 11 is if ⠀⠀⠀ H As of May 30, 2022 TOP500 and GREEN500 lists. #*****### HEIKERERE 3rd Gen EPYC CPU "Milan" DEFETTIDD000 OFORDDIO00 2021 Commercial HPC Users Hybrid Cloud and HCI Leading GREEN500 and TOP500 Lists Collaboration, Social Media, eCommerce "Milan-X" 2022#821st Gen EPYC™ CPU OUR JOURNEY OF OPTIMIZED SERVER SOLUTIONS IRRINE BISSISESS FILEFANTEN Windows Server 2016 Certified ITE 111111111111 1||||||||||| 50+ Solutions ************ ***IEBTEITIE #hadoop vmware® VSAN Ready Nodes 2017 SUSE redhat MySQL 2nd Gen EPYC CPU ⠀⠀ 11361 1 ##1818181818 BERRIRIBUR BEIRERIRINI Ansys SYNOPSYS Windows Server 2019 Certified (ELEFEITO000 CHOROOO000 500+ Solutions ********* 2019 CLOUDERA A ALTAIR simplivity NUTANIX HPE DX Microsoft SQL Server DELL VXRAIL Lenovo ThinkAgile Appliance ## #*****### HEIKERERE HH 111111 ⠀⠀⠀ 3rd Gen EPYC CPU DELETEDDO000 OFORORDO00 2021 SIEMENS 1000+ Solutions mongoDB CISCO VALIDATED DESIGN ⠀⠀⠀RIS (111 FreeBSD ORACLE Tiger Graph docker Windows Server 2022 Certified splunk> ****** SHEARWATER cadence EMERSON citrix Azure Stack HCI 2022#83(-) Alibaba Cloud BAIDU AI CLOUD IBM Cloud ORACLE CLOUD OUTSTANDING MOMENTUM WITH 3RD GEN AMD EPYC™ CPUs aws Linked in . EQUINIX METAL NETFLIX salesforce UCLOUD 11 Microsoft Azure Meituan Cloud Google Cloud Tencent Cloud 火山引擎 Meta 1.6X VM density Up to 30% lower TCO AWE RESEARCH Nik hef CERN JÜLICH FORSCHUNGSZENTRUM SG NSCC 鹏城实验室通讯 Peng Cheng Laboratory Commumications 中国电信 CHINA TELECOM Lawrence Livermore National Laboratory OAK RIDGE National Laboratory PURDUE UNIVERSITY. UnitedHealthcare HPC FRONTIER NERSC 之江实验室 ZHEJIANG LAB PAWSEY supercomputing centre SDSC SAN DIEGO SUPERCOMPUTER CENTER 9 Up to 1.5X FP performance Up to 100% more simulations per day autos f5 INDUSTRIAL LIGHT & MAGIC NIO See endnotes EPYC-025, MLNTCO-020, MLNX-033A, MLNXTCO-001, MLN-101, MLNTCO-021. Results may not be typical. BASF Ford MathWorks DBS (GE) SAMSUNG mastercard PIXAR QUALCOMM WELLS FARGO Dropbox Goldman Sachs Enterprise Micron weta DIGITAL salesforce 1.6X on-line transactions per second Up to 41% lower TCO for HCI#84THE OPPORTUNITY AHEAD Cloud service proliferation Expanding use cases for Al and CPU inference Harnessing value in massive data Enterprise $10.5B Workload Diversity Based on AMD internal data $42B Long-Term TAM HPC $5.5B Cloud $26B#85dddd THE NEXT ERA OF LEADERSHIP 3 ¦ ¦ Highest Performing General Purpose Silicon O AMDET EPYC Optimized Silicon for Diverse Workloads Full Stack Solutions Accelerating Customer Time To Value Ecosystem Scale and Partnerships#862nd Gen EPYC™ CPU General Purpose "Rome" 2019 INDUSTRY LEADING OPTIMIZED SILICON 3rd Gen EPYC CPU 53 General Purpos Technical "Milan" "Milan-X" General Purpose "Genoa" 4th Gen EPYC CPU Cloud Native "Bergamo" All roadmaps are subject to change. Technical "Genoa-X" Telco "Siena" 5th Gen EPYC CPU "Turin" Family 2024#87AMDA EPYC >75% Faster Enterprise Java® Performance vs. 3rd Gen AMD EPYC™ CPU* Launching Q4 2022 "GENOA" EXTENDING COMPUTE LEADERSHIP Leadership Socket and Per-Core Performance Up to 96 "Zen 4" Cores in 5nm Leadership Memory Bandwidth and Capacity 12 Channels DDR5 Next Generation 1/0 PCIe® Gen 5 | Memory Expansion with CXL™ Advances in Confidential Computing Memory Encryption | Direct and CXL Attached *96c "Genoa" vs. EPYC 7763. See Endnotes SP5-003, SP5-005, SP5-007. Preliminary data and projections subject to change#88AMDA EPYC 2X Cloud Container Density vs. 3rd Gen AMD EPYC™ CPU* Available 1H 2023 "BERGAMO" CLOUD NATIVE LEADERSHIP Leadership Scale Out Performance Up to 128 "Zen 4c" Cores Highest Thread Density Up to 256 Threads SP5 Platform Compatible 12 Memory Channels | PCIe® Gen 5 "Zen 4" ISA Compatible No Software Port Required *128c "Bergamo" vs. EPYC 7763. See endnote SP5-006. Preliminary data and projections subject to change. See endnote BGM-001.#89CREMOON4131PCEtern CH ERPRICOM SOMERJALCES INTRODUCING "GENOA-X" Optimized for technical computing and databases Up to 96 "Zen 4" Cores in 5nm 1+ GB L3 Cache Per Socket Coming in 2023 See endnote GD-204. 9) 5G 0 O "SIENA" Optimized for intelligent edge and telco Up to 64 "Zen 4" Cores Optimized Performance- Lower Cost Platform Coming in 2023 Per-Watt#90OPTIMIZED SOLUTIONS FOR DATA AND AI Database and Analytics Industry Leading Solutions OEM ready nodes, reference designs and integrated OEM products ▪ 60+ world records See endnote EPYC-022B. Al Inference ZenDNN Inference Software Supports TensorFlow, PyTorch and ONNX Run Time Deployed in data centers today with leadership recommendation engine performance#91GROWING ECOSYSTEM WITH 4TH GEN AMD EPYC™ PROCESSORS aws Altair anjuna Ansys O BEAMR DASSAULT SYSTEMES MEMSQL Micron Quo DATASTAX: Google Cloud Radisys (-) Alibaba Cloud docker IBM Cloud Microsoft Red Hat CLOUD SERVICES TigerGraph Microsoft Azure ORACLE CLOUD BROADCOM cadence CANONICAL elastic esi Excelero get it right NEURAL redislabs EQUINIX TRANS WARP METAL SOLUTIONS Tencent Cloud mongoDB MySQL ROBIN SAMSUNG VELOCIX™ FreeBSD NETSCOUT. ScaleMP VERTICA Atos ASRock ASUS casa systems hadoop Inventec Lenovo MITAC NS NOKIA NUTANIX. WEKA CISCO ceph Juniper JMA NETWORKS SK hynix vmware* DELL Technologies msi PLATFORMS citrix CLOUDERA splunk> QCT 2X Number of Solutions KIOXIA ORACLE //// StorMagic Western Digital. MariaDB SUPERMICR TYANO PGS SOFTWARE SUSE FOXCONN GIGABYTE H3C wynn Couchbase MAVENIR PostgreSQL SYNOPSYS Hewlett Packard Enterprise wistron#92ACCELERATING OUR GROWTH 104 010818 Large and Growing Opportunity $42B TAM across key segments 13 4- AMD EPYC DA WAWE Product Leadership and Momentum Today Leading cloud, HPC and enterprise performance See MLN-016C, MLNX-047, MLN-092B. Expanding Portfolio of Silicon and Solutions Delivering Si and software optimized to the workload#93Endnotes EPYC-022B: For a complete list of world records see http://amd.com/worldrecords. EPYC-025: 64-core 3rd Gen EPYC 7xx3 CPUs compared to a maximum 40-core 3rd Gen Intel Xeon Platinum 8380. GD-204: "Technical Computing" or "Technical Computing Workloads" as defined by AMD can include: electronic design automation, computational fluid dynamics, finite element analysis, seismic tomography, weather forecasting, quantum mechanics, climate research, molecular modeling, or similar workloads SP5-003: Estimated SPECrate 2017_int_base comparison based on internal AMD estimates. The 4th Gen EPYC score is based on engineering projections and the EPYC 7763 system score is a measured SPECrate score, using two different AMD reference systems as of 4/5/2022. OEM published scores and performance/CPU TDP Watt will vary based on system configuration and use of production silicon. SPEC®, SPEC CPU®, and SPECrate® are registered trademarks of the Standard Performance Evaluation Corporation. See www.spec.org for more information. SP5-005: Server-side Java multiJVM workload demo comparison based on AMD measured testing as of 6/2/2022. Configurations: 2x 96-core AMD 4th Gen EPYC (pre-production silicon) on a reference system versus 2x 64-core EPYC 7763 on a reference system. Java version JDK18. OEM published scores will vary based on system configuration and use of production silicon. SP5-006: 128-core 4th Gen EPYC CPU compared to a maximum 64-core 3rd Gen EPYC 77x3 options for 2x the container/core density SP5-007: 4th Gen EPYC CPUs have 12 channels of DDR5 memory vs. 3rd Gen Intel Xeon Scalable CPUs with 8 channels of DDR4 memory and announced standard "Sapphire Rapids" with 8 channels of DDR5 memory MLN-016C: Results as of 03/28/2022 using SPECrate 2017_int_base. The top scores of EPYC 7773X and 7763 are statistically tied. The AMD EPYC 7763 scored 861, http://spec.org/cpu2017/results/res2021q4/cpu2017-20211121- 30148.html and 7773X scored 864, http://spec.org/cpu2017/results/res2022q1/cpu2017-20220228-31120.html which is higher than all other 2P scores published on the SPEC® website. SPEC®, SPECrate® and SPEC CPUⓇ are registered trademarks of the Standard Performance Evaluation Corporation. See www.spec.org for more information. MLN-101: SAP SD 2-tier comparison based on best performing systems published at www.sap.com/benchmarks as of 5/5/2021, 2x AMD EPYC™ 7763 scored 75,000 benchmark users (https://www.sap.com/dmc/benchmark/2021/Cert21021.pdf) which supports 56% more benchmark users than the top "Ice Lake" 3rd Gen 2x Intel® Xeon® Platinum 8380 that scored 48,000 benchmark users (https://www.sap.com/dmc/benchmark/2021/Cert21026.pdf). 2x AMD EPYC 7H12 scored 69,499 benchmark users, https://www.sap.com/dmc/benchmark/2020/Cert20023.pdf). SAP and SAP logo are the trademarks or registered trademarks of SAP SE (or an SAP affiliate company) in Germany and in several other countries. MLN-092B: SPECjbb® 2015-MultiJVM Critical comparison based on best performing 2P systems published at www.spec.org as of 10/26/2021, 2x AMD EPYC™ 7763 scored 313,824 SPECjbb 2015-MultiJVM Critical-jOPS (339,338 max- JOPS, https://www.spec.org/jbb2015/results/res2021q3/jbb2015-20210701-00688.html) which has 47% higher critical server-side Java operations than the top "Ice Lake" 2x Intel® Xeon® Platinum 8380 that scored 213,195 critical- JOPS (269,094 max-jOPS, https://www.spec.org/jbb2015/results/res2021q3/jbb2015-20210810-00701.html ). 2x AMD EPYC 7H12 scored 248,942 critical-jOPS (315,663 max-jOPS, http://spec.org/jbb2015/results/res2020q2/jbb2015- 20200423-00550.html ). SPEC® and SPECjbb® are registered trademarks of the Standard Performance Evaluation Corporation. See www.spec.org for more information. MLNX-033A: SPECrate®2017_fp_base comparison based on best performing systems published at www.spec.org as of 10/26/2021. Configurations: 2x AMD EPYC 7773X (745 SPECrate®2017_fp_base, http://spec.org/cpu2017/results/res2022q1/cpu2017-20220228-31118.html, $17600 1Ku price total, 560W total TDP) versus 2x Intel Xeon Platinum 8380 (489 SPECrate 2017_fp_base, http://spec.org/cpu2017/results/res2021q2/cpu2017-20210521-26361.html, $17332 1Ku price total, 540W total TDP) for 1.52x the performance at 1.5x the score per total CPU $ and 1.47x the performance/Watt. AMD 1Ku pricing and Intel ARK.intel.com specifications and pricing as of 3/15/22. SPEC®, SPEC CPU®, and SPECrate® are registered trademarks of the Standard Performance Evaluation Corporation. See www.spec.org for more information. MLNX-047: SPECrate@2017_fp_base comparison based on best performing systems published at www.spec.org as of 03/28/2022. A server with 2x AMD EPYC 7773X (745 SPECrate 2017_fp_base, http://spec.org/cpu2017/results/res2022q1/cpu2017-20220228-31118.html) is higher than all other 2P SPECrate 2017_fp_base scores. SPEC®, SPEC CPU, and SPECrate® are registered trademarks of the Standard Performance Evaluation Corporation. see www.spec.org for more information.#94Endnotes MLNTCO-020: 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 Bare Metal Server Greenhouse Gas Emissions TCO (total cost of ownership) Estimator Tool compares the selected AMD EPYC™ and Intel® Xeon® CPU based server solutions required to deliver a TOTAL_PERFORMANCE of 10000 units of integer performance based on the published scores for Intel Xeon and AMD EPYC CPU based servers. This estimation reflects a 3-year time frame. This analysis compares a 2P AMD EPYC EPYC_7763 powered server with a SPECrate 2017_int_base score of 861, https://spec.org/cpu2017/results/res2021q4/cpu2017- 20211121-30148.pdf; compared to a 2P Intel Xeon Platinum_8380 based server with a SPECrate 2017_int_base score of 602, https://spec.org/cpu2017/results/res2021q2/cpu2017-20210521-26364.pdf. Both AMD EPYC and Intel based servers use the same cost for the following elements of the analysis: server chassis size of 2RU at a cost of $2500 per chassis; internal storage $380; physical servers managed per admin: 30; fully burdened cost per admin $110500; server rack size of 42; space allowance per rack of 27 sq feet; monthly cost of data center space $20 per sq foot; cost per kW for power $0.12; power drop per rack of 8kW; and a PUE (power usage effectiveness) of 1.7. The EPYC powered solution is estimated to take: 12 total 2P EPYC_7763 powered servers at a hardware only acquisition cost of $23748 per server, which includes $7890 per CPU, total system memory of 1024GB, which is 8GB of memory / core and a total system memory cost of $5088; internal storage cost of $380. The total estimated AMD EPYC hardware acquisition cost for this solution is $284976. Each server draws ~755.1412kWhr per month. For the 3 years of this EPYC powered solution analysis the: total solution power cost is ~$66548.88 which includes the PUE factor; the total admin cost is ~$132600, and the total real estate cost is ~$38880, using 2 racks. The total 3 TCO estimate for the AMD solution is $523004.88. The Intel based solution is estimated to take 17 total 2P Platinum_8380 powered servers at a hardware only acquisition cost of $24206 per server, which includes $8099 per CPU, total system memory of 1024GB, which is 12.8GB of memory / core and a total system memory cost of $5088; internal storage cost of $380. The total estimated Intel hardware acquisition cost for this solution is $411502. Each server draws ~751.4912kWhr per month. For the 3 years of this Intel based solution analysis the: total solution power cost is ~$93822.048 which includes the PUE factor; the total admin cost is ~$187851, and the total real estate cost is ~$58320 using 3 racks. The total 3 TCO estimate for the Intel solution is $751495.048. AMD EPYC powered servers have a $228490 lower 3-year TCO. Delivering 10000 estimated score of SPECrate 2017_int_base performance produces the following estimated results: the AMD EPYC solution requires 29% fewer servers [1-(AMD server count / Intel server count)]; 33% less space [1-(AMD rack count / Intel rack count)]; 29% less power [1-(AMD power cost / Intel power cost)]; providing a 30% lower 3-year TCO [1-(AMD TCO / Intel TCO)]. delivering ~98 or ~1% Better w/ AMD SPECrate 2017_int_base solution score AMD EPYC_7763 powered servers save ~227276.4kWh of electricity for the 3 years of this analysis. Leveraging this data, using the Country / Region specific electricity factors from the '2020 Grid Electricity Emissions Factors v1.4 - September 2020', and the United States Environmental Protection Agency 'Greenhouse Gas Equivalencies Calculator', the AMD EPYC powered server saves ~103.01 Metric Tons of CO2 equivalents. This results in the following estimated savings based on United States data, Emissions Avoided equivalent to one of the following: 22 USA Passenger Cars Not Driven for 1 year; or; or 7.45 USA Passenger Cars Not Driven Annually; or; or 11640 Gallons of Gasoline Not Used; or; oror Carbon Sequestered equivalent to: 1700 Tree Seedlings Grown for 10 years in USA; or; or 41.2 Acres of USA Forests Annually. The 2020 Grid Electricity Emissions Factors v1.4 - September 2020 data used in this analysis can be found at https://www.carbonfootprint.com/docs/2020_09_emissions_factors_sources_for_2020_electricity_v14.pdf and the US EPA Greenhouse Gas Equivalencies Calculator used in this analysis can be found athttps://www.epa.gov/energy/greenhouse-gas-equivalencies-calculator AMD processor pricing based on 1KU price as of Sept 2021. Intel® Xeon® Scalable Gen 1 and Gen 2 CPU data and pricing from https://ark.intel.com as of September 2021. Intel Xeon Gen3 Scalable Ice Lake pricing and data from https://newsroom.intel.com/wp-content/uploads/sites/11/2021/05/3rd-Gen-Intel-Xeon-Scalable- Processor-SKU-Stack-with-RCP.pdf on 09/01/2021. All pricing is in USD. SPECrate scores as of Jan 14, 2022. SPEC®, SPECrate® and SPEC CPU are registered trademarks of the Standard Performance Evaluation Corporation. See www.spec.org for more information. AMD EPYC performance numbers based on the identified benchmark reported scores or the user provided score where indicated. Product and company names are for informational purposes only and may be trademarks of their respective owners. Results generated by: AMD EPYC™ BARE METAL SERVER and GREENHOUSE GAS EMISSIONS TCO ESTIMATION TOOL; VERSION 4.2#95Endnotes MLNTCO-021: 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 EPYC™ SERVER VIRTUALIZATION and GREENHOUSE GAS EMISSIONS TCO ESTIMATION TOOL tool compares the 2P AMD EPYC™ and the 2P Intel® Xeon® server solutions required to deliver 1200 total virtual machines (VM), requiring 1 core and 8GB of memory per VM. The analysis includes both hardware and virtualization software components. Hardware costs (CPU + memory + storage + chassis): The 2P AMD 64 core EPYC_7713 processor used in this solution analysis provides 128 total cores per server, each processor cost $7060 and the server uses 32 x 32GB DIMN to achieve the minimum required memory footprint, in a 1RU server chassis that cost $2200, and requires 1 server racks. The AMD solution has a total estimated hardware acquisition cost of $217880. The 40 core Intel Xeon Platinum_8380 processor used in this solution analysis provides 80 total cores per server. Each processor cost $8666 and the server uses 16 x 64GB DIMMS to achieve the minimum required memory footprint, in a 2RU server chassis that cost $2500 and requires 2 server racks. The Intel solution has a total estimated hardware acquisition cost of $390060. OPERATING COSTS: The core assumptions for this analysis are as follows: Cost of power @ $0.12 with kwatts (kW) of power to each rack and a PUE (power usage effectiveness) of 1.7 and a server rack size of 42RU. Each server has 1 hard drives drawing 3 watts each. Server Admin annual salary is $85000 managing 30 physical servers with a salary burden rate of 30%. The VM Admin salary is $85000, with a burden rate of 30% and managing 400 VMS.AMD has estimated OpEx costs as follows: a hardware admin cost of $110500, a real estate cost of $19440, and a power cost of $40208.4, for a total estimated 3 year TCO cost (hardware cost and operating expense) of $388028 with AMD. Estimated OpEx costs for Intel are: hardware admin cost of $165750, real estate cost of $38880, and power cost of $58704.HARDWARE TCO: This is the CapEx and OpEx directly associated with the hardware. The AMD EPYC_7713 solution requires 10 - 2P servers with a CapEx of $217880 with a total estimated 3 year TCO cost (CapEx plus OpEx) of $388028. The Intel Platinum_8380 processor requires 15 - 2P servers with a CapEx of $390060 with a total estimated 3 year TCO cost (CapEx plus OpEx) of $653394. The AMD solution has an estimated 41% lower hardware TCO for this virtualization solution, 1 - ($388028 ÷ $653394) = 41%, than the Intel solution.VIRTUALIZATION TCO: Analysis is based on the following estimates: 3 year Virtualization (hardware, operating, and software cost) for the Intel solution is $2005974 and $1621248 for the AMD solution. This means that the AMD solution is ~19% less expensive over three years. 1 -($1621248 ÷ $2005974) = 19%. The EPYC solution 1st year TCO is $844816 and the Intel 1st year TCO is $1167418. The AMD solution 1st year TCO per VM of $704.01 where the Intel 1st yr. solution is $972.85. The AMD 1st year TCO per VM is $268.83, or ~28% lower than Intel. The 1st year TCO per VM is calculated by taking the 1-year TCO (hardware, software, and 1st year OpEx) and dividing it by the total number of VMs. The virtualization software used in this analysis is VMware with a VMware vSphere Enterprise Plus w/ Production support license. This analysis uses license pricing of $5968 per Socket + Core with 3 year support. More information on VMware software can be found @ https://store-us.vmware.com/vmware-vsphere-enterprise-plus-284281000.html. For 1200 VMS with 1 core(s) per VM, and 8 GB of memory per VM, the Intel Platinum_8380 processor requires 15 servers, and 60 licenses. The AMD EPYC_7713 solution requires 10 servers and 40 licenses. The AMD solution requires 33% fewer servers than the Intel solution. The AMD server and virtualization software license cost are $456600, and the Intel cost are $748140. Hardware and virtualization cost are ~$291540 or ~39% Lower w/ AMD. AMD EPYC_7713 powered servers save ~154132.2kWh of electricity for the 3 years of this analysis. Leveraging this data, using the Country/Region specific electricity factors from the '2020 Grid Electricity Emissions Factors v1.4 - September 2020', and the United States Environmental Protection Agency 'Greenhouse Gas Equivalencies Calculator', the AMD EPYC powered server saves ~69.86 Metric Tons of CO2 equivalents. This results in the following estimated savings based on United States data, Greenhouse Gas Emissions Avoided of one of the following: 15 USA Passenger Cars Not Driven for 1 year; or; 5 USA Passenger Cars Not Driven Annually; or; 173382 Miles Driven by Avg Passenger Car; or; or CO2 Emissions Avoided from: 7894 Gallons of Gasoline Not Used; or; 77261 Pounds of Coal Not Burned in USA; or; 9 USA Homes' Electricity Use for 1 year; or; 3 USA Homes' Electricity Use Annually; or;or Carbon Sequestered equivalent to: 1153 Tree Seedlings Grown for 10 years in USA; or; 84 Acres of USA Forests in 1 year; or; 27.94 Acres of USA Forests Annually. The 2020 Grid Electricity Emissions Factors v1.4 - September 2020 data used in this analysis can be found at https://www.carbon footprint.com/docs/2020_09_emissions_factors_sources_for_2020_electricity_v14.pdf and the US EPA Greenhouse Gas Equivalencies Calculator used in this analysis can be found athttps://www.epa.gov/energy/greenhouse-gas-equivalencies-calculatorVirtualization software pricing sourced online as of 09/14/2021. Third party names are for informational purposes only and may be trademarks of their respective owners. All pricing is in USD.AMD CPU pricing based on 1KU price as of January 2022. Intel® Xeon® Scalable CPU data and pricing from https://ark.intel.com as of January 2022. All pricing is in USD. Results generated by: AMD EPYC™ SERVER VIRTUALIZATION and GREENHOUSE GAS EMISSIONS TCO ESTIMATION TOOL - V10.13#96Endnotes MLNXTCO-001: 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 EPYC™ AMD 3D V-Cache™ VALUE ANALYSIS & GHG TOOL compares the selected AMD EPYC™ and Intel® Xeon® CPU based server solutions required to deliver a TOTAL_PERFORMANCE of 4600 jobs per day with Ansys® cfx-50 using the performance scores in this analysis for Intel Xeon and AMD EPYC CPU based servers. This estimation reflects a 3-year time frame. This analysis compares a 2P AMD EPYC_7573X (32 cores per CPU) powered server with a Ansys® cfx-50 jobs per day of 484.67; to a 2P Intel Platinum_8362 (32 cores per CPU) based server with a Ansys cfx-50 jobs per day of 239.51. A server powered by the EPYC_7573X can deliver up to 102% more jobs per day than the Platinum_8362 based server. Both AMD EPYC and Intel based servers use the same cost for the following elements of the analysis: server chassis size of 2RU at a cost of $2500 per chassis; internal storage $380; physical servers managed per admin: 30; fully burdened cost per admin $110500; server rack size of 42; space allowance per rack of 27 sq feet; monthly cost of data center space $20 per sq foot; cost per kW for power $0.12; power drop per rack of 12kW; and a PUE (power usage effectiveness) of 1.7. The AMD EPYC powered solution is estimated to take 10 total 2P EPYC_7573X powered servers at a hardware only acquisition cost of $19564 per server, which includes $5590 per CPU, total system memory of 1024GB, which is 16GB of memory / core and a total system memory cost of $5504; internal storage cost of $380. The total estimated AMD EPYC hardware acquisition cost for this solution is $195640. Each server draws ~754kWhr per month. For the 3 years of this analysis the: EPYC total solution power cost is ~$55406 which includes the PUE factor; the total admin cost is ~$110499, and the total real estate cost is ~$19440 using 1 rack(s). The total 3-year TCO estimate for the EPYC solution is $370802. The Intel based solution is estimated to take 20 total 2P Platinum_8362 powered servers at a hardware only acquisition cost of $20080 per server, which includes $5828 per CPU, total system memory of 1024GB, which is 16GB of memory / core and a total system memory cost of $5504; internal storage cost of $380. The total estimated Intel hardware acquisition cost for this solution is $401600. Each server draws ~743kWhr per month. For the 3 years of this analysis the: Intel total solution power cost is ~$109203 which includes the PUE factor; the total admin cost is ~$221001, and the total real estate cost is ~$38880 using 2 rack(s). The total 3-year TCO estimate for the Intel solution is $750318. AMD EPYC powered servers have a $379516 or 51% lower 3-year TCO.Delivering a minimum score of 4600 for Ansys cfx-50 produces the following estimated results: the EPYC_7573X solution requires 50% fewer servers; takes 50% less RU space; 49% less power. AMD EPYC_7573X powered servers save ~448315 kWh of electricity for the 3-years of this analysis. Leveraging this data, using the Country/Region specific electricity factors from the '2020 Grid Electricity Emissions Factors v1.4 - September 2020', and the United States Environmental Protection Agency 'Greenhouse Gas Equivalencies Calculator', the AMD EPYC powered server saves ~203.19 Metric Tons of CO2 equivalents. This results in the following estimated savings based on United States data, for any one of the following: Greenhouse Gas Emissions Avoided: 44 USA Passenger Cars Not Driven for 1 year; or 15 USA Passenger Cars Not Driven Annually; or 510604 Miles Driven by Avg Passenger Car; oror CO2 Emissions Avoided from: 22960 Gallons of Gasoline Not Used; or 224520 Pounds of Coal Not Burned in USA; or 37 USA Homes' Electricity Use for 1 year; or 12 USA Homes' Electricity Use Annually; oror Carbon Sequestered equivalent to: 3353 Tree Seedlings Grown for 10 years in USA; or 244 Acres of USA Forests in 1 year; or 81.27 Acres of USA Forests Annually. The 2020 Grid Electricity Emissions Factors v1.4 - September 2020 data used in this analysis can be found at https://www.carbon footprint.com/docs/2020_09_emissions_factors_sources_for_2020_electricity_v14.pdf and the US EPA Greenhouse Gas Equivalencies Calculator used in this analysis can be found at https://www.epa.gov/energy/greenhouse-gas-equivalencies-calculator.Pricing per CPU is 1kU pricing for AMD and Intel published pricing at https://ark.intel.com/, January 2022. All pricing is in USD. All performance numbers are based on AMD internal testing, February 2022. AMD tests were run with pre-production B1 CPUs on AMD reference platforms. Intel tests were run on production platforms. Product and company names are for informational purposes only and may be trademarks of their respective owners. Results generated by the: AMD EPYC™ AMD 3D V-Cache" VALUE ANALYSIS & GHG TOOL: v3.10#97FINANCIAL ANALYST DAY 2022 together we advance_ AMD Building on Data Center Leadership..... Forrest Norrod Senior Vice President and General Manager, Data Center Solutions Business Group#98Cautionary Statement This presentation contains forward-looking statements concerning Advanced Micro Devices, Inc. (AMD) including, but not limited to, the timing, availability, features, functionality and expected benefits of AMD's data center products; trends and TAM of the data center market; and AMD's data center roadmap, 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.#99Fwww Cloud MEGATRENDS DRIVING COMPUTE IN DATA CENTERS BAR ********** AI AI HPC 19 20 Data Analytics Video Streaming#100AMD EPYC Server CPUs POWERING THE MODERN DATA CENTER PERFORMANCE AMD INSTINCT GPU Accelerators AMD PENSANDO XILINX SmartNICs and DPUs EFFICIENCY isa XILINX XILINX TM SECURITY TM FPGAs and Adaptive SoCs#101EXPANDING DATA CENTER TAM Server CPU $42B FPGA and Adaptive SoC $13B $125B LONG-TERM TAM Based on AMD internal data $64B GPU & AI $6B DPU & Infrastructure Acceleration#102OUR JOURNEY IN GPU ACCELERATION 2020 AMD Instinct™ MI100 AMD CDNA™ Ecosystem Growth First purpose-built GPU architecture for the data center AMD Instinct™ MI200 AMD CDNA™M 2 TM Driving HPC and Al to a New Frontier First multi-die data center GPU expands scientific discovery and brings choice to Al training Roadmaps Subject to Change AMD INSTINCT AMD Instinct™ M1300 AMD CDNA™ 3 Data Center APU Breakthrough architecture designed for leadership efficiency and performance for HPC and Al 2023#103AMD INSTINCT™ MI200 DEFINING LEADERSHIP IN HPC Performance Metric FP64 Vector FP64 Matrix FP32 Vector FP16, BF16 Memory Size Memory Bandwidth Open MM HPL MI250X Advantage vs A100 4.9X 4.9X 2.5X 1.2X 1.6X 1.6X ~ 2.4X ~ 2.8X 3RD GEN AMD INFINITY ARCHITECTURE CPU & GPU Memory Coherence Exceptional System Bandwidth & Performance See Endnotes MI200-01, MI200-07, M1200-24A, MI200-26B#104AMD EPYC INSTINCT LEADING THE EXASCALE ERA Powering World's #1 Supercomputer First to break Exascale barrier Powering World's #1 Green Supercomputer 8 of top 10 most efficient systems rely on AMD Powering World's #1 Al Supercomputer More than 3X the previous record holder 95% growth in TOP500 systems Year-over-Year Powering more than half of all new systems TOP500, Green500, and HPL-Al lists, as of May 30,2022 CRAY CRAY 11 1 OAK RIDGE Na tional Laboratory U.S. DEPARTMENT OF ENERGY C.] Hewlett Packard Enterprise AMD RONTIER FROM#105AMD INSTINCT ~1.3X GPT2-large Al Training Performance ~1.4X Res Net BRINGING CHOICE TO AL TRAINING Nvidia A100 SXM ~1.4X DenseNet AMD Instinct™ MI250X OAM See Endnotes M1200-57, MI200-59, MI200-61, MI200-63 ~1.2X BERT AMDA INSTINCT#106AMD INSTINCT Microsoft "As part of our long-term partnership with AMD, I'm excited to share that Azure will be the first public cloud to deploy clusters of AMD's flagship MI200 GPUs for large scale Al training. We've already started testing these clusters using some of our own Al workloads with great performance." Kevin Scott Executive Vice President and Chief Technology Officer, Microsoft O MAD CEPECTE T - LIO Pist Cin CHET TIE nat KE HEIDKI N O HILL AMID INSTINCT AL **** O 11 :: ==== 77 11#107AMD INSTINCT™ M1300 THE WORLD'S FIRST DATA CENTER APU ■ 4th Gen AMD Infinity Architecture: AMD CDNA™ 3 and EPYC™ CPU "Zen 4" Together CPU and GPU cores share a unified on-package pool of memory Groundbreaking 3D Packaging CPU | GPU | Cache | HBM Designed for Leadership Memory Bandwidth and Application Latency APU Architecture Designed for Power Savings Compared to Discrete Implementation Available 2023 AMDO INSTINCT >8X Expected Al Training Performance vs. M1250X See Endnote M1300-03. Preliminary data and projections, subject to change.#108#109SCALING THE SOFTWARE DEFINED DATA CENTER INFRASTRUCTURE ACCELERATION AND NETWORKING ARE ESSENTIAL Cloud and Virtualization Cloud services overhead can consume up to 30% of CPU cores CPU X 15.4561.10 Shift of Compute to Edge Extend visibility and management beyond the data center boundary AMD Estimate 90+ 90+ Security of Data Traffic Secure data at rest and in flight at all endpoints#110AMD NETWORKING TECHNOLOGY Solarflare™ Ultra Low Latency NICs Ubiquitous in Financial Exchanges Crucial for High Frequency Trading Nasdaq NYSE Euronext SEERA **** SN1000 ALVEO. Alveo™ FPGA-Based Adaptive Acceleration Programmable Hardware for Custom Network Functions Extreme Packet Processing Rates Alibaba Cloud Tencent Cloud PENSANDO Pensando™M P4 Packet Processing Engines Tens of Millions of Network Flows Multiple Concurrent Services at Line Rate Goldman Sachs ORACLE#111799 MOND EXILINX 回回 TJENJ ONSE ERCINE #un Camam MU:UNZI MEME mm CRNO B ALVEO™ ADAPTIVE NETWORK ACCELERATION Shipping to Hyperscale Customers Accelerates Custom and Evolving Network Functions Extends Confidential Computing to Network Interface 2 x 200G | 400M Packets per Second Next Generation in 2024#112PENSANDO™ WORLD'S MOST INTELLIGENT DPU 144 P4 Packet Processors Fully Programmable Control, Data, and Management Planes Supports Tens of Millions of Network Flows Concurrent Services at Line Rate Performance Network Security | Storage | Telemetry 2nd Generation | 2x 200G | 7nm In Production Today THE B HELL TURI IRUNE 199 LEEE "ELBA" DPU#113PENSANDOT SOFTWARE TURNKEY INFRASTRUCTURE SOLUTION FOR THE SOFTWARE DEFINED DATA CENTER ▪ Based on Open Standards and APIs Easily Incorporate and Accelerate New Services ▪ "Zero Trust Security" Throughout Works with Existing Management Tools ▪ Solutions Deployed Today Networking Management Encryption Storage: Firewall Telemetry#114DPU ACCELERATION ACROSS THE DATA CENTER PENSANDO DELL Technologies TAR POCE P4 SmartNIC Hewlett Packard Enterprise ---- Hapan aruba Smart Switch IBM. Microsoft NetApp Storage NetApp E ORACLE vmware#115AMD EPYC AMD PENSANDO AMD EPYC AMD PENSANDO AMD EPYC AMD PENSANDO AMD INSTINCT { { AMD SECURING THE DATA CENTER COMPREHENSIVE SECURITY FEATURES FROM CORE TO EDGE XILINX ALVEO XILINX XILINX XILINX VERSAL XILINX XILINX VERSAL Distributed Security Services Extend the perimeter Edge and Beyond Smart Switch & Appliance Help secure east-west traffic and legacy endpoints Infrastructure Compute Node 9 Appliance Advanced SmartNIC Help secure every endpoint Confidential Compute Help secure application environment Root of Trust Help secure CPU at boot |||||||| i ‒‒‒‒‒‒‒‒‒‒‒‒ ARRATIA $ Smart Switch Modern Server Legacy Server 1 1 0 1 1 0110 0 0 1 0 0 1 0 0 1 1 0 0 1 0 0 100 0 1 1 0 1 1 0 1 1 1 1 0 1 1 0 1 1 0 0 0 1 0 0 1 0 0 1 1 1 0 1 1 0 1 1 0 0 0 1 0 0 1 0 0 1 0 1 1 0 1 1 011 1 1 1 1 1 1 1 1 1 001001001 Advanced Telemetry Monitor and detect threats in real time 010010010 1 1 0 1 1 0110 0 0 1 0 0 1001 0 0 1 0 0 1 0 0 1 1 1 1 1 1 1 1 1 1 0 1 1 0 1 1 011 001001001 111111111 elastic FORTINET. Guardicore splunk>#116#117OUR PATH FORWARD UNMATCHED DATA CENTER TECHNOLOGY Leadership Performance ▪ World-Class Efficiency Comprehensive Security Features AMD EPYC INSTINCT PENSANDO AMD AMD XILINX { ALVEO. VERSAL. {#118Endnotes MI200-01 - World's fastest data center GPU is the AMD Instinct™ MI250X. Calculations conducted by AMD Performance Labs as of Sep 15, 2021, for the AMD Instinct™ MI250X (128GB HBM2e OAM module) accelerator at 1,700 MHz peak boost engine clock resulted in 95.7 TFLOPS peak theoretical double precision (FP64 Matrix), 47.9 TFLOPS peak theoretical double precision (FP64), 95.7 TFLOPS peak theoretical single precision matrix (FP32 Matrix), 47.9 TFLOPS peak theoretical single precision (FP32), 383.0 TFLOPS peak theoretical half precision (FP16), and 383.0 TFLOPS peak theoretical Bfloat16 format precision (BF16) floating-point performance. Calculations conducted by AMD Performance Labs as of Sep 18, 2020 for the AMD Instinct™ MI100 (32GB HBM2 PCIe card) accelerator at 1,502 MHz peak boost engine clock resulted in 11.54 TFLOPS peak theoretical double precision (FP64), 46.1 TFLOPS peak theoretical single precision matrix (FP32), 23.1 TFLOPS peak theoretical single precision (FP32), 184.6 TFLOPS peak theoretical half precision (FP16) floating-point performance. Published results on the NVidia Ampere A100 (80GB) GPU accelerator, boost engine clock of 1410 MHz, resulted in 19.5 TFLOPS peak double precision tensor cores (FP64 Tensor Core), 9.7 TFLOPS peak double precision (FP64). 19.5 TFLOPS peak single precision (FP32), 78 TFLOPS peak half precision (FP16), 312 TFLOPS peak half precision (FP16 Tensor Flow), 39 TFLOPS peak Bfloat 16 (BF16), 312 TFLOPS peak Bfloat16 format precision (BF16 Tensor Flow), theoretical floating-point performance. The TF32 data format is not IEEE compliant and not included in this comparison. https://www.nvidia.com/content/dam/en-zz/Solutions/Data-Center/nvidia-ampere-architecture-whitepaper.pdf, page 15, Table 1. MI200-01 MI200-07 - Calculations conducted by AMD Performance Labs as of Sep 21, 2021, for the AMD Instinct™ MI250X and MI250 (128GB HBM2e) OAM accelerators designed with AMD CDNA™ 2 6nm FinFet process technology at 1,600 MHz peak memory clock resulted in 128GB HBM2e memory capacity and 3.2768 TFLOPS peak theoretical memory bandwidth performance. MI250/MI250X memory bus interface is 4,096 bits times 2 die and memory data rate is 3.20 Gbps for total memory bandwidth of 3.2768 TB/s ((3.20 Gbps*(4,096 bits*2))/8). The highest published results on the NVidia Ampere A100 (80GB) SXM GPU accelerator resulted in 80GB HBM2e memory capacity and 2.039 TB/s GPU memory bandwidth performance. https://www.nvidia.com/content/dam/en-zz/Solutions/Data-Center/a100/pdf/nvidia-a100-datasheet-us-nvidia-1758950-r4-web.pdf MI200-07 MI200-24A - Testing Conducted by AMD performance lab as of 10/12/2021, on a single socket Optimized 3rd Gen AMD EPYC™ CPU server with 1x AMD Instinct™ MI250X OAM (128 GB HBM2e) 560W GPU with AMD Infinity Fabric™ technology using benchmark OpenMM_amoebagk v7.6.0, (converted to HIP) and run at double precision (8 simulations *10,000 steps) plus AMD optimizations to OpenMM_amoebagk that are not yet upstream resulted in a median score of 387.0 seconds or 223.2558 NS/Day Vs. Nvidia DGX dual socket AMD EPYC [email protected] CPU server with 1x NVIDIA A100 SXM 80GB (400W) using benchmark OpenMM_amoebagk v7.6.0, run at double precision (8 simulations *10,000 steps) with CUDA code version 11.4 resulted in a median score of 921.0 seconds or 93.8111 NS/Day. Information on OpenMM: https://openmm.org/ Server manufacturers may vary configurations, yielding different results. Performance may vary based on use of latest drivers and optimizations MI200-24A MI200-26B - Testing Conducted by AMD performance lab as of 10/14/2021, on a single socket Optimized 3rd Gen AMD EPYC™ CPU (64) server, with 1x AMD Instinct™ MI250X OAM (128 GB HBM2e, 560W) GPU with AMD Infinity Fabric™ technology using benchmark HPL v2.3, plus AMD optimizations to HPL that are not yet upstream. Vs. Nvidia DGX dual socket AMD EPYC 7742 (64C) @2.25GHz CPU server with 1x NVIDIA A100 SXM 80GB (400W) using benchmark HPL Nvidia container image 21.4-HPL. Information on HPL: https://www.netlib.org/benchmark/hpl/ Nvidia HPL Container Detail: https://ngc.nvidia.com/catalog/containers/nvidia:hpc-benchmarks Server manufacturers may vary configurations, yielding different results. Performance may vary based on use of latest drivers and optimizations MI200-26B MI200-57- Testing Conducted by AMD performance lab as of 25/5/2022 using SuperBench v 0.4.0 benchmark, GPT2-Large. EPYC/Instinct system: Dual socket, 64 core, 2nd Gen AMD EPYC™ 7002 Series CPU powered server with 8x AMD Instinct™ MI250X OAM (128 GB HBM2e) 500W GPUs with AMD Infinity Fabric™ technology. Benchmark: GPT2-Large with AMD | Microsoft optimized batch sizes tuned for GPT2-Large results for system configurations that are not yet available upstream. Benchmark Results: GPT2-large resulted in a median throughput of 8x MI250X = 761.08 Samples (Throughput)/ sec. Training model separates copies of model on each GPU; total system throughput obtained by calculating the sum of the throughput obtained on each GPU. Vs. EPYC/Nvidia system: NVIDIA DGXA100, Dual AMD EPYC 7002 Series CPUs with 8x NVIDIA A100 SXM 80GB (400W) Benchmark: GPT2-Large Commit(Container): superbench/superbench:v0.4.0- cuda11.1.1 from here: (https://hub.docker.com/r/superbench/superbench) Benchmark Results: GPT2-large resulted in a median throughput of 8x A100 = 589.435 Samples (Throughput)/ sec. Server manufacturers may vary configurations, yielding different results. Performance may vary based on use of latest drivers and optimizations MI200-57 MI200-59 - Testing Conducted by AMD performance lab as of 25/5/2022 using SuperBench v 0.4.0 benchmark, DenseNet 169/201, Framework PyTorch 1.9. EPYC/Instinct system: Dual socket, 64 core, 2nd Gen AMD EPYC™ 7002 Series CPU powered server with 8x AMD Instinct™ M1250X OAM (128 GB HBM2e) 500W GPUs with AMD Infinity Fabric™ technology, ROCm™ 5.1.0 Benchmark: DenseNet model (Median scores of DenseNet169, DenseNet201 datasets) with AMD | Microsoft optimized batch sizes tuned for DenseNet results for system configurations that are not yet available upstream. Commit(Container): computecqe/superbench:rocm5.1.3_superbench04 from here Benchmark Results: DenseNet testing resulted in median throughput scores of 8x MI250X: DenseNet169 = 6567.769, DenseNet201 = 5254.561 Samples (Throughput)/ sec. Training model separates copies of model on each GPU; total system throughput obtained by calculating the sum of the throughput obtained on each GPU. Vs. EPYC/Nvidia system: NVIDIA DGXA100, Dual AMD EPYC 7002 Series CPUs with 8x NVIDIA A100 SXM 80GB (400W), CUDA 11.6 and Driver Version 510.47.03, Commit(Container): superbench/superbench:v0.4.0-cuda11.1.1 from here: (https://hub.docker.com/r/superbench/superbench) Benchmark Results: DenseNet testing resulted in median throughput of 8x A100: DenseNet169 = 4712.705, DenseNet201 = 3877.668 Samples (Throughput)/ sec. Details on SuperBench found here Server manufacturers may vary configurations, yielding different results. Performance may vary based on use of latest drivers and optimizations MI200-59 MI200-61- Testing Conducted by AMD performance lab as of 5/25/2022 using SuperBench v 0.4.0 benchmark, Bert-Base. EPYC/Instinct system: Dual socket, 64 core, 2nd Gen AMD EPYC™ 7002 Series CPU powered server with 8x AMD Instinct™ MI250X OAM (128 GB HBM2e) 500W GPUs with AMD Infinity Fabric™ technology, ROCm™ 5.1.0, PyTorch 1.9 Benchmark: Bert-Base with AMD | Microsoft optimized batch sizes tuned for Bert-Base results for system configurations that are not yet available upstream. Commit (Container): computecqe/superbench:rocm5.1.3_superbench04 from here Benchmark Results: Bert-Base resulted in a median throughput of 8x MI250X = 6230.021 Samples (Throughput)/ sec. Training model separates copies of model on each GPU; total system throughput obtained by calculating the sum of the throughput obtained on each GPU. Vs. EPYC/Nvidia system: NVIDIA DGXA100, Dual AMD EPYC 7002 Series CPUs, CUDA 11.6 and Driver Version 510.47.03, Commit(Container): superbench/superbench:v0.4.0-cuda11.1.1 from here Server manufacturers may vary configurations, yielding different results. Performance may vary based on use of latest drivers and optimizations MI200-61#119Endnotes (Cont.) MI200-63- Testing Conducted by AMD performance lab as of 25/5/2022 using SuperBench v 0.4.0 benchmark, Resnet 50/101/152 datasets. All configurations used Framework: PyTorch 1.9 EPYC/Instinct system: Dual socket, 64 core, 2nd Gen AMD EPYC™ 7002 Series CPU powered server with 8x AMD Instinct™ MI250X OAM (128 GB HBM2e) 500W GPUs with AMD Infinity Fabric™ technology, ROCm™ 5.1.0 Benchmark: Resnet model (Median scores of Resnet50, Resnet 101, Resnet152 datasets) with AMD | Microsoft optimized batch sizes tuned for Resnet results for system configurations that are not yet available upstream. Commit(Container): computecqe/superbench:rocm5.1.3_superbench04 from here Benchmark Results: Resnet testing resulted in median throughput scores of 8x MI250X: resnet50 = 9708.873, resnet101 = 6705.041, resnet152 = 5250.635 Samples (Throughput)/ sec. Training model separates copies of model on each GPU; total system throughput obtained by calculating the sum of the throughput obtained on each GPU. Vs. EPYC/Nvidia system: NVIDIA DGXA100, Dual socket 64-core AMD EPYC 7742 Series CPUs, with 8x NVIDIA A100 SXM 80GB (400W), CUDA 11.6 and Driver Version 510.47.03, OS: Benchmark: Resnet model (Median scores of Resnet50, Resnet 101, Resnet152 datasets) Commit(Container): superbench/superbench:v0.4.0-cuda11.1.1 from here Benchmark Results: Resnet testing resulted in median throughput of 8x A100: resnet50 = 7057.052, resnet101 = 4968.324, resnet152 = 3806.318 Samples (Throughput)/ sec. Details on SuperBench found here Server manufacturers may vary configurations, yielding different results. Performance may vary based on use of latest drivers and optimizations MI200-63 MI300-03 - Measurements by AMD Performance Labs June 4, 2022. MI250X FP16 (306.4 estimated delivered TFLOPS based on 80% of peak theoretical floating-point performance). M1300 FP8 performance based on preliminary estimates and expectations. Actual results based on production silicon may vary. M1300-03#120FINANCIAL ANALYST DAY 2022 together we advance_ AMD Adaptive and Embedded Leadership Victor Peng President, Adaptive and Embedded Computing Group.#121Cautionary Statement This presentation contains forward-looking statements concerning Advanced Micro Devices, Inc. (AMD) including, but not limited to, the timing, availability, features, functionality and expected benefits of AMD's adaptive computing products; AMD's momentum and TAM; AMD's silicon roadmap; and AMD's new revenue opportunities and path forward, 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.#122XILINX MOMENTUM SIG Data Center 5G Telecommunications Automotive Platform Transformation $3.2B FY21* $3.9B *Xilinx fiscal year ended April 3, 2021 **Xilinx trailing 4 quarters ended March 26, 2022 FY22*** Building on Momentum as Part of AMD $33B CY25 TAM#123AECG MISSION Exceeding customer expectations with high performance, adaptive, and intelligent solutions for the data center, edge, and endpoints#124AECG STRATEGY I Delivering leadership adaptive compute products and technology Increase data center growth in networking, compute, and Al acceleration with adaptive and customizable solutions Drive growth in embedded markets by enabling customer differentiation with a broad product portfolio and Al technology#125SIGNIFICANT TAM EXPANSION Data Center* $13B Communications $32B $1058 Long-Term TAM Over 3X Xilinx CY25 TAM *Included in Data Center TAM Based on AMD Internal Data Embedded $33B Automotive $27B#126a hata VERSAL HELILETT 111111 1 A CAN ALVEO. ADAPTIVE COMPUTING DATA CENTER LEADERSHIP Compute Acceleration - Alibaba Cloud SN1000 ALVEO Adaptive SmartNIC aws SAMSUNG IBM Stace Drive Deployed Across 10 of the Largest Hyperscalers XILINX. KINTEX. X Computational Storage Microsoft Azure **** ARR.. XILINX. VERSAL XCVC1900 KORZENNYIREM ****** *AR BAR ***** 1 XILINX. VIRTEX UltraScale+ General Purpose Tencent Cloud#127XILINX. VIRTEX UltraScale+ **** ********* **** XILINX VERSAL ADVC1900 ************ MALENNETRADA ADAPTIVE COMPUTING COMMUNICATIONS LEADERSHIP Metro/Transport and Converged Access ......... CISCO ALVEO XNX VIRTEX 5G Core CHEK FUJITSU ALVEO KLINK VIRTEX Vodder Central Unit EXILINX \Orchestrating a brighter world NEC T1 Telco Accelerator Distributed Unit Deployed at 6 of the Top 7 5G Wireless Equipment Manufacturers XILINX ZYNQ. NOKIA Tala ORA Radio Unit SAMSUNG#128XILINX ZYNQ UltraScale M BYD XAZU3 -- ADAPTIVE COMPUTING AUTOMOTIVE LEADERSHIP VERSAL HY טיין אחוור Autonomous Control ADAS/AD Domain Controller Surround View Automated Parking ECU Continental XILINX ZYNQ UltraScale XAZU5 External Sensors DAIMLER Lidar Radar Forward-Looking Camera Designed-in at the Top 10 Manufacturers MAGNA AMDA SUBARU RYZEN EMBEDDED Internal Safety and Comfort Infotainment XILINX® ZYNQ® Occupant Monitoring TESLA 7F#129RYZEN me XILINX ZYNQ UltraSCALE Ma E 320 W C4% Healthcare and Sciences 5M Units Deployed in Medical Equipment in Last 12 Months* ADVANTECH ADAPTIVE COMPUTING EMBEDDED LEADERSHIP VERSAL (GE RYZEN EMBEDDED Industrial PUSS #1 Logic IC Vendor in the Industrial Market** XOLOG ZYNO. INTUITIVE SURGICAL HANARAMARN Aerospace and Defense XILINX VIRTEX 30+ Years of Innovation, Longevity of Supply and Reliability 6000+ Unique Customers Across Diverse Embedded Markets nu OLYMPUS *Based on AMD Internal Data **According to Omdia Industrial Semiconductor Market Tracker Database 4021, Xilinx holds the #1 spot for Logic IC vendor on the top ten semiconductor suppliers list for the industrial market, which captures 2019-2020 revenue and market share. Test, Measurement and Emulation #1 Logic IC Vendor for Test and Measurement* ** SIEMENS Gamesa ENERGY AMDD RENEWABLE RYZEN EMBEDDED HIFPULCO Edge Server Performance Leadership and Advanced Reliability Tektronix#130SILICON ROADMAP 2021 7nm HPC Versal™ HBM Adaptive SoCs 7nm Signal Processing Versal™ AIRF Adaptive SoCs 7nm Edge Al Computing Versal™ Al Edge Adaptive SoCs Adaptive Silicon Data Center and Edge 3nm Next-Gen Adaptive SoC 2025 2021 All roadmaps are subject to change. 7nm EPYC™ Embedded CPUs 7nm Ryzen™ Embedded CPUs 5nm Next-Gen EPYC™ Embedded CPUs 6nm Next-Gen Ryzen™ Embedded CPUs Embedded x86 Processors Networking, Storage & Automotive 2024#131Automotive Communications Aerospace and Defense Industrial and Vision Data Center Healthcare and Sciences AECG GROWTH ACCELERATION Leadership adaptive computing platforms ▪ Sustained double digit growth Diverse high-margin markets#132AMD INSTINCT ERAL AMD + XILINX OPPORTUNITIES EXILINX. ALVEO. Al Inference and Training AMD EPYC SN1000 ALVEO TAMPS OXUN Data Center and Communications VERSAL. RYZEN EMBEDDBO XILINX ZYNQ UltraScale Automotive mill WWW VERSAL AMDA EPYC RYZEN EMBEDDED Embedded Leverage Expanded Customer Base, Leadership Products, and Al Platforms#133******* PERVASIVE AI AAAAA HARRY ū#134Cloud Data Center EPYC™ CPUs, AMD Instinct™ GPUs I ■ Al Training Al Inference Commercial & Enterprise EPYC™, Ryzen™ PRO CPUs ■ ■ ■ Server PC Workstation PERVASIVE AI Digital Home Ryzen™, Radeon™ CPUs PCs and Consoles Metaverse#135Cloud Data Center EPYC™ CPUs, AMD Instinct™ GPUs, Alveo™ Accelerators, Kintex® FPGAs Al Training Al Inference Security Automatic Speech Recognition Commercial & Enterprise EPYC™, Ryzen™ PRO CPUs ■ ■ Server PC Workstation Edge Healthcare & Life Sciences Zyng Adaptive SoCs ▪ Imaging and Diagnosis Surgical Robotics PERVASIVE AI Transportation Zyng Adaptive SoCs ADAS Autonomous Vehicles Smart Retail Zynq Adaptive SoCs Cashier-less payment Customer Reidentification Inventory Management Digital Home Ryzen™, Radeon™ GPUs, Zyng Adaptive SoCs PCs and Consoles Metaverse Smart Home Smart City Zyng Adaptive SoCs ■ . . Security Endpoints Traffic Optimization Smart Grid Intelligent Factory Zyng Adaptive SoCs ■ ■ . Machine Vision Predictive Maintenance Al Robotics Communications Versal Adaptive SoCs I 5G Wireless Networking Security#136Cloud Data Center EPYC™ CPUs, AMD Instinct™ GPUs, Alveo™ Accelerators, Kintex® FPGAs Al Training Al Inference Security Automatic Speech Recognition Commercial & Enterprise EPYC™, Ryzen™ PRO CPUs ■ I Server PC Workstation Edge Healthcare & Life Sciences Zyng Adaptive SoCs ▪ Imaging and Diagnosis Surgical Robotics PERVASIVE AI Transportation Zynq Adaptive SoCs ADAS Autonomous Vehicles Smart Retail Zynq Adaptive SoCs Cashier-less payment Customer Reidentification Inventory Management Digital Home Ryzen™, Radeon™ GPUs, Zyng Adaptive SoCs PCs and Consoles Metaverse Smart Home Smart City Zyng Adaptive SoCs ■ . ■ Security Endpoints Traffic Optimization Smart Grid Intelligent Factory Zyng Adaptive SoCs ■ ■ . Machine Vision Predictive Maintenance Al Robotics Communications Versal Adaptive SoCs I 5G Wireless Networking Security#137PERVASIVE AI STRATEGY Leadership Al IP with AMD CDNA™ and scalable AIE architecture Broaden AMD Al product portfolio across cloud, edge, and endpoint applications Unified Al stack to empower developers across the AMD portfolio#138AMD XDNA: ADAPTIVE ARCHITECTURE IP ▪ Dataflow architecture optimal for Al and signal processing applications Highly-scalable array of engines with local memory and data movers Leverages deep expertise of compiling algorithms to FPGAs and adaptive SoCs Adaptive Interconnect AI Engine AI Engine Al Engine (AIE) AI Engine Local Mem. Local Mem. Local Mem. AI Engine AI Engine Local Mem. Local Mem. Local AI Engine Mem. Source: The McClean Report, May 2022 High-Performance and Energy Efficiency for Al and Signal Processing FPGA LOGIC Adaptive Interconnect FPGA LOGIC FPGA Fabric FPGA LOGIC (FPGA) Local Mem. Local Mem. Local Mem. FPGA LOGIC FPGA LOGIC FPGA LOGIC Local Mem. Local Mem. Local Mem. Leading FPGA for broad set of applications and Al#139AMD AI ENGINE: ADAPTIVE DATAFLOW PROCESSOR L1 Deep Neural Network (DNN) L2 L3 L4 L5 L6 DNN Runs Optimally on AIE L2 L1 Adaptive Interconnect L3 ↓ L4 L6 L5 Data "flows" from layer to layer, connections between layers are often “sparse" Dataflow architecture, sparsity, efficient datatypes deliver high performance and low power High-Performance, Energy Efficient, and Customizable for Al Workloads#140Ryzen™ CPUS Inference AI APPLICATION COVERAGE Training Small-medium models (EPYC™ CPUs only) Mid-size models, mid-range performance and memory BW, high memory capacity (e.g., face recognition, recommendation systems) EPYC™ CPUS Increasing Processing Requirements Radeon™ GPUS Versal® Al Adaptive SoCs Medium-large models Larger models, high performance and more memory BW (e.g., video analytics) AMD Instinct™ GPUS Large-very large models Very large models, efficient data types, and high memory BW (e.g., NLP, large recommendation systems)#141Ryzen™ CPUS Inference AI APPLICATION COVERAGE Training Small-medium models (EPYC™ CPUs only) Mid-size models, mid-range performance and memory BW, high memory capacity (e.g., face recognition, recommendation systems) EPYC™ CPUS Ryzen™ CPUS AMD XDNA AIE Increasing Processing Requirements Radeon™ GPUS Versal® Al Adaptive SoCs EPYC™ CPUS AMD XDNA AIE Medium-large models Larger models, high performance and more memory BW (e.g., video analytics) Large-very large models AMD Instinct™ GPUS Adaptive SoCs Very large models, efficient data types, and high memory BW (e.g., NLP, large recommendation systems)#142CPU Stack WinML** ONNX Runtime Optimized Inference Models Pytorch Windows Runtime AMD AI SOFTWARE TODAY ML Graph Compiler ZenDNN, AOCL Optimized Library ZEN Studio (AOCC) Ryzen™ CPUs *Ryzen uses Radeon ML and WinML **WinML for Ryzen Only TensorFlow Linux Runtime EPYC™ CPUs ROCM™ Platform TM Optimized Inference Models ONNX Runtime Pytorch AMD Instinct™ GPUS MIGraphX: Al Development Tools MIGraphX: ML Graph Compiler MIOpen, ROCBLAS ROCM™ HIP Compiler and tools TensorFlow Runtime Radeon™ GPUS Al Development Tools: Quantizer, Sparsity/Pruning Tool Al Deployment Tools: Inference Server Vitis™ Al Platform Optimized Inference Models ONNX Runtime Pytorch Vitis Al Development & Deployment Tools Vitis Al ML Graph Compiler Vitis ML Libraries Vitis™ SW Platform (AIE Compiler, HLS and tools) CNN Overlay TensorFlow VersalⓇ Adaptive SoCs Runtime Transformer Overlay ZynqⓇ Adaptive SoCs#143WinML* Windows Runtime ML Graph Compiler ZenDNN, AOCL Optimized Library ZEN Studio (AOCC) Ryzen™ CPUs *WinML for Ryzen Only AMD UNIFIED AI STACK 1.0 ONNX Runtime Linux Runtime Unified Inference Frontend (UIF) for Al Developers EPYC™ CPUs Optimized Inference Models Pytorch Unified Al Development & Deployment Tools (Quantization, Pruning, Inference Server) MIGraphX: ML Graph Compiler MIOpen, ROCBLAS ROCM™ HIP Compiler and tools Runtime AMD Instinct™ GPUs Radeon™ GPUS All roadmaps are subject to change. TensorFlow Ryzen™ CPUs AMD XDNA AIE Vitis Al ML Graph Compiler Vitis ML Libraries Vitis™ SW Platform (AIE Compiler, HLS and tools) CNN Overlay Runtime Transformer Overlay VersalⓇ Adaptive SoCs ZynqⓇ Adaptive SoCs#144WinML* Windows Runtime Ryzen™ CPUs *WinML for Ryzen Only ZEN Studio (AOCC) + Vitis AIE tools Ryzen™ CPUs AMD XDNA AIE AMD UNIFIED AI STACK 2.0 Seamless Workload Partitioning with UIF, Graph Compiler and Library APIs ONNX Runtime Linux Runtime EPYC™ CPUs Pytorch Unified Al Development & Deployment Tools (Quantization, Pruning, Inference Server) Unified ML Graph Compiler Common Library APIs for inference operators ROCm™ HIP Compiler and tools Optimized Inference Models EPYC™ CPUs AMD XDNA AIE TM AMD Instinct™ GPUS Runtime Radeon™ GPUS All roadmaps are subject to change. TensorFlow Vitis™ SW Platform (AIE Compiler, HLS and tools) CNN Overlay VersalⓇ Adaptive SoCs Runtime Transformer Overlay ZynqⓇ Adaptive SoCs#145AMD INSTINCT & RAL y AMD + XILINX OPPORTUNITIES EXILINX. ALVEO. Al Inference and Training AMD EPYC SN1000 ALVEO #EXUN Data Center and Communications >$10B wwwWWWWWW VERSAL Based on AMD Internal Data A M AMG-2 RYZEN EMBEDORO XILINX ZYNQ LitraScalo Automotive wwwwww VERSAL ****** LE AND AMDA EPYC RYZEN IDENTIFIED REVENUE OPPORTUNITIES Embedded#146ACCELERATING AMD GROWTH TAM Expansion Large and diversified markets KON XLNX ZYNQ EAN ALVEO EXILINX 71 Telco Accelerato Leadership Products Broad portfolio of platforms Pervasive AI Drive cloud, edge and endpoints to scale#147FINANCIAL ANALYST DAY 2022 together we advance_ Performance and Growth Saeid Moshkelani Senior Vice President and General Manager, Client AMD#148Cautionary Statement This presentation contains forward-looking statements concerning Advanced Micro Devices, Inc. (AMD) including, but not limited to, the total addressable PC market and AMD's opportunity; the timing, availability, features, functionality and expected benefits of AMD's products; and AMD's notebook and desktop roadmaps, which are made pursuant to the Safe Harbor provisions of the Private Securities Litigation Reform Act of 1995. Forward-looking statements are commonly identified by words such as "would," "may," "expects," "believes," "plans," "intends," "projects" and other terms with similar meaning. Investors are cautioned that the forward-looking statements in this presentation are based on current beliefs, assumptions and expectations, speak only as of the date of this presentation and involve risks and uncertainties that could cause actual results to differ materially from current expectations. Such statements are subject to certain known and unknown risks and uncertainties, many of which are difficult to predict and generally beyond AMD's control, that could cause actual results and other future events to differ materially from those expressed in, or implied or projected by, the forward-looking information and statements. Investors are urged to review in detail the risks and uncertainties in AMD's Securities and Exchange Commission filings, including but not limited to AMD's most recent reports on Forms 10-K and 10-Q. AMD does not assume, and hereby disclaims, any obligation to update forward-looking statements made in this presentation, except as may be required by law.#149GLOBAL PC MARKET ENORMOUS OPPORTUNITY Commercial $22B Gaming $9B $50B LONG-TERM TAM Consumer $19B Based on AMD internal data. PCs are Essential & Mission Critical +$7B TAM growth over the past two years Strong Market Growth Drivers Commercial with hybrid workforce Gaming and premium consumer Huge Opportunity for AMD Sustained revenue and share growth#150. МИНИИН ки мини ини **********E AMD RYZEN 6000 SERIES IIIIIIIIMI I IIMIMIIIIIIII ** AF MAINTAINING LEADERSHIP ACROSS ALL PC MARKETS AMD 10 ⠀⠀⠀⠀ Is Build the Best ***** RYZEN AMD RADEON AMG ADVANTAGE Deliver Leading Experiences Driving Sustained Growth acer hp ASUS Lenovo Deepen OEM Relationships DELL Microsoft Surface#151OUR JOURNEY INCREDIBLE NOTEBOOK MOMENTUM Premium Consumer Growth % Share of >$800 Notebook Market 6% O 2019 o 23% 2021 Gaming Growth % Share of Notebook Market 15% O 2019 o 32% 2021 Source: IDC; Exit shares, x86 processors only Commercial Growth % Share of >$500 Notebook Market 9% O 2019 o 16% 2021#152AMD NOTEBOOK FOCUS MOBILE USES FOR MODERN LIFESTYLES Hybrid Work Environment AE Anywhere Gaming Continuous Streaming Advanced Content Creation Users First, Experience Focused Unleashed Computing#153AMD RYZEN™ 6000 SERIES CONTINUED NOTEBOOK LEADERSHIP Leading All-Around Compute Performance 1.3X Faster CPU Performance Best Ultrathin Notebook Gaming Ever 2X More Graphics Performance Incredible Power Efficiency Up to 29 Hours of Battery Life Cutting-Edge Connectivity Wi-Fi 6E, 5G, Bluetooth 5.2 and USB4™ See endnote RMP-39, RMB-13 TM HIIIIIII "Zen 3+" CPU Core нннннн ши HHHH HHH II II III AMD RYZEN 6000 SERIES IIMIMIIIIIIII RDNA™ 2 With Raytracing Support 6nm Technology#154AMD RYZEN™ PRO 6000 SERIES NOTEBOOK ADVANCED COMMERCIAL PROCESSORS Ultimate Performance +17% MS Office Performance while Using Teams Conferencing Intel Core i7-1260P AMD Ryzen™ 7 PRO 6000 U-series Extreme Portability +45% Battery Life Performance for Teams Conferencing See endnote RMP-31, RMP-32 Built for Professionals Leading-edge security features for a hybrid workforce Wireless remote, enterprise manageability#155AMD PLATFORM EXPANSION PARTNERING TO DELIVER THE BEST AMD Platform Momentum 185 117 Premium Designs 68 2020 261 204 Premium Designs 57 2021 294 249 Premium Designs 45 2022 Deepening OEM Partnerships Lenovo Yoga 7 HP EliteBook 845/865 G9 Driving Co-Innovation Premium Platforms ROG Zephyrus G14 Lenovo Thinkpad Z Alienware Aurora R14 Razer Blade 14 Multi-Year Roadmap Planning#156AT THE FOREFRONT OF PC INNOVATION AI Acceleration Adaptive computing Image Signal Processing Incredible image quality Heterogeneous Architecture Advanced scalability FRONTER AMD ** Advanced Packaging Breakthrough performance VER Extreme Power Management Even longer battery life#157AMD NOTEBOOK ROADMAP SUSTAINED PERFORMANCE LEADERSHIP 2021 7nm AMD RYZEN 5000 SERIES "Zen 3" "Vega" 6nm AMD RYZEN 6000 SERIES "Zen 3+" RDNA™ 2 4nm "Phoenix Point" Roadmaps Subject to Change "Zen 4" RDNA 3 AIE Advanced Node "Strix Point" "Zen 5" RDNA 3+ AIE 2024#158AMD DESKTOP FOCUS DELIVERING ULTIMATE USER EXPERIENCES Pushing the envelope for gamers Unleashed performance for enthusiasts Extreme workstation workloads#159AMD RYZEN™M 7000 SERIES THE NEXT FRONTIER FOR DESKTOPS Fastest Desktop Processor >5.5 GHz Clock Speed ~8% IPC Uplift in Desktop Workloads >15% Single-Thread Uplift The Most Efficient Desktop Processor >25% More Performance-Per-Watt All New AM5 Socket Platform More Performance and Faster Data Access See endnote RPL-001, Z4-001, Z4-003 [[[[ TAIT DAN NO HA HAF HAF "Zen 4" CPU Core AMD RYZEN 11111 IIII AAA HIBE H MIRIA IA A ID RDNA™ 2 Architecture D IIII [[[[ TATAL 5nm Technology#160AMD THREADRIPPER™ PRO EXPANSION WORKSTATION 2X Intel® Xeon® 8280 Intel® Xeon® 2X Platinum 8280 +43% Single Thread Performance Leadership +95% LEADERSHIP Multi Thread +51% Graphics AMD Threadripper™ PRO CPU 5000 +138% Storage +100% Performance to Power Ratio See endnote CGP-40, CGP-30, CGP-29, CGP-24 Portfolio Growth x4 DELL DELL Lenovo#161AMD DESKTOP ROADMAP HIGH PERFORMANCE MOMENTUM 2021 AMD RYZEN "Zen 3" 5000 Series 7nm AMD RYZEN "Zen 3" 3D V-Cache TM AMD RYZEN THREADRIPPER "Zen 3" AMD RYZEN "Zen 4" 7000 Series 5nm AMD RYZEN "Zen 4" 3D V-Cache All roadmaps are subject to change. AMD RYZEN THREADRIPPER "Zen 4" Advanced Node "Granite Ridge" "Zen 5" 2024#162OUR PATH FORWARD SUSTAINED GROWTH ACROSS PC MARKETS ▪ Execute strong roadmap ▪ Deliver the best user experiences Deepen OEM relationships#163ENDNOTES ■ ■ GD-150: Max boost for AMD Ryzen processors is the maximum frequency achievable by a single core on the processor running a bursty single-threaded workload. Max boost will vary based on several factors, including, but not limited to: thermal paste; system cooling; motherboard design and BIOS; the latest AMD chipset driver; and the latest OS updates. GD-168-A: Based on testing by AMD Labs as of 4/11/22. Battery life evaluated in hours of continuous 1080p local video playback with a premium laptop configured with an AMD Ryzen 7 PRO 6850U processor, 76 WHr battery, 150 nit screen ghtness. Actual battery life will vary based on several factors, including, but not limited to: product configuration and usage, software, operating conditions, wireless functionality, power management settings, screen brightness and other factors. The maximum capacity of the battery will naturally decrease with time and use. RMB-13: Based on testing by AMD as of 12/14/2021. CPU performance evaluated with a geomean of 9 multi-threaded content creation and CPU tests. GPU performance evaluated with 3DMark® Time Spy. System configuration for Ryzen™ 7 5800U CPU/GPU performance: HP ProBook 635 Aero G8 configured with 2x8GB DDR4-3200 (22-22-22), Windows® 11 22000.282, Samsung 980 Pro 1TB SSD, 15W nominal processor TDP, GPU driver 27.20.21026, BIOS T83. System configuration for Ryzen™ 7 6800U CPU/GPU performance: AMD reference motherboard configured with 4x4GB LPDDR5-6400 (40-39-45-90), Windows® 11 22000.282, Samsung 980 Pro 1TB SSD, 28W nominal processor TDP, GPU driver 30.0, BIOS TRM0081D. RMP-31: Based on testing by AMD as of 4/1/22. Productivity performance evaluated with simultaneous operation of nine-participant Microsoft Teams video conferences using the UL Procyon Office Productivity benchmark. System configuration for Intel® Core™ i7-1260P CPU/GPU performance: Lenovo ThinkPad X1 Carbon, Intel Iris Xe Graphics, 2X8 GBytes RAM (LPDDR5-5500), 1TB SSD, BIOS version N3AET45W (1.10), Windows 11 Pro. System configuration for Ryzen™ 7 PRO 6860Z: Lenovo ThinkPad Z13, 2x16GB LPDDR5-6400, Windows 11 Pro, 1TB SSD, AMD Radeon 680M graphics, GPU driver 30.0, BIOS N3GET12WE (0.12 ). Performance may vary. RMP-32: Based on testing by AMD as of 4/1/22. Battery life evaluated in hours using a nine-participant Microsoft Teams video conference with camera on, 200 nit brightness, slider position AC#2 (Balanced), with 95% utilization. Battery life results normalized for battery capacity differences. System configuration for Intel® Core™ i7-1260P CPU/GPU performance: Lenovo ThinkPad X1 Carbon, 57-watt hour battery, Intel Iris Xe Graphics, 2X8 GB RAM (LPDDR5-5500), 1TB SSD, BIOS version N3AET45W (1.10), Windows 11 Pro. System configuration for Ryzen™ 7 PRO 6860Z: Lenovo ThinkPad Z13, 50-watt hour battery, 2x16GB LPDDR5-6400, Windows 11 Pro, 1TB SSD, AMD Radeon 680M graphics, GPU driver 30.0, BIOS N3GET12WE (0.12 ). Actual battery life will vary based on several factors, including, but not limited to: product configuration and usage, software, operating conditions, wireless functionality, power management settings, screen brightness, and other factors. The maximum capacity of the battery will naturally decrease with time and use. RMP-39: Based on testing by AMD Labs as of 4/11/22. Battery life evaluated in hours of continuous 1080p local video playback with a HP Elitebook 865 G9 configured with an AMD Ryzen 7 PRO 6850U processor with Radeon 680M graphics, 76 WHr battery, 150 nit screen brightness, 256GB HDD, 8GB memory, Win 10 Pro, video resolution of 1920 x 1200 x 60 Hz and the power slider set to "better battery." Actual battery life will vary based on several factors, including, but not limited to: product configuration and usage, software, operating conditions, wireless functionality, power management settings, screen brightness and other factors. The maximum capacity of the battery will naturally decrease with time and use. RPL-001: Testing as of May 5, 2022, by AMD Performance Labs. Single-thread performance evaluated with Cinebench R23 1T. AMD Ryzen 9 5950X System: ASUS ROG Crosshair VIII Hero X570, 2x8 DDR4-3600C16. AMD Ryzen 7000 Series: AMD Reference X670 Motherboard, 16-core pre-production processor sample, 2x16GB DDR5-6000CL30. All systems configured with Radeon™ RX 6950XT GPU (driver: 22.10 Prime), Windows 11 Build 22000.593, Samsung 980 Pro 1TB SSD, Asetek 280MM liquid cooler. Results may vary.#164ENDNOTES ■ CPG-04: Based on AMD performance lab testing as of January 31, 2022, using Chaos V-Ray, PugetBench for Adobe After Effects, Puget Bench for Davinci Resolve, SPECapc® for Maya 2017 CPU Composite metric, SPECapc® for Maya 2017 Graphics Interactive Composite, Cinebench 1T, Cinebench NT, Chromium Compilation, Unreal Engine Compilation, Cadalyst AutoCAD 3D Graphics, Cadalyst AutoCAD CPU, Ansys CFX Pump, Ansys CFX LeMans Car and Keysot benchmarks to compare the performance of an AMD Ryzen Threadripper 5995WX reference system configured with 8x32GB DDR4, NVIDIA Quadro RTX A5000, 1TB SSD, Win 11 vs. a similarly configured BOXX APEXX4 workstation with TWO Intel® Xeon®Platinum 8280 processors. Results may vary CGP-24: Based on internal AMD analysis of benchmarks as of January 31, 2022, evaluating the V-Ray rendering performance and TDP of an AMD Ryzen Threadripper Pro 5995WX reference system (280W) configured with 8x32GB DDR4, NVIDIA Quadro RTX A5000, 1TB SSD, Win 11 vs. a similarly configured BOXX APEXX4 workstation with TWO Intel® Xeon® W-8280 server processors (410W). Results may vary. CGP-24 CGP-29: Based on AMD Labs testing as of January 31, 2022 using the Crystal Disk Mark 8 benchmark, to compare the PCIe® storage performance of a Ryzen Threadripper PRO 5995WX reference system configured with 8x32GB DDR4, NVIDIA Quadro RTX A5000, 1TB SSD, Win 11 vs. a similarly configured BOXX APEXX4 workstation with 2X Intel® Xeon® W-8280 processor. CGP-30: Based on AMD performance lab testing on January 31, 2022, using the SPECapc® for Maya 2017 Graphics Interactive Composite metric to compare graphics performance of (5) AMD Ryzen™ Threadripper™ PRO 5000WX-Series reference systems configured with 8x32GB DDR4, NVIDIA Quadro RTX A5000, 1TB SSD, Win 11 vs. (5) similarly configured BOXX APEXX4 workstations with Intel® Xeon® W-3300 series processors. Results may vary. Z4-001: IPC uplift based on the average of estimated/published 2017 SPECint® and 2017 SPECfp scores and internal estimates/testing on Cinebench R23 1T and Geekbench 51T.for "Zen4" and "Zen 3" processors Z4-003: Testing as of May 31, 2022, by AMD Performance Labs. Power measured at CPU socket only (Watts), CPU performance ("points") measured with Cinebench R23 nT. AMD Ryzen 9 5950X System: AMD Reference X570 Motherboard, 2x8 DDR4-3200. AMD Ryzen 7000 Series: AMD Reference X670 Motherboard, Ryzen 7000 Series 16-core pre-production processor sample, 2x16GB DDR5-5200. All systems configured with Radeon™ RX 6950XT GPU (driver: 22.10 Prime), Windows 11 Build 22000.593, Samsung 980 Pro 1TB SSD, Asetek 280MM liquid cooler. Results may vary when final products are released in market.#165FINANCIAL ANALYST DAY 2022 together we advance_ AMD Building on Gaming Leadership Rick Bergman Executive Vice President, Computing and Graphics Business Group#166Cautionary Statement This presentation contains forward-looking statements concerning Advanced Micro Devices, Inc. (AMD) including, but not limited to, the timing, availability, features, functionality and expected benefits of AMD's graphics products; gaming TAM and market; AMD's gaming GPU roadmap; expected Ryzen and Radeon systems in 2022; projected annual console sales and CAGR; new growth opportunities; silicon TAM; and AMD's growth opportunities and path forward, 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.#167INSATIABLE DEMAND FOR GRAPHICS COMPUTE HEE Source: AMD internal research Gaming Entertainment Creators Machine Learning Photorealistic Gaming Metaverse GFLOPS 500,000 100,000 2016 2021 >3X 2024 Potential Metaverse Trajectory Graphics Compute 2030#168n Gaming Desktops Game Consoles Gaming Laptops Automotive AMD RADEON PRO Workstations Smartphones TETRI AMD GRAPHICS EVERYWHERE Handheld Gaming Tencent Cloud Cloud Gaming#169AMD GAMING LEADERSHIP MILLINE CTRL Mi#170GAMING IS THE FASTEST GROWING FORM OF ENTERTAINMENT Source: Newzoo, Statista, AMD Research GAMERS >3B WORLDWIDE ~70% cross-platform players GAME ON PC 57% AND CONSOLE ~$176B spent on games in 2021 Number of Gamers 2B 2015 2020 4B 2030#171$37B SILICON TAM Growing Market DELIVERING INCREDIBLE GAMING EXPERIENCES AMD RDNA Established RDNA" Architecture TM PP SAG Complete Radeon™ Product Line-Up Source: AMD Estimates Console Leadership#172AMD RADEON™ GRAPHICS DELIVERING PRODUCT LEADERSHIP ▪ Top-to-bottom product stack with focus on premium segments Strong partnership with ISVs to deliver world-class software solutions ■ Projected performance-per-watt leadership ▪ AMD Advantage™ solutions deliver unique, state-of-the-art, market-leading solutions See endnote RM-041 RADEON RADEON RADEON#173AMD RADEON™ RX 6000 SERIES Growing family of powerful and efficient gaming GPUs 22+ Models Offered 300+ SI, AIB and OEM Solutions 3X Notebook Designs AMOLT RADSON ANDO RADDON X6800 X AHDD RADEON AY 6800 RADEON#174AMD FIDELITYFX™ SUPER RESOLUTION TECHNOLOGY 4K FSR OFF 73 FPS VON Arant "DEATHLOOP" Varen Riddles BR AMD CHAOS CHUO BICH FSR 2.0 ON "Performance" Mode 121 FPS E TREASURE OF THE ICE ALL YOU CAN EAT AMD FidelityFX Super Resolution 2.0 Only cross-platform, open-source upscaling solution FSR 1.0 delivers up to 2.5X performance over native 4K resolution across select titles FSR 2.0 launched on May 12 Near native quality with higher FPS 100+ game titles Endnotes: RS-430#175AMD “NAVI 3" GRAPHICS AT THE FOREFRONT OF GRAPHICS INNOVATION Industry-leading performance-per-watt ▪ Advanced packaging technologies System level efficiency ▪ Advanced multimedia capabilities See endnote RX-785 RDNAT™ 3 GPU Core TM Chiplet Architecture 5nm Technology Coming later this year#176AMD GAMING GPU ROADMAP SUSTAINED PERFORMANCE LEADERSHIP 2019 7nm AMD RDNA AMD RADEON RX 5000 SERIES 7nm AMD RDNA 2 AMD RADEON RX 6000 SERIES 5nm AMD RDNA 3 "Navi 3x" All roadmaps are subject to change. Advanced Node AMD RDNA 4 "Navi 4x" 2024#177THE POWER OF RYZEN™ CPU + RADEON™ GPU TOGETHER ALIENWARE R THOSE WHO DARE AMD 12 SYSTEMS in 2020 RYZEN RADEON FOR THOSE WHO DARE FOR THOSE WHO LEGION LEGION 33 SYSTEMS in 2021 COMEN 50+ SYSTEMS Expected in 2022#178AMD RADEON AMD ADVANTAGE WIT GAMING FG PASION AMD SmartShift Max 2X More Boost Up to 21121 ANDORON AMD Eco AMD RYZEN ‒‒‒‒‒‒‒‒‒‒ AMD SmartShift Eco Enabled ‒‒‒‒ SmartShift 2X Up to Projected Gaming Time on Battery 1 ++ Culls & ::: A ANDON AMD Graphics SmartAccess AMD RADEON Up to AMDO RYZEN 29% More Game Performance In Doom Eternal See endnotes RM-040, RM-041, RM-045, RM-062 + AMD AMD Software Adrenalin Edition SmartAccess Memory 0 Up to 30% More Game Performance In Assassin's Creed Valhalla#179GAME CONSOLES ARE GROWING & EVOLVING 50MU Historical 2013 Source: AMD Internal Analysis Annual Console Sales +Handheld + New Consoles and User Growth 2021 Number of U.S. gamers, total time spent gaming and total spend are increasing Expanding demand driven by significant growth in female and 45+ age gamers ▪ Gaming communities are fast growing, driven by multi-player and online play ▪ Strong handheld gaming growth driven by on-the-go demand#180A BUILDING THE BEST CUSTOM SILICON #1 Market Share Microsoft Xbox Series X | S SONY PlayStation 5 VALVE Steam Deck Increasing Silicon Content 4K Immersive Experiences Fast Load Times Backward Compatibility Strong Partnerships Deep hardware and software co-engineering Innovative end user experiences Broadened gaming ecosystem#181NEW MARKET GROWTH#182NEW GROWTH OPPORTUNITIES Metaverse Growth Opportunity Custom Silicon Leadership#183METAVERSE OPPORTUNITIES TOOLL OUTI **** Datacenter EPYC™ Processors AMD Instinct™ Accelerators Alveo™ Accelerators User Data ||| Mobile Network Wired Broadband Edge Radeon Pro™ Graphics EPYC™ Processors Versal™ FPGAs AR/VR Video Near Compute Radeon™ PRO Graphics EPYC™ Processors Versal™ FPGAs Endpoint Radeon™ Graphics Ryzen™ Processors#184AMD ACCELERATES THE METAVERSE AMD EPYC ONEG Expanding 3D Creation Beyond Movies AMD RYZEN THREADRIPPER u UNREAL ENGINE Accelerating 3D Creation Beyond Games AMD AMD EPYC RADEON CLOUD GAMING PRO AMD Custom Silicon Powering Cloud Gaming Infrastructure Gear 2-38 N Trans Gros STATOR Magic Leap Prototype AMD Custom Silicon Interacting Within Metaverse#185UNLOCKING THE FULL POTENTIAL OF CUSTOM SILICON Automotive Data Center Cloud Gaming $30B Long-Term TAM AR/VR - Telco/Edge Based on AMD Internal Data ▪ Extend proven custom silicon, co-engineering model into new markets Leverage multi-generational chiplet experience Continue to build deep & long-term partnerships with market makers ▪ 200+ million custom solutions shipped#186AMD RDNA AMD XDNA EXPANDING OUR CUSTOM SILICON & SOLUTIONS AMD CDNA ZEN Unmatched IP Portfolio ● ●● e a arm synopsYS cadence Customer and 3rd Party IP Integration High-Performance SoC Design Custom-Ready Chiplet Platform#187OUR PATH FORWARD HUGE GROWTH OPPORTUNITY ▪ Insatiable demand for graphics ▪ Leadership products today Next generation graphics coming Expanded market opportunity in custom silicon and metaverse REE TO#188ENDNOTES GD-18 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 presentation, 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 presentation, 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 presentation. Terms and limitations applicable to the purchase or use of AMD's products are as set forth in a signed agreement between the parties or in AMD's Standard Terms and Conditions of Sale. GD-18 RM-040 Testing done by AMD performance labs December 10, 2021, comparing the performance of a system configured with a Radeon™ 6850M XT GPU and a Ryzen™ 6900H Series processor with SmartShift Max enabled vs a similarly configured system with a Ryzen™ 5900H Series processor and SmartShift enabled. Games tested: Control (High); Godfall (Ultra); Shadow of the Tomb Raider (Medium); Strange Brigade (Ultra); Tom Clancy's The Division 1 (Ultra). Performance will vary. RM-40. RM-041 Engineering projections are not a guarantee of final performance. Performance projection by AMD engineering staff based on AMD DC power guidelines on a system configured with Radeon™ RX 6850M XT mobile GPU. Projection is subject to change when final product is released in market. RM-041 RM-045 Testing done by AMD performance labs 7/20/2021 with a reference laptop design with AMD Ryzen 7 5800XH, Radeon RX 6600M with Smart Access Memory enabled, 32GB DDR4-3200MHz, Win10 Pro x64 19042.985 and AMD Driver 21.10.01.21. Testing of "Hybrid Mode" vs. "Discrete Graphics Mode" showed an average uplift of 15% more performance across the following titles: Assassin's Creed Valhalla; Battlefield 5; Borderlands 3; Call of Duty: Black Ops Cold War; Cyberpunk 2077; DOOM Eternal; Horizon Zero Dawn; Red Dead Redemption 2; Fortnite; Valorant; World of Warcraft: Shadowlands. Performance may vary. RM-045 RM-062 Testing done by AMD performance labs, February 14, 2022, on Asus ROG Strix G513 (pre-production sample) equipped with Ryzen 9 5900HX, Radeon RX 6800M. Tested at 1080p with Smart Access Memory ON with driver 22.2.2 versus SAM OFF in driver 22.2.1 in the following game and settings: Assassin's Creed Valhalla, DX12, Ultra High Settings. Performance may vary. RM-062#189ENDNOTES RS-430 AMD Radeon™ RX 6000 Series graphics cards with AMD FidelityFX Super Resolution (FSR) "Performance" mode ON offer on average 2.5x better performance at 4K across select titles vs. FSR OFF. Testing by AMD Performance Labs as of July 22, 2021, on a test system configured with an AMD Ryzen™ 9 5900X CPU, AMD Radeon™ 6900 XT/6800 XT/6700 XT/6600 XT graphics cards, 16GB DDR4-3200 RAM, and Windows 10 Pro, with Smart Access Memory enabled and using FSR "Performance" mode vs. FSR off. Actual performance will vary. Games tested: Baldur's Gate 3 ("Ultra"); Black Desert Global Lab ("Epic"); Chernobylite ("Epic"); Deathloop™ ("Ultra"); Far Cry 6 ("Ultra"); Gamedec ("Very High") Grounded ("Epic"); Necromunda: Hired Gun™ ("Epic"); Resident Evil™ Village ("Epic"); Vampire: The Masquerade - Bloodhunt ("Ultra"). Benchmark tests: Baldur's Gate 3, Vulkan, 3840 x 2160, "Ultra" settings, FSR OFF/FSR "Performance" mode. Black Desert, DX11, 3840 x 2160, "Epic" settings, FSR OFF/FSR "Performance" mode. Vampire: The Masquerade - Bloodhunt, DX12, 3840 x 2160, "Ultra" settings, FSR OFF/FSR "Performance" mode. Chernobylite, DX11, 3840 x 2160, "Epic" settings, FSR OFF/FSR "Performance" mode. Deathloop™, DX12, 3840 x 2160, "Ultra" settings, FSR OFF/FSR "Performance" mode vs. the same system with FSR off. Far Cry® 6, DX12, 3840 x 2160, "Ultra" settings, FSR OFF/FSR "Performance" mode. Gamedec, DX12, 3840 x 2160, "Very High" settings, FSR OFF/FSR "Performance" mode. Grounded, DX12, 3840 x 2160, "Epic" settings, FSR OFF/FSR "Performance" mode. Necromunda: Hired Gun™, DX12, 3840 x 2160, "Epic" settings, FSR OFF/FSR "Performance" mode. Resident Evil™ Village, DX12, 3840 x 2160, "Epic" Settings, FSR OFF/FSR "Performance" mode. RS-430. RX-785 Testing done by AMD performance labs May 31, 2022, on (11) AMD Radeon™ RX 6000 Series graphics cards, using systems configured with Ryzen™ 9 5900X and Ryzen™ 5 5600X CPUs, each with 16GB DDR4-3600MHz and AMD Smart Access Memory enabled, Win 10 Pro versus similarly configured systems with (11) Nvidia GeForce RTX 3000 Series, GeForce GTX 1650 and GTX 1050 Ti GPUs, each with ReBAR enabled. Performance tested across 20 games at 4K, 1440P and 1080P resolutions at MAX settings, and across 12 games at 1080P Medium settings for each one of the (22) AMD and NVIDIA GPUs. Performance per watt and per die size calculated using the total board power (TBP) and die sizes of the (22) individual AMD and NVIDIA GPUs over the average FPS scores. Performance may vary. RX-785. Slide 16 (Building the Best Custom silicon) #1 market share based on sales of Playstation and Xbox consoles. Source: IDC, Worldwide Home Video Game Console Market Shares, 2021: The Ninth-Generation Race Tightens, doc #US47709922, May 2022#190FINANCIAL ANALYST DAY 2022 together we advance_ Financial Momentum Devinder Kumar Executive Vice President, Chief Financial Officer and Treasurer AMD#191Cautionary Statement This presentation contains forward-looking statements concerning Advanced Micro Devices, Inc. (AMD) including, but not limited to, AMD being positioned for long-term cash generation; the expected benefits of AMD's acquisition of Xilinx, Inc., including being non-GAAP EPS and free cash flow accretive in 2022; AMD's expanding growth opportunities and TAM; AMD's long-term financial priorities and model, including potential drivers; long-term revenue growth drivers; AMD's 2022 financial outlook, including revenue growth, non-GAAP gross margin, non-GAAP operating expenses as a percentage of revenue, non-GAAP tax rates and diluted share count; Xilinx momentum and opportunities; AMD's capital allocation priorities; and accelerating financial momentum, 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 In this presentation, in addition to historical GAAP financial results, AMD has provided historical non-GAAP financial measures including non-GAAP gross margin, non-GAAP operating margin, non-GAAP earnings per share and free cash flow. AMD uses a normalized tax rate in its computation of the non-GAAP income tax provision to provide better consistency across the reporting periods. For fiscal year 2019, 2020 and 2021, AMD used a non-GAAP tax rate of 3%, 3%, and 15%, respectively, which excludes the tax impact of pre-tax non-GAAP adjustments. In addition, AMD provided combined AMD and Xilinx pro forma non-GAAP financial measures for fiscal year 2021, including rev nue and gross profit, as supplemental information. AMD is ding 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 contains forward-looking non-GAAP measures concerning AMD's 2022 financial outlook and long-term financial model such as gross margin, operating expenses, operating margin, tax measures and free cash flow margin. For 2022 financial outlook and long-term financial model, AMD uses a projected non-GAAP tax rate of 13%, which excludes the tax impact of pre-tax non-GAAP adjustments, reflecting currently available information, and a projected non- GAAP cash tax rate of approximately 10% and 20%, respectively, that includes the projected current income tax liability plus known foreign withholding tax obligations paid expressed as a percentage of non-GAAP profit before tax. 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 typically include stock-based compensation, amortization of acquired intangible assets, income tax provision, equity income in investee, and other non-recurring items such as loss on debt redemption/conversion, impairment charges and acquisition-related costs. The timing and impact of such adjustments are dependent on future events that are typically uncertain or outside of AMD's control; therefore, a reconciliation to equivalent GAAP measures is not practicable at this time. All statements made in this presentation are based on current expectations as of June 9, 2022. AMD undertakes no intent or obligation to publicly update or revise any statement in this presentation whether as a result of new information, future events or otherwise, except to the extent that disclosure may be required by law.#192T A OUR JOURNEY#193TH KEY FINANCIAL PRIORITIES 2020 Increasing Operating Margin & Profitability Strong Revenue Growth Expanding Gross Margin Significant Cash Generation#194OUR RESULTS BEST-IN-CLASS REVENUE GROWTH $6.7B 2019 56% CAGR $9.8B 2020 $16.4B 2021#195OUR JOURNEY SIGNIFICANT REVENUE GROWTH & DIVERSIFICATION ~$1B ~15% Revenue PCs & Gaming Data Center & Embedded 2019 $6.7B VS. 2021 $16.4B Quadrupled Data Center & Embedded Revenue in 2 Years ~$4B ~25% Revenue#196OUR JOURNEY OUTSTANDING FINANCIAL PERFORMANCE¹ 43% 2019 Gross Margin +560 bps 45% 2020 48% 2021 Operating Margin +1230 bps 12% 2019 17% 2020 25% 2021 (1) Non-GAAP financial measures. See appendices for GAAP to non-GAAP reconciliation. Earnings per Share 109% CAGR $0.64 2019 $1.29 2020 $2.79 2021#197OUR JOURNEY SIGNIFICANT CASH GENERATION¹ $0.3B 2019 Free Cash Flow $0.8B 2020 $3.2B 2021 Free Cash Flow Margin 4% 2019 (1) Non-GAAP financial measures. See appendices for GAAP to non-GAAP reconciliation. 8% 2020 20% 2021#198TH Technology and Product Leadership filte XILINX TRANSFORMATIONAL ACQUISITION HIGH-PERFORMANCE AND ADAPTIVE COMPUTING LEADER Large and Diversified Markets New Customer Opportunities Non-GAAP Margin Expansion Non-GAAP EPS and Free Cash Flow Accretive for 2022#199AMD AND XILINX 2021 COMBINED REVENUE -$4B ~25% Revenue PCs & Gaming Data Center & Embedded $16.4B VS. PROFORMA 1 $20.1B Higher Data Center & Embedded Revenue Mix (1) See appendices for reconciliation.. ~$8B ~40% Revenue#200A ADVANCING OUR JOURNEY#201NEW STRATEGIC MARKETS EXPANDING LONG-TERM GROWTH OPPORTUNITIES $50B $37B PCs Gaming Look $125B Data Center $33B Embedded ~$300B TAM Based on AMD Internal data. $27B $32B Automotive Communications#202Fillite 11 LONG-TERM FINANCIAL PRIORITIES Revenue Growth and Diversification Expanding Margins and Profitability Disciplined Capital Allocation Significant Shareholder Returns#203STRONG AND COMPELLING LONG-TERM FINANCIAL MODEL ¹,2 Revenue Growth Based on 2021 Proforma (AMD + Xilinx) Gross Margin Operating Expenses Operating Margin Effective Tax Rate FCF Margin ~20% CAGR >57% ~23% -24% ~mid-30s% ~13% >25% > > Continued Roadmap Execution Market Share Momentum Richer Product Mix Cost Improvements Investing in Growth Expanding Margins and Profitability Combined Company Tax Structure Significant Cash Generation. (1) Non-GAAP financial measures except for Revenue; includes 2022 Financial guidance. See Cautionary Statement on Slide 2. (2) Long-term model timeframe is ~3 - 4 years#204LONG-TERM REVENUE GROWTH ~$8B ~40% Revenue PCs & Gaming 2021 PROFORMA ¹ $20.1B Data Center & Embedded VS. LTM ~20% CAGR >50% Revenue Significantly Higher Data Center & Embedded Revenue Mix (1) See appendices for reconciliation.#205LONG-TERM GROSS MARGIN DRIVERS¹ 48% 2021 AMD Xilinx 52% 2021 AMD + Xilinx Proforma Server, PCs & Gaming Embedded & New Data Center Cost & Mix Improvement Diversification & Scale Driving Margin Expansion. (1) Non-GAAP financial measures. See appendices for 2021 GAAP to non-GAAP reconciliation and Cautionary Statement on Slide 2. >57% Long Term Model#206XILINX SYNERGIES COST SYNERGIES >$10B Al, Data Center, Communications, Automotive, Embedded >$400M Higher than previous ~$300M expectation + IDENTIFIED REVENUE OPPORTUNITIES#207LONG-TERM CAPITAL ALLOCATION PRIORITIES Invest in the Business M&A Shareholder Returns Balance Sheet Technology, Products, Go-to-Market Infrastructure and Talent Strategic M&A Technology tuck-ins and talent acquisition Return >40% of Free Cash Flow Significant Net Cash Position. Strong Investment Grade Rating Capital Expenditures ~2% -3% of Revenue#2082022 FINANCIAL OUTLOOK¹ Revenue Growth Gross Margin Operating Expenses / Revenue Effective Tax Rate Diluted Share Count ~$26.3B | +~60% Y/Y ~54% ~24% ~13% of pre-tax income ~1.58 billion shares (1) Non-GAAP financial measures except for Revenue. See Cautionary Statement on Slide 2.#209NEW FINANCIAL REPORTING SEGMENTS ▪ Data Center Server CPU, Data Center GPU, FPGAS and Adaptive SoCs Embedded Xilinx (excluding Data Center) and AMD Embedded Client Desktop and Notebook Processors ▪ Gaming Gaming GPUs and Game Console SoCs Starting Q2 2022 Historical annual 2020 and quarterly 2021 Revenue and Operating Income to be provided#210141 FINANCIAL MOMENTUM Significant Margin Expansion Large and Diverse Markets Strong Revenue Growth Disciplined Capital Allocation Positioned for Sustained Growth and Higher Profitability#211APPENDICES#212RECONCILIATION OF GAAP TO NON-GAAP FINANCIAL MEASURES (Millions, except percentages) GAAP Gross Margin GAAP Gross Margin % Stock-based compensation Non-GAAP Gross Profit Non-GAAP Gross Margin % GAAP Operating Income GAAP Operating Margin % Stock-based compensation Acquisition-related costs Loss contingency on legal matter Non-GAAP Operating Income Non-GAAP Operating Margin % $ $ SA GA 2019 2,868 43% 6 2,874 43% 631 9% 197 12 840 12% $ $ $ 2020 4,347 45% 6 4,353 45% 1,369 14% 274 14 1,657 17% $ $ $ $ 2021 7,929 48% 5 7,934 48% 3,648 22% 379 42 4,069 25%#213RECONCILIATION OF GAAP TO NON-GAAP FINANCIAL MEASURES (Millions, except per share data) GAAP Net Income / earnings per share Loss on debt redemption/conversion (Gains) losses on equity investments, net Non-cash interest expense related to convertible debt Stock-based compensation Equity income in investee Acquisition-related costs Loss contingency on legal matter Release of valuation allowance on deferred tax assets Income tax provision Non-GAAP Net Income / earnings per share Shares used and net income adjustment in earnings per share calculation Shares used in per share calculation (GAAP) (1) Interest expense add back to GAAP net income (1) Shares used in per share calculation (Non-GAAP) Interest expense add back to Non-GAAP net income 2019 $ 341 176 22 197 - 12 8 $ 756 $ $ 0.30 0.15 0.02 0.16 0.01 0.64 1,120 $ 1,209 $ 16 2020 $ 2,490 54 $ 6 274 (5) 14 (1,301) 43 1,575 $ $ 2.06 0.04 0.22 0.01 (1.07) 0.03 1.29 1,207 $ 1 1,228 $ 4 $ 3,162 7 (56) $ 2021 379 (6) 42 I I (93) 3,435 $ 2.57 $ (0.04) 0.31 0.03 I (0.08) 2.79 1,229 $ 1,229 (1) For the year ended December 28, 2019, GAAP diluted EPS calculation did not include shares related to the assumed conversion of the Company's 2026 Convertible Notes and the associated interest expense add-back to net income under the "if converted" method as their inclusion would have been anti-dilutive. For the year ended December 26, 2020, GAAP diluted EPS calculations include 3 million shares related to the assumed conversion of the Company's 2026 Convertible Notes and the associated $1 million interest expense add-back to net income under the "if converted" method. For the year ended December 26, 2020 and December 28, 2019, Non-GAAP diluted EPS calculations include 24 million and 89 million shares, respectively, related to the assumed conversion of the Company's 2026 Convertible Notes and the associated $4 million and $16 million interest expense, respectively, add-back to net income under the "if converted" method.#214RECONCILIATION OF GAAP TO NON-GAAP FINANCIAL MEASURES Shares (millions) (1) Basic Shares Dilutive impacts from: Employee Equity Grants (2) 75 million share Warrant (3) Diluted Shares (without 2026 Convertible Notes) 2026 Convertible Notes - GAAP (4) Diluted Shares (with 2026 Convertible Notes) - GAAP 2026 Convertible Notes - Non-GAAP (4) Diluted Shares (with 2026 Convertible Notes) - Non-GAAP 2019 Actual 1,091 22 7 1,120 1,120 89 1,209 2020 Actual 1,184 20 I 1,204 3 1,207 24 1,228 2021 Actual 1,213 16 1,229 1,229 1,229 The table above provides actual share count for FY'19, FY'20, and FY'21 that may be used when calculating GAAP and non-GAAP diluted earnings per share for respective periods. (1) Share counts are weighted average shares. (2) The dilutive impact of employee equity grants is based on the Treasury Stock method and is dependent upon the average stock price during the period. The FY'19, FY'20, and FY'21 average stock price was $29.77, $65.05, and $100.04, respectively. (3) The dilutive impact of the warrant to purchase 75 million shares (Warrant) granted in 2016 to a wholly owned subsidiary of Mubadala Investment Company PJSC, West Coast Hitech L.P. (WCH), in consideration for limited waiver and rights under the sixth amendment to our Wafer Supply Agreement with GLOBALFOUNDRIES Inc. is based on the Treasury Stock method and is dependent upon the average stock price during the period. The Warrant was exercised, and common stock shares were issued on February 13, 2019. Therefore, FY'19 includes the dilutive impact through the date of exercise. (4) The dilutive impact from the 2.125% Convertible Senior Notes due 2026 (2026 Convertible Notes) is based on the If-Converted method, where the interest costs associated with the 2026 Convertible Notes are added back to the Net Income and the shares underlying the 2026 Convertible Notes are assumed to be converted and are added to the share count. The impact from the 2026 Convertible Notes, if dilutive, is included in diluted EPS calculation. For the GAAP computation, the add-back to net income includes cash and non-cash interest expense, while only the cash interest expense is added back to the non-GAAP net income. The dilutive shares associated with the 2026 Convertible Notes reflect the weighted average shares subject to conversion during each period. During FY'19, FY'20, and FY'21, 69 million, 28 million, and 3 million shares of the Company's common stock were issued to convert $554 million, $225 million, and $25 million of the outstanding 2026 Convertible Notes, respectively.#215RECONCILIATION OF GAAP TO NON-GAAP FINANCIAL MEASURES (MILLIONS) GAAP net cash provided by operating activities Operating cash flow margin % Purchases of property and equipment Free cash flow Free cash flow margin % 2019 493 7% (217) 276 4% GA 2020 1,071 11% (294) 777 8% GA 2021 3,521 21% (301) 3,220 20%#216RECONCILIATION OF COMBINED COMPANY PRO FORMA (Millions, except percentages) Net revenue GAAP Gross Margin GAAP Gross Margin % Stock-based compensation Amortization of acquisition-related intangibles Acquisition-related costs Non-GAAP Gross Profit Non-GAAP Gross Margin % 2021 AMD Classic $ $ $ 16,434 7,929 48% 5 7,934 48% 2021 Xilinx Classic (1) $ $ 3,676 2,514 68% 16 37 1 2,568 70% $ 2021 Proforma 20,110 10,443 52% 21 37 1 10,502 52% (1) The unaudited Xilinx Classic Proforma revenue and gross margin are based on the unaudited financial results for the trailing four-quarter period ended January 1, 2022, and the Non-GAAP gross margin is adjusted to align with AMD Non-GAAP calculation.#217FINANCIAL ANALYST DAY 2022 together we advance_ Advancing Our Journey Dr. Lisa Su, Chair and CEO AMD#218Large and Compelling $300B TAM FINANCIAL ANALYST DAY 2022 OUR NEXT JOURNEY Unmatched Technology Leadership Expanding Data Center Leadership and Pervasive AI World-Class Execution and Focus Best-in-Class Growth Driving Compelling Shareholder Returns#219AMD together we advance_