《英偉達（NVIDIA）：2025財年英偉達公司概況概覽報告（英文版）（37頁）.pdf》由會員分享，可在線閱讀，更多相關《英偉達（NVIDIA）：2025財年英偉達公司概況概覽報告（英文版）（37頁）.pdf（37頁珍藏版）》請在三個皮匠報告上搜索。

1、Company OverviewFebruary 2025Except for the historical information contained herein,certain matters in this presentation including,but not limited to,statements as to:expectations with respect to growth,performance and benefits of NVIDIAs products,services,and technologies,including Blackwell,and re

2、lated trends and drivers;expectations with respect to supply and demand for NVIDIAs products,services,and technologies,including Blackwell,and related matters including inventory,production and distribution;our financial position;projected market growth and trends,market opportunity,demand,and growt

3、h drivers;our financial and business outlook;our dividend program;expectations with respect to NVIDIAs third party arrangements,including with its collaborators and partners;third parties adopting our products and technologies;expectations with respect to technology developments and related trends a

4、nd drivers;expectations with respect to AI and related industries;our sustainability goals;and other statements that are not historical facts are forward-looking statements.These forward-looking statements and any other forward-looking statements that go beyond historical facts that are made in this

5、 presentation are subject to risks and uncertainties that may cause actual results to differ materially.Important factors that could cause actual results to differ materially include:global economic and political conditions;NVIDIAs reliance on third parties to manufacture,assemble,package and test N

6、VIDIAs products;the impact of technological development and competition;development of new products and technologies or enhancements to NVIDIAs existing product and technologies;market acceptance of NVIDIAs products or NVIDIAs partners products;design,manufacturing or software defects;changes in con

7、sumer preferences and demands;changes in industry standards and interfaces;unexpected loss of performance of NVIDIAs products or technologies when integrated into systems and other factors;and changes in applicable laws and regulations.NVIDIA has based these forward-looking statements largely on its

8、 current expectations and projections about future events and trends that it believes may affect its financial condition,results of operations,business strategy,short-term and long-term business operations and objectives,and financial needs.These forward-looking statements are subject to a number of

9、 risks and uncertainties,and you should not rely upon the forward-looking statements as predictions of future events.The future events and trends discussed in this presentation may not occur and actual results could differ materially and adversely from those anticipated or implied in the forward-loo

10、king statements.Although NVIDIA believes that the expectations reflected in the forward-looking statements are reasonable,the company cannot guarantee that future results,levels of activity,performance,achievements or events and circumstances reflected in the forward-looking statements will occur.Ex

11、cept as required by law,NVIDIA disclaims any obligation to update these forward-looking statements to reflect future events or circumstances.For a complete discussion of factors that could materially affect NVIDIAs financial results and operations,please refer to the reports we file from time to tim

12、e with the SEC,including NVIDIAs most recent Annual Report on Form 10-K,Quarterly Reports on Form 10-Q,and Current Reports on Form 8-K.Copies of reports we file with the SEC are posted on NVIDIAs website and are available from NVIDIA without chargeMany of the products and features described herein r

13、emain in various stages and will be offered on a when-and-if-available basis.The statements within are not intended to be,and should not be interpreted as a commitment,promise,or legal obligation,and the development,release,and timing of any features or functionalities described for our products is

14、subject to change and remains at the sole discretion of NVIDIA.NVIDIA will have no liability for failure to deliver or delay in the delivery of any of the products,features or functions set forth herein.NVIDIA uses certain non-GAAP measures in this presentation including non-GAAP operating income,no

15、n-GAAP operating margin,and free cash flow.NVIDIA believes the presentation of its non-GAAP financial measures enhances investors overall understanding of the companys historical financial performance.The presentation of the companys non-GAAP financial measures is not meant to be considered in isola

16、tion or as a substitute for the companys financial results prepared in accordance with GAAP,and the companys non-GAAP measures may be different from non-GAAP measures used by other companies.Further information relevant to the interpretation of non-GAAP financial measures,and reconciliations of thes

17、e non-GAAP financial measures to the most comparable GAAP measures,may be found in the slide titled“Reconciliation of Non-GAAP to GAAP Financial Measures.”NVIDIAs invention of the GPU in 1999 sparked the growth of the PC gaming market,redefined computer graphics,revolutionized accelerated computing,

18、ignited the era of modern AI,and is fueling industrial digitalization across markets.Today,two transitions are occurring simultaneouslyaccelerated computing and generative AItransforming the computer industry and every other industry worldwide,and NVIDIA is enabling these transitions with our full-s

19、tack computing platform and data-center-scale offerings.NVIDIAs platform is installed in several hundred million computers,is available in every cloud and from every server maker,powers over 75%of the TOP500 supercomputers,and has 5.9 million developers.NVIDIAHeadquarters:Santa Clara,CA|Headcount:36

20、,000Grace BlackwellMGX NodeNVLink SwitchQuantum SwitchSpectrum-X SwitchChips Purpose-Built for AI SupercomputingGPU|CPU|DPU|NIC|NVLink Switch|IB Switch|ENET SwitchCUDA DOCA NCCLCluster-Scale SoftwareSystem SoftwareChip SoftwareCUDA-X LibrariesNIMCUDA-AcceleratedAgentic AI LibrariesOmniverse CUDA-Acc

21、eleratedPhysical AI LibrariesAccelerated Software StackGB200 NVL72 SuperPODNVIDIAs Accelerated Computing PlatformData center scale innovation across chips,networking,systems,software,and algorithmsNVIDIA has accelerated software and compute by a 1,000,000X in the last decade,far surpassing Moores la

22、w.Accelerated computing requires full-stack innovationoptimizing across every layer of computingfrom chips and systems to software and algorithms,demanding deep understanding of the problem domain.Our platform extends from the cloud and enterprise data centers to supercomputing,edge computing,PCs,an

23、d robotics.What Is Accelerated Computing?Not just a superfast chipaccelerated computing is a full-stack combination of:Chip(s)with specialized processors Algorithms in acceleration libraries Domain experts to refactor applicationsTo speed up compute-intensive parts of an applicationA full-stack appr

24、oach:silicon,systems,softwareFor example:If 90%of the runtime can be accelerated by 100X,the application is sped up 9X If 99%of the runtime can be accelerated by 100X,the application is sped up 50X If 80%of the runtime can be accelerated by 500X,or even 1,000X,the application is sped up 5X Amdahls l

25、aw:The overall system speed-up(S)gained by optimizing a single part of a system by a factor(s)is limited by the proportion of execution time of that part(p).=11 +Why Accelerated Computing?Accelerated computing is needed to tackle the most impactful opportunities of our timelike AI,climate simulation

26、,drug discovery,ray tracing,and robotics.NVIDIA is uniquely dedicated to accelerated computing working top-to-bottom,refactoring applications and creating new algorithms,and bottom-to-top inventing new specialized processors,like RT Cores and Tensor Cores.Advancing computing in the post-Moores law e

27、ra102103104105106107109108198019902000201020202030GPU-Computing perf2X per year1,000XIn 10 yearsSingle-threaded CPU perf1.5X perf per year1.1X per yearTrillions of Operations per Second(TOPS)“Its the end of Moores law as we know it.”John Hennessy,Oct 2018“Moores law is dead.”Jensen Huang,GTC 2013 Th

28、e NVIDIA accelerated computing platform has attracted the largest ecosystem of developers,supporting a rapidly growing universe of applications and industry innovation.Developers can engage with NVIDIA through CUDAour parallel computing programming model introduced in 2006or at higher layers of the

29、stack,including libraries,pretrained AI models,SDKs,and other development tools.NVIDIAs Accelerated Computing EcosystemDevelopersCUDA Downloads*AI StartupsGPU-Accelerated Applications202120247K19K202120245.1M2.5M2021202453M26M202120243,7001700*CumulativeAI Driving a Powerful Investment Cycle and Sig

30、nificant ReturnsAI Agents will take action to automate tasks at superhuman speed,transforming businesses and freeing workers to focus on other tasks.Copilots based on LLMs will generate documents,answer questions,or summarize missed meetings,emails,and chatsadding hours of productivity per week.Spec

31、ialized for fields such as software development,legal services or education and can boost productivity by as much as 50%.Social media,search,and e-commerce apps are using deep recommenders to offer more relevant content and ads to their customers,increasing engagement and monetization.Creators can g

32、enerate stunning,photorealistic images with a single text promptcompressing workflows that take days or weeks into minutes in industries from advertising to game development.Call center agents augmented with AI chatbots can dramatically increase productivity and customer satisfaction.Drug discovery

33、and financial services are seeing order-of-magnitude workflow acceleration from AI.Manufacturing workflows are reinvented and automated through generative AI and robotics,boosting productivity.AI can augment creativity and productivity by orders of magnitude across industriesSource:Goldman Sachs,Cow

34、en,Statista,Capital One,Wall Street Journal,Resource Watch,NVIDIA internal analysisAI Content Creation50M creators globallyFinancial Services678B annual credit card transactionsManufacturing$50T of heavy industryCustomer Service 15M call center agents globallyAI Agents&CopilotsOver 1B knowledge work

35、ersLegal Services,Education 1M legal professionals in the US9M educators in the USSearch&Social Media$700B in digital advertising annuallyAI Software Development30M software developers globallyDrug Discovery1018 molecules in chemical space40 exabytes of genome dataGenerative AIThe most important com

36、puting platform of our generationThe era of generative AI has arrived,unlocking new opportunities for AI across many different applications.Generative AI is trained on large amounts of data to find patterns and relationships,learning the representation of almost anything with structure.It can then b

37、e prompted to generate text,images,video,code,or even proteins.For the very first time,computers can augment the human ability to generate information and create.1,600+generative AI companies are building on NVIDIA.TEXTSOUNDTEXTTEXTIMAGEVIDEOSPEECHMULTI-MODALAMINO ACIDBRAINWAVESSPEECHIMAGEVIDEOIMAGE

38、3DANIMATIONMANIPULATIONPROTEINLearn and UnderstandEverythingDelivering Tremendous Value to CustomersSignificant reduction in TCO with each generationHopper|8,000 GPUs|15MWBlackwell GB200 NVL72|2,000 GPUs|4MWLLM Training Workload:GPT-MoE-1.8T|Train in 90 days|H100 vs GB200 NVL724XReduction in PowerAI

39、 FactoriesA New Class of Data CentersAI factories are a new form of computing infrastructure.Their purpose is not to store user and company data or run ERP and CRM applications.AI factories are highly optimized systems purpose-built to process raw data,refine it into models,and produce monetizable t

40、okens with great scale and efficiency.In the AI industrial revolution,data is the raw material,tokens are the new commodity,and NVIDIA is the token generator in the AI factory.Every company will produce digital intelligence.Tokens will be transformed into intelligent responses and actions of digital

41、 nurses,tutors,customer service agents,chip designers,manufacturing robots and autonomous cars,Weather prediction agents will warn us of storms.Some companies will build and operate AI factories,while others will rent.Countries are awakening to the need to treat their data as a national resource and

42、 AI factories as an essential national infrastructure.Data encodes a nations history,knowledge,and culture,and can be transformed into the sovereign AI for its companies,startups,universities,and governments.NVIDIA builds the complete AI system and licenses NVIDIA AI Enterprise,the AI stack and oper

43、ating system for AI factories.Production of digital intelligence tokensDataTokensAI FactoryExtending NVIDIA Networking to Scale Up&Scale Out AI in Any Data CenterNew NVLink and Spectrum-X increase networking opportunity beyond InfiniBand to every data centerGenerative AI Is a Data Center-Scale Compu

44、ting WorkloadLimitless scaling with NVLINK+InfiniBand or Spectrum-XNVIDIA NVLinkFastest Interconnect for GPU Scale-UpNVIDIA InfiniBandSupercomputing and Dedicated AI FactoriesNVIDIA Spectrum-XEthernet Optimized for Multi-Tenant AI Factories#of GPU in a Data CenterNVIDIA Spectrum-X AI Ethernet Fabric

45、101001K100K10K1M+AI PerformanceInfiniBandNVLink+InfiniBand/Spectrum-XTraditional EthernetAccelerated Computing Starts With CUDA LibrariesUnlike CPU general-purpose computing,GPU-accelerated computing requires software and algorithms to be redesigned.Software is not automatically accelerated in the p

46、resence of a GPU or accelerator.NVIDIA CUDA libraries encapsulate NVIDIA-engineered algorithms that enable applications to be accelerated on NVIDIAs installed base.They deliver dramatically higher performancecompared to CPU-only alternativesacross application domains,including AI and high-performanc

47、e computing,and significantly reduce runtime,cost,and energy,while increasing scale.With over 400 CUDA libraries,NVIDIA can address many major workloads across a wide range of industries.As new libraries become available,they unlock new markets adding to our long-term opportunity.Delivering up to 20

48、0X speedup across major workloadsSpeech AIRiva,TensorRT,Triton Inference Server,NeMo,cuBLAS,cuDNN,cuFFT,CUTLASS 30X30XRecommender SystemsMerlin,HugeCTR,TensorRT,Triton Inference Server,cuBLAS,cuDNN,cuFFT,cuSPARSE,CUTLASS,Magnum IO,NCCL,cuVS 50X50XDeep LearningcuDNN,CUTLASS,Megatron,TensorRT,TRT LLM,

49、NCCL,NV-Triton,CUDA-optimized PyTorch,Tensorflow,Triton,Jax 100X100XScienceEarth-2 CorrDiff,Holoscan,Parabricks,Monai,Modulus,Warp,cuLitho,cuQuantum,CUDA-Q,AmgX,cuDSS,cuFFT,cuSOLVER,cuBLAS,cuSPARSE,cuTENSOR,cuGraph,Magnum IO,NCCL,NVSHMEM,RAFT,cuNumeric,Sionna 100X100XAgentic and Physical AIACE,Riva,

50、Nemo,Tokkio Digital Human,Holoscan,Metropolis,Omniverse,Isaac,DRIVE,cuLitho,cuMotion,cuOpt,Aerial CUDA-accelerated RAN,Sionna,fVDB,PhysX,Warp,NVblox 100X100XComputer VisionCV-CUDA,Deepstream,TAO,Holoscan,cuCIM,TensorRT,Triton Inference Server,DALI,nvImageCodec,cuDNN,nvJPEG,nvJPEG2000,nvTIFF,NPP,Vide

51、o Codec SDK,Magnum IO,NCCL,cuVS,DALI 200X200XData ProcessingcuVS,cuDF-Spark,cuDF-pandas,cuDF-Polars,cuGraph,cuML,XGBoost,RAPIDS,NeMo Curator,cuSOLVER,cuIO 200X200XAccelerated Computing Is Sustainable ComputingOrder of magnitude more energy efficientEnergy Usage in AI InferenceGPT-MoE-1.8T energy per

52、 tokenAccelerated computing requires higher peak power consumption than CPUs,however,completes workloads significantly faster and consumes less total energyTime(sec)Server TDP(kW)Kepler42000 Joules/TokenPascal17640 J/TokenVolta1200 J/TokenAmpere150 J/TokenHopper10 J/TokenBlackwell0.4 J/Token20142024

53、Accelerated computing enables full-stack optimization from algorithm to GPU architecture,such as Tensor Core Transformer Engine;LLM energy efficiency improved 100,000X in the past 10 yearsCPUNVIDIAAI Scaling Laws Drive Exponential Demand for ComputeNew OpenAI o1 Long“thinking time”creates a new way

54、to scaleInference compute scales exponentially with larger models,multimodality,large context,low latency,and nowlong“thinking time”Training compute scales exponentially with larger models,multimodality,reinforcement learning,and synthetic data generationTrainingComputeInferenceComputeo1NVIDIA Is th

55、e Leading Inference PlatformInference compute scales exponentially with“long thinking”Flash AttentionKVCache PageAttentionDistillationPruning&QuantizationNeural Architecture SearchDisaggregated ServingSpeculative DecodingMulti-GPU,Multi-NodeHopper inference performance increased 5X in 1 year with ra

56、pid algorithm innovationsenabled by rich NVIDIA CUDA ecosystemInstalled base&CUDA rapid software innovation performance lower inference cost increase demand increase installed baseInference compute scaling exponentially with large multimodal models,chain-of-thought,reasoning,agents,and low-latency r

57、esponsesGB200 NVL72NVLink Switches130 TB/s All-to-All BWOne Giant Blackwell1.44 EF FP4576TB/s HBM3eNVIDIA NVLink Enables New Level of AI Training&Inference ScalingNVLink SwitchGB200 NVL72NVLink Switches130 TB/s All-to-All BWOne Giant Blackwell1.44 EF FP4576TB/s HBM3eEcosystemNVIDIA AI Platform and E

58、cosystem Reaches Every MarketEvery workload to address the worlds industriesAccelerate Every WorkloadFull-Stack,Entire AI InfrastructureSoftwareData Pre-ProcessingTrainingPostTraininge.g.,SDGAgentic AIInferenceRobotic AIInferenceAI InfrastructureAI TechnologyEvery CloudOEMs and ODMsPC and Workstatio

59、nsEdge and RoboticsFull-StackCompute-to-NetworkingSovereign AIRegional CSPsTelcosInternetServicesPublicCloudsHeavyIndustriesEnterprisesAuto,Healthcare,Logistics,Energy,FSI,etc.AIStartupsSaaSSocialMediaPlatformsNVIDIA AI Enterprise Enables IT Ecosystem With State-of-the-ArtAI Models and Libraries to

60、Build Agentic AI.Cloud and On-Prem InfrastructureSystem Integrators.Enterprise ISVsNVIDIA AI Enterprise EcosystemNVIDIA NeMoUserNVIDIA NeMo RetrieverVector DatabaseActionStructuredDatabaseNVIDIA NIMDataFlywheelAI Agent.Data Platforms.NVIDIA Omniverse and AI Revolutionizing Manufacturing&Robotics10M

61、Factories200K Warehouses100M CarsBillions in FutureThe next AI wave is physical AImodels that can perceive,understand,and interact with the physical world.Physical AI will embody robotic systemsfrom autonomous vehicles to industrial robots and humanoids,to warehouses and factories.Three computers an

62、d software stacks are required to build physical AI:NVIDIA AI on DGX to train the AI model,NVIDIA Omniverse on OVX to teach,test,and validate the AI models skills,and NVIDIA AGX to run the AI software on the robot.Enterprises license NVIDIA Omniverse at$4,500 per GPU per year.Sovereign AINations pro

63、duce AI using their own data,infrastructure,workforce,and business networksJapanNational Institute of Advanced Industrial Science and Technology(AIST)SwitzerlandSwisscom GroupEcuadorTelconetSingaporeSingapore Telecommunications Limited(Singtel)FranceScalewayVietnamFPT Smart CloudSpainBarcelonaSuperc

64、omputing CenterLocation of NVIDIA sovereign AI partnersSovereign AINations are awakening to the imperative to produce AI using their own infrastructure,data,workforces,and business networks.Nations are building domestic computing capacity.Some governments operate sovereign AI clouds in collaboration

65、 with state-owned telecommunications providers or utilities.Other governments partner with local cloud providers to deliver a shared AI computing infrastructure for public and private-sector use.NVIDIAs ability to help build AI infrastructure with our end-to-end compute-to-networking technologies,fu

66、ll-stack software,AI expertise,and rich ecosystem of partners and customers allows sovereign AI and regional cloud providers to jump-start their countries AI ambitions.$6,803$12,690$9,040$37,134$86,78941%47%34%61%67%30%40%50%60%70%$0$10,000$20,000$30,000$40,000$50,000$60,000$70,000$80,000$90,000FY21

67、FY22FY23FY24FY25 Operating Income(Non-GAAP,$M)Operating Margin(Non-GAAP)$16,675$26,914$26,974$60,922$130,497FY21FY22FY23FY24FY25Driving Strong and Profitable GrowthFiscal year ends in January.Refer to Appendix for reconciliation of Non-GAAP measures.Operating margins rounded to the nearest percent.R

68、evenue($M)$4.7B$8.0B$3.8B$26.9B$60.7BFY21FY22FY23FY24FY25Strong Cash Flow GenerationFiscal year ends in January.Refer to Appendix for reconciliation of Non-GAAP measures.Share RepurchaseUtilized$33.7B of cash for repurchases in FY25$38.7B remaining authorization as of the end of Q4Dividend$834M in F

69、Y25Dividend increased by 150%in Q2 FY25Plan to Maintain1Strategic InvestmentsGrowing Our Talent Platform Reach and EcosystemFree Cash Flow(Non-GAAP)Capital Allocation1 Subject to continuing determination by our Board of Directors.DRIVE Hyperion sensor architecture with AGX computeDRIVE AV&IX full-st

70、ack software for ADAS,AV,and AI cockpitDGX/HGX/MGX/IGX systemsGPU|CPU|DPU|NetworkingNVIDIA AI softwareOur Market Platforms at a GlanceFY25 Revenue$115.2B5-YR CAGR 108%FY25 Revenue$11.4B5-YR CAGR 16%FY25 Revenue$1.9B5-YR CAGR 9%FY25 Revenue$1.7B5-YR CAGR 19%GeForce GPUs for PC gamingGeForce NOW cloud

71、 gamingNVIDIA RTX GPUs for workstationsOmniverse softwareProfessional VisualizationAutomotive9%of FY25 Revenue1%of FY25 Revenue1%of FY25 RevenueData Center88%of FY25 RevenueGaming$6,696$10,613$15,005$47,525$115,186FY21FY22FY23FY24FY25Data CenterThe leading accelerated computing platformLeader in AI

72、and HPCNo.1 in AI training and inferenceUsed by all hyperscalers,major cloud computing providers,and over 40,000 companiesPowers over 75%of the TOP500 supercomputersGrowth DriversBroad data center platform transition from general-purpose to accelerated computingEmergence of AI factoriesoptimized for

73、 refining data and training,inferencing,and generating AIBroader and faster product launch cadence to meet a growing and diverse set of AI opportunitiesNVIDIA AI Enterprise/NIM for building and running enterprise AI applications108%5-YR CAGR Through FY25Revenue($M)RubinBlackwellHopperBlackwell-Ultra

74、NVLink Switch900 GB/secCX7 SuperNICHopper GPU6S HBM3Hopper+GPU6S HBM3eBF3 SuperNICQuantum-X400Infiniband SwitchGrace CPUNVLink 6 Switch3600 GB/secCX9 SuperNIC1600 Gb/secRubin GPU8S HBM4Vera CPUX1600IB/Ethernet SwitchRubin Ultra GPU12S HBM4202220242026202320252027One-Year Rhythm|Supercluster Scale|Fu

75、ll-Stack|CUDA EverywhereSupercharge AI scaling lawX-FactorsX-FactorsX-FactorsCX8 SuperNICSpectrum Ultra X800Ethernet Switch 512-RadixBlackwell Ultra GPU288GB HBM3eMore AI FLOPS$7,759$12,462$9,067$10,447$11,350FY21FY22FY23FY24FY25GamingGeForceworlds largest gaming platformLeader in PC GamingStrong No

76、.1 market position15 of the top 15 most popular GPUs on SteamLeading performance and innovation200M+gamers on GeForceGrowth DriversRising adoption of NVIDIA RTX in gamesExpanding universe of gamers and creatorsGaming laptops and generative AI on PCsGeForce NOW cloud gaming16%5-YR CAGR Through FY25Re

77、venue($M)$1,053$2,111$1,544$1,553$1,878FY21FY22FY23FY24FY25Professional VisualizationWorkstation graphicsLeader in Workstation Graphics95%+market share in graphics for workstations45M designers and creatorsStrong software ecosystem with over 100 RTX accelerated and supported applicationsGrowth Drive

78、rsGenerative AI adoption across design and creative industriesEnterprise AI development,model fine-tuning,cross-industryRay tracing revolutionizing design and content creationExpanding universe of designers and creatorsOmniverse for digital twins and collaborative 3D designHybrid work environmentsRe

79、venue($M)9%5-YR CAGR Through FY25$536$566$903$1,091$1,694FY21FY22FY23FY24FY25AutomotiveAutonomous vehicles and AI cockpitsLeader in Autonomous Driving NVIDIA DRIVE an end-to-end autonomous vehicle(AV)and AI cockpit platform featuring a full software stack and powered by NVIDIA SoCs(systems-on-a-chip

80、)in vehiclesDRIVE Orin SoC ramp began in FY23Next-generation DRIVE Thor SoC ramp to begin in FY26Over 40 customers including 20 of top 30 EV makers,7 of top 10 truck makers,8 of top 10 robotaxi makersGrowth DriversAdoption of centralized car computing and software-defined vehicle architecturesAV sof

81、tware and services:Mercedes-BenzJaguar Land RoverRevenue($M)19%5-YR CAGR Through FY25The$1T installed base of general-purpose CPU data center infrastructure is being modernized to a new GPU-accelerated computing paradigm.The entire computing stack has been reinventedfrom CPU to GPU,from coding to ma

82、chine learning,from software to generative AI.Computers generate intelligence tokens,a new commodity.A new type of data center,AI factories,is expanding the data center footprint to$2T and beyond in the coming years.Eventually,companies in every industry will operate AI factories as the digital twin

83、 of their workforce,manufacturing plants,and products.A new industrial revolution has begun.Accelerated Computing and Generative AI Create Trillion-Dollar OpportunitiesGPU-AcceleratedData CentersAI FactoriesTraditionalData CentersGeneral-PurposeComputingAIAcceleratedComputingFinancialsAnnual Cash&Ca

84、sh Flow Metrics5,8229,1085,64128,09064,089FY21FY22FY23FY24FY256,80312,6909,04037,13486,789FY21FY22FY23FY24FY254,6778,0493,75026,94760,724FY21FY22FY23FY24FY2511,56121,20813,29625,98443,210FY21FY22FY23FY24FY25Cash balance is defined as cash and cash equivalents plus marketable securitiesRefer to Appen

85、dix for reconciliation of non-GAAP measuresFree Cash Flow(Non-GAAP)$MCash Balance$MOperating Income(Non-GAAP)$MOperating Cash Flow$MCorporate SustainabilityFast Company Magazines Worlds Most Innovative CompaniesFortunes Worlds Most Admired CompaniesTime Magazines 100 Most Influential CompaniesWall S

86、treet Journals Management Top 250“Americas Most Sustainable Companies”BARRONS“Americas 100 Best Companies to Work For”FORTUNE“Americas Most Responsible Companies”NEWSWEEKNVIDIA Blackwell GPUs are as much as 20X more energy efficient than CPUs for certain AI and HPC workloadsOn track to source 100%re

87、newable electricity for offices and data centers under operational control by end of FY25A Place for People to Do Their Lifes WorkEnvironmentally ConsciousOn track to engage manufacturing suppliers comprising at least 67%of scope 3 category 1 GHG emissions with the goal of effecting supplier adoptio

88、n of science-based targets by end of FY26Management92%of directors are independentCorporate Governance“Best Places to Work”GLASSDOORReconciliation of Non-GAAP to GAAP Financial MeasuresOperating Income and Margin($in Millions and Margin Percentage)Non-GAAPAcquisition Termination CostAcquisition-Rela

89、tedand Other Costs(A)Stock-Based Compensation(B)Other(C)GAAPFY 2021$6,803(836)(1,397)(38)$4,53240.8%(5.0)(8.4)(0.2)27.2%FY 2022$12,690(636)(2,004)(9)$10,04147.2%(2.5)(7.4)37.3%FY 2023$9,040(1,353)(674)(2,710)(79)$4,22433.5%(5.0)(2.5)(10.0)(0.3)15.7%FY 2024$37,134(583)(3,549)(30)$32,97261.0%(1.0)(5.8

90、)(0.1)54.1%FY 2025$86,789(602)(4,737)3$81,45366.5%(0.5)(3.6)62.4%Reconciliation of Non-GAAP to GAAP Financial MeasuresA.Consists of amortization of acquisition-related intangible assets,inventory step-up,transaction costs,compensation charges,and other costsB.Stock-based compensation charge was allo

91、cated to cost of goods sold,research and development expense,and sales,general and administrative expenseC.Comprises of legal settlement cost,contributions,restructuring costs and assets held for sale related adjustments($in Millions)Free Cash FlowPurchases Related to Property and Equipment and Inta

92、ngible AssetsPrincipal Payments on Property and Equipment and Intangible AssetsNet Cash Provided by Operating ActivitiesFY 2021$4,6771,12817$5,822FY 2022$8,04997683$9,108FY 2023$3,7501,83358$5,641FY 2024$26,9471,06974$28,090FY 2025$60,7243,236129$64,089Reconciliation of Non-GAAP to GAAP Financial Measures(contd.)