《凯捷：2025人工智能（AI）时代的半导体产业研究报告：未来需求驱动创新（英文版）（92页）.pdf》由会员分享，可在线阅读，更多相关《凯捷：2025人工智能（AI）时代的半导体产业研究报告：未来需求驱动创新（英文版）（92页）.pdf（92页珍藏版）》请在三个皮匠报告上搜索。

1、The semiconductor industry in the AI eraInnovating for tomorrows demands#GetTheFutureYouWantTable of contentsWho should read this report and why?Snapshot of research methodologyOrganizations anticipate surging semiconductor demand The semiconductor industry is innovating,but softwarization remains a

2、 challengeDownstream industries express concerns over semiconductor supply 2Capgemini Research Institute 2025The semiconductor industry in the AI era040708102032Executive summaryResilience and sustainability gather momentumExploring big techs shift to in-house chip designPartnerships are taking the

3、industry forwardHow the semiconductor industry can capitalize on emerging opportunitiesResearch methodologyConclusion3Capgemini Research Institute 2025The semiconductor industry in the AI era465457607271Executive summary AI adoption is powering a surge in demand for semiconductors:While semiconducto

4、r industry organizations forecast a 15%rise in two years,the downstream organizations(those reliant on semiconductor supply for their products or services and operations)anticipate their demand for chips to increase at a higher growth rate of 29%.Increased adoption of artificial intelligence(AI)and

5、generative AI(Gen AI)is driving the need for specialized neural processing units(NPUs)and high-performance graphics processing units(GPUs)that can handle massive computations and large datasets efficiently.Additionally,downstream organizations expect their demand for AI chips,custom silicon chips,an

6、d memory-intensive chips to increase over the next 12 months.Amid buoyant demand lies concern:Over half of downstream organizations doubt the semiconductor industrys ability to meet their needs.Technological advancements,including GPU computing for AI and machine learning and inference acceleration

7、are important to these industries,which are continually seeking to enhance customization,introduce more comprehensive application programming interfaces(APIs)and software development kits(SDKs),as well as reinforcing security features.Consequently,one in three downstream organizations is exploring o

8、r is actively engaged in in-house chip design,enabling greater customization while also gaining more control over their supply chains.Further,sustainability,supply chain resilience,and security are critical concerns for downstream organizations going forward.Amid“softwarization”challenges,innovation

9、 shines through:The semiconductor industry continues to excel in innovation across a number of areas.Although various players stand out in specific areas,our analysis reveals three types of innovation that are consistently prioritized across the industry:Design innovation:Advances in chip architectu

10、res,such as 3D integrated circuit(IC)design and multi-die integration are pushing the boundaries of performance and energy efficiency,while half of design organizations are investing in Gen AI to shorten design cycles.4Capgemini Research Institute 2025The semiconductor industry in the AI eraExecutiv

11、e summary Manufacturing innovation:Advances in extreme ultraviolet(EUV)lithography and the shift towards smaller process nodes(i.e.,3 nanometer and 2 nm)enable the production of more powerful and efficient chips.Nearly half of manufacturers also rely on AI and ML to optimize processes.Packaging inno

12、vation:3D packaging and the use of chiplets(tiny integrated circuits that can be combined to create complex components)are enhancing functionality and performance without increasing physical footprint.Hardware security remains paramount,with significant investment in secure chip design,hardware-base

13、d encryption,and root of trust(RoT,a trusted source within a cryptographic system)technologies.But while there is steady progress in integrating software and hardware to create more adaptable and programmable semiconductor solutions,monetization remains a hurdle.However,the“softwarization of semicon

14、ductors”is falling short of the industrys expectations.While this innovation is crucial for semiconductor companies to expand use cases,extend the chip lifecycle,and enhance customization in an evolving market,the industry finds monetizing its software a challenge.Focus on supply chain resilience an

15、d sustainability:Only two in five semiconductor organizations are confident in the resilience of their supply chains.Organizations focus on onshoring and“friendshoring”(basing supply chains in countries that are geopolitical allies)to enhance stability and reduce single-region dependency.Consequentl

16、y,the industry anticipates domestic sourcing will improve by 17%over the next two years.Three-quarters(74%)of organizations expect to increase their US investments,and 59%will increase investment in the EU.Besides its continued focus on improving power efficiency,the industry is becoming more eco-fr

17、iendly by cutting energy consumption,implementing water recycling and reuse systems,using less toxic alternative chemicals,and minimizing waste.5Capgemini Research Institute 2025The semiconductor industry in the AI eraExecutive summary Recommendations for the semiconductor industry:To capitalize on

18、emerging opportunities,semiconductor organizations should consider the following:Utilize AI and Gen AI to automate design processes,improve production efficiency,and optimize performance to meet the specialized needs of emerging applications.Invest in cutting-edge fabrication methods such as 3D chip

19、 stacking and accelerate research in emerging fields such as advanced silicon photonics integration.Diversify supplier networks across multiple regions while investing in R&D for alternative materials and technologies.Implement sustainable manufacturing practices such as green chemistry and utilize

20、renewable energy sources to minimize carbon footprint.Coordinate strategies with governmental initiatives such as grants for R&D and collaborate within the industry ecosystem to drive shared innovation and standardization.Enhance cybersecurity measures to protect data integrity,use advanced security

21、 to safeguard proprietary technologies,and advocate for stronger intellectual property(IP)laws to deter infringement and protect innovation-led competitive advantage.Adopt open standards and open-source collaboration to drive semiconductor innovation.6Capgemini Research Institute 2025The semiconduct

22、or industry in the AI eraWho should read this report and why?This report will be relevant for decision-makers across the semiconductor ecosystem and its down-stream industries.Specifically,this will be useful for:1.Executives in the semiconductor industry:Integrated device manufacturers,fabless desi

23、gn firms,foundries,OSAT companies,material and subsystem companies,and semiconductor capital equipment manufacturers will gain a strategic outlook on industry trends,including advancements in design,manufacturing,and packaging.The report also addresses critical topics like cybersecurity,softwarizati

24、on,supply chain resilience,and sustainabilityenabling executives to align their strategies for the future.2.Leaders in downstream industries:Professionals in automotive,consumer electronics,retail,telecom,aerospace and defense,high tech,medical devices/medical electronics,industrial equipment,financ

25、ial services,and energy industries will gain a deeper understanding of how semiconductor trends impact their industries.Insights into custom silicon,in-house chip design,sustainability,and supply chain dynamics provide guidance for aligning technology adoption with organizational goals.By connecting

26、 perspectives from both semiconductor executives and downstream organizations,this report equips stakeholders with the knowledge and strategies needed to thrive in a rapidly evolving landscape.7Capgemini Research Institute 2025The semiconductor industry in the AI eraSnapshot of research methodology3

27、.Twelve in-depth interviews with executives from the semiconductor industry and downstream industries.1.Survey of 250 semiconductor industry organizations2.Survey of 800 downstream organizations1%6%11%11%10%10%10%10%13%13%6%18%12%10%8%3%48%Integrated device manufacturersFabless designOutsourced semi

28、conductor assembly and test(OSAT)companiesFoundriesMaterial and subsystem companiesSemiconductor capital equipment companiesElectronic design automation(EDA)companiesAerospace and defense AutomotiveConsumer electronicsEnergyFinancial servicesHigh techIndustrial equipmentMedical devices/medical elect

29、ronicsRetailTelecom8Capgemini Research Institute 2025The semiconductor industry in the AI eraDefinitions Neural processing units(NPUs):1 NPU architecture simulates the neural network of the human brain.It processes large amounts of data simultaneously,performing trillions of operations per second.It

30、 uses less power and is far more efficient than a CPU or GPU,while freeing these up for other tasks.Combining an NPU with machine learning(ML)offers lightning-fast,high-bandwidth AI in real time.Data processing units(DPUs):2 A DPU is a“system-on-a-chip”(SoC)that combines:An industry-standard,high-pe

31、rformance,software-programmable,multi-core CPU,tightly coupled with the other SoC components,A high-performance network interface capable of parsing,processing,and efficiently transferring data at line rate,or at least network speed,to GPUs and CPUs,A rich set of flexible and programmable accelerati

32、on engines that offload and improve applications performance for AI and ML,zero-trust security,telecommunications and storage,among others.Graphics processing units(GPUs):3 A GPU is an electronic circuit that can perform mathematical calculations at high speed.Computing tasks such as graphics render

33、ing,ML,and video editing require similar operations on a large dataset.GPUs can perform the same operation on multiple data values simultaneously.This increases processing efficiency for many compute-intensive tasks.AI chips:4 AI chips are specialized computing hardware used in the development and d

34、eployment of AI systems.AI chips are essential for meeting the demand for greater processing power,speed and efficiency.Custom chips:5 Custom or application-specific integrated circuit(ASIC)chips are designed for a specific purpose,such as image processing or inference,at a lower monetary and resour

35、ce cost than a general-purpose processor.ASICs enable ML and other typically high-cost functionality in situations where it would otherwise be impractical.ASICs will not always be the appropriate solution,but are worth consideration.Softwarization:6 Softwarization is the concept of developing a more

36、 standardized,limited set of base chips that can be customized for various industries and solutions by reducing the number of unique hardware designs and instead using software to provide industry-specific functionalities essentially moving the“logic”from silicon to software.Downstream industries:Wh

37、ile nearly all industries rely on semiconductors for their products or services and operations,the scope of this research includes automotive,consumer electronics,retail,telecom,aerospace and defense,high tech,medical devices/medical electronics,industrial equipment,financial services,and energy.9Ca

38、pgemini Research Institute 2025The semiconductor industry in the AI era10Capgemini Research Institute 2024Organizations anticipate surging semiconductor demand 0110Capgemini Research Institute 2025The semiconductor industry in the AI era29%expected increase in demand for semiconductors by downstream

39、 organizations,to the end of 202611Capgemini Research Institute 2024Semiconductors are the backbone of the digital world,powering smartphones,Gen AI,computers,cars,satellites,and virtually every electronic device in use today.In 2023,nearly 1 trillion semiconductors more than 100 times the number of

40、 people on the planet were sold worldwide.7 Despite a cyclical market decline in the first half of 2023,worldwide sales recovered in the second half of the year to$527 billion.8 By 2030,it is anticipated that the semiconductor market size will surpass$1 trillion.9 Semiconductor manufacturers and dow

41、nstream industries expect demand to riseThe semiconductor industry and downstream industries anticipate significant growth in demand over the next two years.The main drivers of this increase are the growing usage of sophisticated electronics,the rise in electric and driverless cars,and the developme

42、nt of smart technologies and high-speed internet.Growth in data-driven applications and demand for energy-efficient solutions are also helping to perpetuate this trend.While the semiconductor industry is cyclical by nature,our research indicates that the semiconductor industry anticipates demand to

43、increase by 15%by 2026,while downstream organizations expect an increase of almost 30%(see Figure 1).While we did not ask the respondents for their projections beyond 2026,extrapolating from the current increase of 15%suggests that the market size could reach approximately$930 billion by 2030,nearin

44、g the$1 trillion estimate mentioned earlier.The semiconductor industrys cautious outlook reflects a slow recovery in markets like automotive despite strong AI-driven demand.11Capgemini Research Institute 2025The semiconductor industry in the AI eraExpected semiconductor demand increase in two years

45、to the end of 2026Semiconductor industryDownstream industries15%29%Figure 1.Downstream industries estimate demand for semiconductors to increase at double the rate of the semiconductor industrys expectationSource:Capgemini Research Institute,Semiconductor survey,November 2024,N=190 semiconductor ind

46、ustry organizations(includes Integrated Device Manufacturers(IDMs),fabless design firms and foundries),N=800 downstream organizations.Gen AI at work:Shaping the future of organizations12Capgemini Research Institute 2025The semiconductor industry in the AI era12The semiconductor industry in the AI er

47、aRapid technological progress is driving demand for more powerful,efficient,and customized chips,prompting semiconductor manufacturers to invest heavily in R&D.Our research indicates that 58%of semiconductor organizations believe that 5G or other next-gen communication protocols are impacting their

48、strategies,while 56%believe that Gen AI is a strong influence.5G and other next-generation communication protocols are foundational to enabling a wide range of emerging technologies and markets,including IoT,autonomous vehicles,AR/VR,and edge computing.They drive demand for advanced semiconductor so

49、lutions that require higher performance,energy efficiency,and integration.Figure 2.Nearly three in five semiconductor organizations say that 5G(or other next-gen communication protocols)and Gen AI are impacting their strategySource:Capgemini Research Institute,Semiconductor survey,November 2024,N=25

50、0 semiconductor organizations.Technology domains impacting semiconductor manufacturing strategy5G or other next-gencommunicationprotocolsEdge computingGenerative AIWearablesQuantumcomputingSpacetechnologyAR/VRBlockchain andcryptocurrency56%51%44%42%21%17%16%58%13Capgemini Research Institute 2025The

51、semiconductor industry in the AI eraThe adoption of AI/Gen AI is driving GPU/NPU demand The adoption of AI and Gen AI is driving demand for model training and inference capabilities,and data centers,while the growth of on-device AI applications further underscores the need for specialized semiconduc

52、tor solutions.Subi Kengeri,VP of AI Systems Solutions at Applied Materials,says,The AI era marks a new wave of growth for the semiconductor industry,propelled by the high returns on investment generated by AIs economic value.For AI systems,the key metric remains Total Cost of Ownership,while for Sil

53、icon,it is Perf/Watt/$.”To supply the demand rising from AI/Gen AI,organizations are ramping up their NPU,GPU,and memory capabilities.A Senior Director at a US-based IDM says,“We focus on optimizing the interplay of compute,memory,and network components to achieve system-level efficiency as customer

54、s adopt AI to unlock its benefits.Computational efficiency minimizes the time required for matrix multiplication,the memory subsystem ensures data is readily available for the compute engine,and the network subsystem enhances throughput,whether in small or large clusters.”Qualcomms NPU,designed spec

55、ifically to take on AI workloads,is an essential enabler of on-device Gen AI capabilities.The NPU offers optimal performance,power,and space efficiency to handle complex ML operations.10 ChatGPT exemplifies the transformative role of GPUs in AI.Leveraging thousands of NVIDIA GPUs,the training and in

56、ference processes for its large language model(LLM)demonstrate the unparalleled efficiency and scalability GPUs bring to AI workloads.This infrastructure supports Gen AI services for over 100 million users,underscoring the critical contribution of GPU technology to cutting-edge semiconductor applica

57、tions in AI-driven innovation.11 A Senior Director at a US-based IDM explains,“One of the key advancements in the semiconductor industry is performance enhancement through parallel computing,exemplified by GPUs.As AI accelerators evolve,various parallelization techniques,including expert parallelism

58、,pipeline parallelism,tensor parallelism,and context parallelism,are being employed.These optimization and architectural innovations are designed to boost overall throughput,whether for training or inference workloads.”Furthermore,todays Gen AI models require more data to enhance outcomes and exploi

59、t new opportunities.Inferencing LLMs such as ChatGPT benefits from advanced memory solutions such as Microns HBM3E(one of many products on the market),which can optimize performance and reduce CPU offload during AI processing,allowing faster training and greater responsiveness to queries.12 In our r

60、esearch,58%of semiconductor organizations mentioned they expect higher demand for NPUs to accompany growth in Gen AI adoption,with 57%anticipating increased need for high-performance chips and 56%for memory-intensive chips,signaling a shift toward advanced processing solutions.14Capgemini Research I

61、nstitute 2025The semiconductor industry in the AI eraFigure 3.Due to Gen AI adoption,nearly three in five semiconductor organizations are seeing increased demand for NPUs,high-performance GPUs,and memory-intensive chipsAreas where organizations anticipate demand for their semiconductor products will

62、 be impacted in the next two years by use of Gen AI applicationsNPUsHigh performanceGPUsMemory-intensivechipsCustom chipsDPUsPower and MEMSchips57%56%40%39%16%58%Source:Capgemini Research Institute,Semiconductor survey,November 2024,N=250 semiconductor organizations.15Capgemini Research Institute 20

63、25The semiconductor industry in the AI eraOrganizations project high demand for AI chips and custom silicon chips The market is also witnessing a surge in diverse applications for AI chips and custom-designed chips,underlining their transformative potential.Intel and AWS announced a multi-year co-in

64、vestment in custom chip design that would encompass Intel products and wafers.13 Sam Geha,EVP of IoT,Compute and Wireless business at Infineon Technologies,explains,“Just a few years ago,our role was simply to produce chips,leaving it to customers to determine how to use them.Today,however,we are ex

65、pected to deliver customized solutions.Software has emerged as a critical differentiator,particularly as our chips have become increasingly complex.Beyond general-purpose software,we now provide specialized solutions for AI and edge AI,enabling customers to effectively train and deploy models.Altern

66、atively,we can offer services to manage the training and deployment for them.”Our research shows that most downstream organizations foresee heightened demand in the next two years:88%anticipate a rise in AI chip needs,81%foresee heightened calls for custom chips,and 79%predict an upturn in demand fo

67、r memory-intensive chips.As figure 3 shows,according to 39%of semiconductor organizations,Gen AI is expected to drive demand for custom chips in the next two years.16Capgemini Research Institute 2025The semiconductor industry in the AI eraFigure 4.Nearly four out of five downstream organizations ant

68、icipate increased demand for AI chips,custom silicon chips,and memory-intensive chips over the next 12 monthsExpected demand for chips88%15%21%14%81%79%AI chips/chips designedfor AI accelerationCustom silicon chips(custom ASIC)Memory-intensive chips(e.g.,HBM,GDDR6)Proportion of organizations expecti

69、ngan increase in demandAverage expected increase overthe next 12 monthsSource:Capgemini Research Institute,Semiconductor survey,November 2024,N=800 downstream organizations.Subi KengeriVP of AI Systems Solutions,Applied MaterialsThe AI era marks a new wave of growth for the semiconductor industry,pr

70、opelled by the high returns on investment generated by AIs economic value.For AI systems,the key metric remains Total Cost of Ownership,while for Silicon,it is Perf/Watt/$.”17Capgemini Research Institute 2025The semiconductor industry in the AI eraRise of custom silicon As AI workloads ramp up energ

71、y demand,electricity and infrastructure suppliers must adapt quickly to support the growing needs of data centers.Meanwhile,companies such as NVIDIA with their commanding market share in AI chips,continue to exerts significant influence over pricing.The need for optimized performance,energy efficien

72、cy,and differentiation in competitive markets such as automotive,high-tech,and medical devices is driving the use of custom silicon.Custom chips allow organizations to better meet specific application demands,reduce costs at scale,and leverage advances in AI and IoT,fueling widespread adoption acros

73、s industries.According to our research,56%of downstream organizations are already using custom chips,while 11%are considering the possibility.A large majority of industrial equipment organizations(85%),and telecom organizations(82%)currently rely on custom silicon chips.Tech giants Microsoft,Amazon,

74、and Meta are also developing in-house chips tailored for AI inferencing.Figure 5.Two in three downstream organizations are either already using custom silicon chips or are considering using them in their productsSource:Capgemini Research Institute,Semiconductor survey,November 2024,N=800 downstream

75、organizations.Usage of custom silicon chips in products56%18%11%16%Yes,we use custom silicon chips in our productsYes,we are considering using custom silicon chips in our productsNo,we may consider using custom silicon chips in the futureNo,we have no such plans18Capgemini Research Institute 2025The

76、 semiconductor industry in the AI era19Jiani Zhang EVP,Chief Software Officer,Capgemini Engineering“Advanced platforms and software are no longer just enablers but critical differentiators in the semiconductor industry,driving efficiency and scalability in design,manufacturing,and deployment.With th

77、e growing complexity of AI,IoT,and edge computing applications,the ability to integrate domain-specific software with hardware accelerators will define leadership.To stay competitive,semiconductor players must embrace co-optimization across the stack,from chip architecture to application interfaces,

78、ensuring they can meet the escalating demands of data-intensive,low-latency markets.”19Capgemini Research Institute 2025The semiconductor industry in the AI era20Capgemini Research Institute 2024Downstream industries express concerns over semiconductor supply 0220Capgemini Research Institute 2025The

79、 semiconductor industry in the AI eraOver half of all downstream organizations are uncertain that the semiconductor industry can cope in 2025As nations compete for control over vital technologies and resources,geopolitical tensions continue to impact the global semiconductor supply chain.The flow of

80、 components,materials,and completed semiconductor products has been hindered by international trade disputes,export restrictions,and tariffs.For example,Taiwans TSMC,the worlds largest semiconductor manufacturer,with a market share of about 55%,is the producer of the worlds most advanced chips.14 Co

81、nsumer supply networks that rely on TSMC could be seriously disrupted by any military escalation involving China and Taiwan.Deteriorating US-China ties have also given rise to setbacks in the form of prohibitions on certain products and more stringent controls.In 2022,the US introduced export contro

82、ls that restrict the Peoples Republic of Chinas(PRCs)ability to obtain advanced computing chips,develop and maintain supercomputers,and manufacture The The Figure 6.More than three out of five organizations believe that geopolitical tensions and inadequate fab capacity impact the reliability of the

83、semiconductor supply chainSource:Capgemini Research Institute,Semiconductor survey,November 2024,N=800 downstream organizations.Factors impacting the reliability of semiconductor supply chainGeopolitical tensionsFab capacityReliance onsmall number ofsemiconductorsuppliersPush for sovereigntyPandemic

84、Availability ofnatural resources65%52%49%46%43%69%21Capgemini Research Institute 2025The semiconductor industry in the AI era69%Percentage of downstream organizations that believe geopolitical tensions impact the reliability of the semiconductor supply chainThe COVID-19 pandemic exposed vulnerabilit

85、ies in the global semiconductor supply chain,disrupting logistics,demand,and production,while heightening concerns over product availability and rising costs.Our research shows that 49%of downstream organizations consider this impact to be ongoing,while 47%had to curtail some product/feature launche

86、s due to chip shortages during the pandemic.Downstream industries share this uncertainty.Around 59%of downstream organizations believe that suppliers ability to meet their semiconductor demand is an ongoing concern.Similarly,only around one-quarter(26%)feel that supply is sufficient.This is particul

87、arly prominent among sectors such as A&D(14%)and organizations headquartered in Sweden(10%)and the United States(11%).advanced semiconductors.15 These rules were revised in 2023,and in December 2024,the US Commerce Department further expanded the list of Chinese technology companies subject to expor

88、t controls and included many that make equipment used to make computer chips,chipmaking tools,and software.China,as a response,announced that it is banning exports to the United States of gallium,germanium,antimony and other key high-tech materials with potential military applications.16 Our researc

89、h indicates that 69%of downstream organizations believe that geopolitical tensions significantly impact the reliability of the semiconductor supply chain.Additionally,65%of downstream organizations consider fab capacity to have a strong impact,while 52%feel that reliance on a small number of semicon

90、ductor suppliers impacts their reliance on the semiconductor supply chain.22Capgemini Research Institute 2025The semiconductor industry in the AI eraFigure 7.Fewer than three in ten downstream organizations believe chip supply is sufficientSource:Capgemini Research Institute,Semiconductor survey,Nov

91、ember 2024,N=800 downstream organizations.GPU computing and AI/ML acceleration are the most relevant advancements for downstream organizationsSemiconductor technology breakthroughs have spurred innovation in consumer industries,breeding smarter,more efficient products.AI/ML acceleration and GPU proc

92、essing have the potential to revolutionize downstream operations.GPUs outperform because they provide high throughput and parallel processing,streamlining real-time inference and model training,in particular for AI and ML applications.Our research suggests that 54%of downstream organizations(those t

93、hat rely on fast data processing and AI-powered automation)believe that GPU computing and AI/ML accelerations are the most relevant advancements for them.Alessandro Miranda,Senior Director of Radio Access Network(RAN)Design and Optimization,at ZTE,explains,“We need specialized hardware and architect

94、ures designed to accelerate processing and optimize algorithms.Graphics Downstream industries perception of chip supply/demand59%47%26%Suppliers ability to meetour semiconductor demandis an ongoing concernfor our organizationDuring the COVID-19pandemic,we had tocurtail some product/featurelaunches d

95、ue to chip shortages The semiconductorindustry is supplying chipsat a rate sufficientfor our needs23Capgemini Research Institute 2025The semiconductor industry in the AI eraprocessing units(GPUs),for instance,which were originally developed solely for rendering images and videos in video games,have

96、now become the cornerstone of AI processing due to their ability to handle parallel data processing efficiently.“Dell Technologies showcases the practical application of GPU computing and AI/ML acceleration by integrating NVIDIAs AI-ready GPUs,networking solutions,and tools like AI Enterprise and Om

97、niverse with its own hardware and expertise.This collaboration offers communications service providers(CSPs)the tools needed to efficiently run AI workloads across networks.17Figure 8.More than half of downstream organizations believe that advancements in GPU computing and AI/ML acceleration can bri

98、ng most valueSource:Capgemini Research Institute,Semiconductor survey,November 2024,N=800 downstream organizations.54%54%49%47%27%24%24%19%17%GPU computingNPU computingAI and ML acceleration5G/next-gen communication technologiesAdvancements in GHz/wattDPUChipletsWearablesNext-generation memory chips

99、(MRAM,ReRAM)Most relevant semiconductor advancements for downstream industries24Capgemini Research Institute 2025The semiconductor industry in the AI eraAlessandro Miranda Senior Director of Radio Access Network(RAN)Design and Optimization,ZTE“We need specialized hardware and architectures designed

100、to accelerate processing and optimize algorithms.Graphics processing units(GPUs),for instance,which were originally developed solely for rendering images and videos in video games,have now become the cornerstone of AI processing due to their ability to handle parallel data processing efficiently.”25

101、25Capgemini Research Institute 2025The semiconductor industry in the AI eraDownstream organizations expect enhanced customization,more comprehensive APIs and SDKs,and stronger securityThe semiconductor industry has seen significant enhancements in recent years,particularly in areas such as customiza

102、tion,security,APIs,software development kits(SDKs),kits,and boards.These advancements are helping businesses develop more tailored,secure,and efficient products.Our research indicates that 47%of downstream organizations have ranked enhanced customization and availability of more comprehensive APIs,S

103、DKs,kits,boards,etc.,in the top three innovations that they look forward to in the semiconductor industry,while 41%have ranked enhanced security among the top three.Companies such as Intel18 design custom ASICs for their CPUs and GPUs,integrating both digital logic and analog components to meet the

104、specific performance and power requirements of their products.Figure 9.Nearly half of downstream organizations are looking for enhanced customization and more comprehensive APIs and SDKsSource:Capgemini Research Institute,Semiconductor survey,November 2024,N=800 downstream organizations.47%47%41%36%

105、29%27%25%25%21%Enhanced customizationAI supportAvailability of more comprehensive APIs,SDKs,kits,boards,etc.Enhanced securityHigher computation capacityLower power consumptionSmaller sizeImproved integrationDrive customer experienceTop ranked technology innovations/improvements desired by downstream

106、 organizations26Capgemini Research Institute 2025The semiconductor industry in the AI eraOne in three downstream organizations is designing chips in-house(or currently exploring the option)In-house chip design allows organizations to tailor semiconductors to their unique specifications,minimizing re

107、liance on external vendors and preserving control over their intellectual property(IP),while boosting speed,power efficiency,and compatibility with other hardware and software.Because tech giants use SoCs and custom processors to differentiate their goods,this strategy has become more popular.As par

108、t of a broad initiative to integrate AI capabilities into its devices,Apple will implement several of its future AI features using data centers furnished with its in-house processors.19 Our research indicates that 34%of downstream organizations are either already designing chips in-house or are expl

109、oring options in this area.Figure 10.More than three out of ten downstream organizations are either already designing their chips in-house or are considering itSource:Capgemini Research Institute,Semiconductor survey,November 2024,N=800 downstream organizations.Has your organization considered/been

110、considering designing your own semiconductor chips?10%44%23%24%Yes,we have already started designing our chips in-houseYes,we are currently exploring designing chips in-houseNo,we have considered it and decided against itNo,we have no such plans27Capgemini Research Institute 2025The semiconductor in

111、dustry in the AI eraHowever,developing a complex design cycle for in-house chips requires a significant investment in talent,cutting-edge design tools,and testing equipment.Supply chain issues,including securing raw materials and managing production,further complicate these choices.Our research sugg

112、ests that 23%of downstream organizations have decided against in-house chip design,while 44%have not considered it.28Capgemini Research Institute 2025The semiconductor industry in the AI eraFigure 11.More than three out of five downstream organizations consider supply chain complexity,rapid technolo

113、gical advancements and high capital expenditure as primary reasons to decide against in-house chip designSource:Capgemini Research Institute,Semiconductor survey,November 2024,N=529 downstream organizations.In todays feverishly competitive tech market,the opportunity to control the hardware ecosyste

114、m is a strong selling point.Although semiconductor design organizations can fall back on decades of experience and IP,they should avoid complacency in a changing industry.The semiconductor and information technology industries are likely to continue to develop in tandem as a result of Meta,Microsoft

115、,Apple,Google,Amazon,and other significant tech companies switching to in-house semiconductor chip design in order to reduce supply chain risk,capitalize on in-house innovation,and keep costs low.Over the next two years,sustainability,supply chain resilience,and security will be customers top priori

116、tiesEnsuring a sustainable,flexible,and resilient supply chain involves adopting eco-friendly practices.Volkswagen,20 for instance,reorganized its procurement with Cross Operational Management Parts&Supply Security(COMPASS)to address Reasons for not considering in-house designComplex supply chainRap

117、id technologicaladvancements inthe semiconductor industryHigh capital expenditureInaccessibility toadvanced process technologiesEconomies of scaleTalent gap65%64%57%54%39%67%29Capgemini Research Institute 2025The semiconductor industry in the AI erathe daily shortage of semiconductors and questionab

118、le supply chain resilience.Security is equally important,as the semiconductor industry is a prime target for cyber threats and IP theft.Organizations must implement stringent security measures to safeguard the integrity of the supply chain and boost customer trust.Together,these elements form the ba

119、ckbone of a reliable,future-proof semiconductor industry.Our research shows that 58%of downstream organizations cite sustainability and supply chain resilience as top priorities,while 51%will prioritize cybersecurity for the coming two years.Figure 12.More than half of downstream organizations plan

120、to prioritize chip sustainability,supply chain resilience,and cybersecurity features in the next two yearsSource:Capgemini Research Institute,Semiconductor survey,November 2024,N=800 downstream organizations.58%58%51%48%46%44%41%38%Sustainability of chips is going to be a crucial aspect in our chip

121、selection.Government policies and regulations will significantly influence our semiconductor sourcing strategies.Supply chain resilience will be a critical factor in our selection of semiconductor suppliers.We will prioritize semiconductor chips that offer advanced cybersecurity features to protect

122、our products and services.Our Gen AI adoption is dependent on how well our semiconductor suppliers are able to meet the demand.We are confident that the semiconductor industry will continue to innovate at a pace that meets our future needs.We prefer to use marketplaces such as the OCP Marketplace to

123、 understand the emerging trends in chips.Customization of semiconductor chips to meet our specific needs will be a key priority.Do you agree with the below statements for your organization over the next two years?30Capgemini Research Institute 2025The semiconductor industry in the AI era31Capgemini

124、Research Institute 2025The semiconductor industry in the AI eraCapgemini Research Institute 2024The semiconductor industry is innovating,but softwarization remains a challenge03The semiconductor industry in the AI era32Capgemini Research Institute 2025The semiconductor industry in the AI eraFigure 1

125、3 highlights the semiconductor industrys design and manufacturing innovation,with nearly two in three organizations rating their abilities in these areas as high.Similarly,hardware security remains a strong area,with nearly two in three excelling in it.Figure 13.While the semiconductor industry exce

126、ls in innovation and security,softwarization poses a challengeSource:Capgemini Research Institute,Semiconductor survey,November 2024,N=250 semiconductor organizations.Organizations that rated their abilities high in the below areasManufacturing innovationHardware securityDesign innovationSoftwarizat

127、ion ofsemiconductors63%62%47%63%Gregg BromleyDirector,AMD“Chiplets offer a way to address the complexities stemming from Moores law.They enable the integration of diverse technologies,which is particularly beneficial for AI applications that demand increased computational power.“33Capgemini Research

128、 Institute 2025The semiconductor industry in the AI eraThe semiconductor industry continues to excel in design and manufacturing innovationDesign innovation The semiconductor industry is turning to advanced architectures and methodologies to meet demand from downstream organizations for increasingly

129、 powerful components.Several organizations already rely on the flexibility of chiplet technology.Having already implemented multi-chiplet architecture for its CPUs and data center GPUs,AMD has filed a patent for a GPU design that is split across multiple GPU chiplet sets.This modular GPU design enab

130、les scalable adjustment of GPU resources and performance to match different operating modes.21 Our research shows that the development of 3D IC techniques and multi-die integration is one of the top three areas of focus for semiconductor organizations.The industry is also exploring open-source archi

131、tectures such as RISC-V,enabling cost-effective customization.(Designing for testing and manufacturability nevertheless remains the top focus area.)Figure 14.Designing for testing and manufacturability remains the top focus area for semiconductor design Source:Capgemini Research Institute,Semiconduc

132、tor survey,November 2024,N=250 semiconductor organizations.69%64%58%55%31%23%Implementation of design-for-test(DFT)and design-for-manufacturability(DFM)techniquesEnhance extensibility/reusability of current IP into new productsDevelopment of 3D IC design techniques and multi-die integrationExplorati

133、on of new architectures(e.g.,RISC-V,chiplets,heterogeneous integration)Integration of photonics and optoelectronics in semiconductor designCutting-edge technologies(such as quantum computing,neuromorphic chips)Please select the top three areas where your organization intends to invest to enhance des

134、ign innovation/engineering in the next two years34Capgemini Research Institute 2025The semiconductor industry in the AI eraIn addition to AI and ML,around half of semiconductor design organizations use Gen AI to shorten their design cycles.Thy Phan,a Senior Director at Synopsys says,EDA tools are le

135、veraging AI to enhance PPA(performance,power,area)and development time by automating iterative design processes.These tools can concurrently optimize performance,power,and area,which are key factors of chip design,to find the best balance among them so that the specifications can be met or exceeded.

136、NVIDIAs engineers have also created a custom LLM for internal use,which is trained on internal data to generate and optimize software and assist human designers.The LLM serves multiple purposes,including generating code snippets in two design languages and automating bug monitoring.22 Bill Dally,NVI

137、DIA s Chief Scientist,says:“This marks an important first step in applying LLMs to the complex work of designing semiconductors.It shows how even highly specialized fields can use their internal data to train useful generative AI models.“Our research further revealed that the industry is expecting i

138、ts R&D budget to increase by around 10%over the next two years.In 2023 alone,Intel spent$16.5 billion on R&D,followed by Qualcomm($8.9 billion),NVIDIA($7.3 billion),and AMD($5.9 billion).23Figure 15.One in two semiconductor organizations is investing in Gen AI for designSource:Capgemini Research Ins

139、titute,Semiconductor survey,November 2024,N=167 IDMs,fabless design firms and EDA firms.Is your organization investing in the design innovations below?50%78%Adopting design automation technologiesto improve performance of the chipsInvesting in Gen AI to shorten our design cycles35Capgemini Research

140、Institute 2025The semiconductor industry in the AI eraManufacturing innovation To keep up with Moores law,semiconductor manufacturers continue to push the boundaries of physics.Extreme ultraviolet(EUV)lithography,for instance,employs light with shorter wavelengths(13.5 nm)to create intricate pattern

141、s on silicon wafers.This allows the production of chips with transistors as small as 3 nm or 5 nm,meaning manufacturers can pack more transistors onto a single chip.With ASML being the sole supplier of these EUV machines,24 and each machine costing$150 million,25 semiconductor manufacturers face a b

142、ottleneck in their supply chains.Our research shows that cost improvement is the top area of focus for 60%of the industry,while 57%are turning to materials research for better yield.Antimonide-and bismuthide-based semiconductors are employed in advanced infrared sensors for medical and military appl

143、ications.Similarly,pyrite,which is abundant and non-toxic,can replace the rare earth element cadmium telluride,currently widely used in solar cells.26 Subi Kengeri from Applied Materials,says,“Organic substances tend to have mechanical reliability and yield challenges beyond a certain substrate size

144、.To surpass a specific size,such as an 80 x 80 mm monolithic substrate,alternative materials are needed for packaging.This is why the industry is turning to glass substrates,which offers more rigidity.”Figure 16.Lowering costs and materials research are the top focus areas for semiconductor manufact

145、uringSource:Capgemini Research Institute,Semiconductor survey,November 2024,N=250 semiconductor organizations.60%57%49%49%45%40%31%23%Manufacturing innovation focused on cost improvementDigital continuity driving better connectedness between manufacturing and peripheral systems for better productivi

146、ty,efficiency and planning effectivenessMaterials research for better yield and/or productivityImproving traceability of materials being issued to manufacturing to the sourceSmart manufacturing-driving better insights in manufacturing to improve outputMaterials research for sustainabilityPlease sele

147、ct the top three areas where your organization intends to invest in the next two years to enhance manufacturing36Capgemini Research Institute 2025The semiconductor industry in the AI eraOur research has found that more than half of semiconductor manufacturers use techniques such as real-time data sh

148、aring across upstream and downstream systems to improve agility and enhance quality control and reliability testing.TSMC has integrated intelligent mobile devices,Internet of Things(IoT)technologies,augmented(AR)and mixed reality(MR)applications,and mobile robots into its intelligent automated mater

149、ial handling systems(AMHS).This integration consolidates wafer data collection and analysis and efficiently utilizes resources,maximizing manufacturing effectiveness.27 Figure 17.Nearly half of all semiconductor manufacturers rely on AI and ML Source:Capgemini Research Institute,Semiconductor survey

150、,November 2024,N=182 integrated device manufacturers(IDMs),foundries,outsourced semiconductor assembly and test(OSAT)firms and semiconductor capital equipment firms.55%We connect manufacturing data on a real-time or near real-time basis to upstream and downstream planning and execution systems.Adopt

151、ing enhanced quality control and reliability testing methods for ensuring the performance and longevity of semiconductor devices.Using advanced AI and ML algorithms for process optimization and predictive maintenance.Please indicate whether your organization uses the manufacturing innovations below:

152、51%49%We connect manufacturing data on a real-time or near real-time basis to upstream and downstream planning and execution systems.Adopting enhanced quality control and reliability testing methods for ensuring the performance and longevity of semiconductor devices.Using advanced AI and ML algorith

153、ms for process optimization and predictive maintenance.37Capgemini Research Institute 2025The semiconductor industry in the AI eraPackaging innovation TSMCs Chip-on-Wafer-on-Substrate(CoWoS)packaging technology allows 2.5D and 3D stacking of components,streamlining efficiency.28 TSMC is doubling its

154、 CoWoS capacity for 2025,with NVIDIA set to receive 50%of this expanded supply,although the Taiwanese manufacturer still anticipates that AI demand will continue to outpace supply capacity.29 Around 40%of organizations in our research are implementing 3D-packaging techniques,while one in three focus

155、es on chiplets.Omer Dossani,VP of Global Test Services,at Amkor Technologies,says,“Silicon continues to grow increasingly complex.Packaging has become more critical than ever as we strive to keep pace with Moores law.The challenge lies in managing high power demands while addressing space constraint

156、s simultaneously.”Gregg Bromley,a Director at AMD,adds,“Chiplets offer a way to address the complexities stemming from Moores law.They enable the integration of diverse technologies,which is particularly beneficial for AI applications that demand increased computational power.“Figure 18.Around two i

157、n five organizations deploy 3D packaging for semiconductorsSource:Capgemini Research Institute,Semiconductor survey,November 2024,N=149 IDMs and OSAT firms for the top statement and N=197 IDMs,fabless design firms,OSAT firms and EDA firms for the bottom statement.41%Implementing 3D packaging techniq

158、uesAdopting chiplet architecture to enhance semiconductor performancePlease indicate whether your organization uses the packaging innovations below:34%38Capgemini Research Institute 2025The semiconductor industry in the AI eraHardware security remains a critical area of focusThe semiconductor indust

159、rys highly complex and interdependent supply chain spans suppliers of IP,equipment,and materials.This means that any vulnerability can jeopardize the entire ecosystem that depends upon semiconductor supply.A US-based semiconductor firm attributed a potential loss of$250 million in sales in 2023 to a

160、 cybersecurity incident at one of its suppliers.30 The industry also faces potential data breaches,IP theft,physical security issues,and attacks on the manufacturing process.In another notable incident,researchers identified a vulnerability in a series of chips from a prominent company that allowed

161、attackers to extract secret encryption keys during cryptographic operations(known as a“side-channel attack”).The vulnerability stems from the microarchitectural design of the silicon;to mitigate it would require a significant trade-off in performance.31Our research has shown that three in five semic

162、onductor design organizations prioritize secure firmware and software integration to prevent vulnerabilities.Additionally,58%also use cryptographic protection,such as hardware-based encryption,to safeguard the data within the chip.Sam Geha from Infineon comments,“We are observing significant industr

163、y trends around security,with customers increasingly requesting features like secure boot and secure firmware over-the-air updates.There is a growing need for enhanced security across various applications,and we believe this trend will only continue to strengthen.“60%Percentage of semiconductor orga

164、nizations focusing on secure firmware and software integration when designing hardware-based chip security39Capgemini Research Institute 2025The semiconductor industry in the AI eraSam GehaEVP of IoT,Compute and Wireless business,Infineon Technologies“We are observing significant industry trends aro

165、und security,with customers increasingly requesting features like secure boot and secure firmware over-the-air updates.There is a growing need for enhanced security across various applications,and we believe this trend will only continue to strengthen.40Capgemini Research Institute 2025The semicondu

166、ctor industry in the AI eraFigure 19.Nearly three in five semiconductor design organizations focus on cryptographic protection to enhance chip security Source:Capgemini Research Institute,Semiconductor survey,November 2024,N=167 IDMs,fabless design firms and EDA firms.Steady progress in softwarizati

167、on,but monetization challenges persistSoftwarization allows for the creation and updating of custom software for different industries without needing to redevelop hardware.This leads to agile adaptation,extended product lifecycles,and tailored solutions across various applications,with significant c

168、ost and efficiency benefits.Our research found that nearly half of all semiconductor design firms are actively developing software-centric solutions,while another quarter are exploring or piloting the integration of software.For instance,NVIDIAs CUDA Compute Unified Device Architecture platform supp

169、orts parallel processing,using GPU acceleration to develop and optimize high-performance applications.32 Similarly,Qualcomm is working with AWS to enable downstream automotive organizations to integrate cloud technology for simulating,testing,validating,and verifying new features and functions prior

170、 to traditional hardware testing,expediting the development process.33 60%58%54%42%41%34%Secure firmware and software integration(e.g.,firmware updates)Cryptographic protection(e.g.,hardware-based encryption)Authentication and access control(e.g.,hardware root of trust)Supply chain security(e.g.,sec

171、uring third-party components)Physical security mechanisms such as tamper-resistant packagingProtection against side-channel attacks(e.g.,electromagnetic leakage prevention)Which of the below security measures are considerations when designing hardware-based chip security?41Capgemini Research Institu

172、te 2025The semiconductor industry in the AI eraFigure 20.Nearly half of semiconductor organizations are actively developing software-centric solutionsSource:Capgemini Research Institute,Semiconductor survey,November 2024,N=165 IDMs and fabless design firms.Actively developing software-centric soluti

173、ons alongside hardwareExploring or piloting software integration in our productsPlanning to adopt softwarization in the near futureNo current involvement or plans related to softwarizationWhat is your companys current level of involvement in the softwarization of chips?8%17%26%48%Brett Bonthron,Indu

174、stry Leader for Tech and Digital from Capgemini,adds,“Softwarization is transforming the semiconductor industry,enabling unprecedented flexibility and scalability.By decoupling hardware from software,companies can rapidly adapt to evolving demands,optimize performance across diverse applications,and

175、 unlock new capabilities through updates.This shift not only accelerates innovation but also fosters collaboration across the ecosystem,as software becomes the bridge connecting hardware excellence with cutting-edge functionality.“42Capgemini Research Institute 2025The semiconductor industry in the

176、AI eraOur research found that two in three semiconductor organizations anticipate increasing their use of software in the next two years.However,around 60%of organizations find monetizing these software-based offerings a challenge,due either to customer expectations or IP complexities.Figure 21.Mone

177、tization of software remains a challenge for three in five semiconductor organizationsSource:Capgemini Research Institute,Semiconductor survey,November 2024,N=250 semiconductor organizations.66%We anticipate using more software to control our chips in the next two yearsWe are encountering difficulti

178、es in monetizing our software offerings due to customer expectations that software features should be included with the hardware at no additional costWe face challenges in establishing effective software monetization models because of complexities in intellectual property rights and licensing agreem

179、ents64%56%66%Percentage of semiconductor organizations anticipating the use of more software to control chips in the next two years43Capgemini Research Institute 2025The semiconductor industry in the AI eraOmer DossaniVP of Global Test Services,Amkor Technologies“Silicon continues to grow increasing

180、ly complex.Packaging has become more critical than ever as we strive to keep pace with Moores law.The challenge lies in managing high power demands while addressing space constraints simultaneously.”44Capgemini Research Institute 2024The semiconductor industry in the AI era45Capgemini Research Insti

181、tute 2024The semiconductor industry in the AI era46Capgemini Research Institute 2024Resilience and sustainability gather momentum0446Capgemini Research Institute 2025The semiconductor industry in the AI era17%Semiconducor industrys anticipated improvement in domestic sourcing over the next two years

182、The global semiconductor chip shortage in 202023 wrought devastation in many other economic areas,underscoring the far-reaching significance of the industry.Factory shutdowns,US trade restrictions on Chinese chip imports,extreme weather and fires disrupting global production,and a neon shortage caus

183、ed by the Russian invasion of Ukraine all exacerbated supply chain issues.Conversely,the surge in cryptocurrency markets,stay-at-home orders,and the shift to remote work drove significant spikes in demand.Only two out of five organizations rate their resilience as high,while less than half(46%)rate

184、their sustainability achievements as high.Organizations turn to onshoring and friendshoring to enhance resilienceOur research found that the semiconductor industry expects to improve its domestic sourcing from 40%currently to 47%over the next two years to mitigate the risks associated with internati

185、onal logistics.The challenge is reliance on a limited set of suppliers or regions for certain components or pieces of equipment.ASML,for instance,is the only supplier of EUV lithography machines(see page 36).34 As of 2022,Ukraine was the supplier of 90%of the highly purified,semiconductor-grade neon

186、 for chip production in the US industry.35 On the manufacturing front,TSMC has a market share of above 60%in contract manufacturing and over 90%in advanced chip manufacturing.36 Similarly,Spruce Pine,a small town in North Carolina,US,is home to two mines that produce more than half of the worlds hig

187、h-purity quartz,and production at these mines was shut down for two weeks due to Hurricane Helene in 2024.3747Capgemini Research Institute 2025The semiconductor industry in the AI eraOur research also found that the industry anticipates a significant shift to nearshoring over the next two years(arou

188、nd one-quarter of sourcing and one-fifth of manufacturing capacity).Three-quarters of organizations expect to increase US investments,while 59%plan to increase EU investments.In 2022,the US allocated nearly$53 billion in loans,grants,and incentives for US semiconductor research,development,manufactu

189、ring,and workforce development via the CHIPS and Science Act,a US federal statute signed into law by President Joe Biden that is intended to strengthen domestic semiconductor production.38 The Act awarded$8.5 billion in grants and$11 billion in federal loans to Intel,and grants and federal loans to

190、firms including GlobalFoundries,Micron,TSMC,and Amkor Technology,among others.39 With up to$6.6 billion in direct funding allocated through the CHIPS Act,TSMC is investing more than$65 billion in three US greenfield fabs.This investment will support the development of chips for AI,high-performance c

191、omputing,and 5G/6G communications.One of these fabs is scheduled to manufacture chips using Figure 22.Semiconductor industry anticipates domestic sourcing to improve by 17%over the next two yearsSource:Capgemini Research Institute,Semiconductor survey,November 2024,N=202 IDMs,foundries,OSAT firms,ca

192、pital equipment firms,material and subsystem organizations.Sourcing capacity by location23%24%29%47%36%40%TodayNext two yearsDomestic/onshore Offshore Nearshore48Capgemini Research Institute 202448Capgemini Research Institute 2025The semiconductor industry in the AI era2 nm or even more advanced pro

193、cess technology.40 Since the introduction of the CHIPS Act,semiconductor organizations have announced more than 90 new US manufacturing projects,representing nearly$450 billion in investment capital.41 In a similar vein,the EU in April 2023 announced a 43 billion subsidy plan through the EU CHIPS Ac

194、t,with the objective of building large-scale capacity and innovation,ensuring the self-sufficiency of the EU,and galvanizing its resilience against future supply crises.42 Organizations are also deploying strategies to ensure production continuity.Four in five semiconductor organizations plan to mai

195、ntain strategic inventory levels,while three-quarters are prioritizing supplier diversification.Figure 23.Investment will focus on the US and the EU over the next two yearsSource:Capgemini Research Institute,Semiconductor survey,November 2024,N=250 semiconductor organizations.16%18%28%20%17%5%12%37%

196、28%18%22%26%48%22%4%37%34%6%ChinaUnited Kingdom EU United StatesInvestment likely to increase significantly Investment likely to increase slightly No change Investment likely to decrease slightly Investment likely to decrease significantly 49Capgemini Research Institute 2025The semiconductor industr

197、y in the AI era10%Expected increase of around 10%in semiconductor industrys R&D budget over the next two years60%81%Maintaining strategic inventory levels to ensure production continuityActive diversification of supplier baseInvesting in new solutions or advanced technologies(e.g.,data and analytics

198、,AI,IoT)Improving collaboration and communication with suppliers during periods of supply chain instabilityConducting regular supply chain risk assessments to identify potential vulnerabilities and develop preemptive mitigation strategiesWhich resilience strategies is your organization implementing

199、or planning to implement to combat supply chain disruptions?74%62%62%60%Figure 24.Three-quarters of semiconductor organizations are focusing on active diversification of their supplier basesSource:Capgemini Research Institute,Semiconductor survey,November 2024,N=250 semiconductor organizations.50Cap

200、gemini Research Institute 202450Capgemini Research Institute 2025The semiconductor industry in the AI eraThe industry is making strides with various sustainability initiativesFor a number of years,the semiconductor industry has continued its focus on improving power efficiency.Our research found tha

201、t the semiconductor industry is also focusing on energy conservation,water use reduction,hazardous chemicals management,and waste minimization.Alessandro Miranda from ZTE highlights,We are heavily investing in more energy-efficient manufacturing processes,striving to reduce waste and water usage eac

202、h year.This effort is a global initiative,implemented across all our manufacturing sites worldwide.Figure 25 shows the top three deployments across each of these categories.Intel,for instance,pledged to achieve net-zero greenhouse gas(GHG)emissions in its global operations by 2040.43 To achieve this

203、,the organization plans to invest around$300 million by 2030,targeting 4 billion cumulative kilowatt hours(KwH)of energy savings.Additionally,it will launch a cross-industry R&D initiative to identify greener chemicals and develop new abatement equipment.ASML is focusing on circularity and aspires t

204、o send zero waste from its operations to landfills or incineration.44 Chiplet-based modular design allows for interchangeable component configurations,boosting flexibility while cutting manufacturing costs and waste.By producing its fourth-generation EPYC CPUs with eight separate compute chiplets,ra

205、ther than a single monolithic die,AMD was able to save 50,000 metric tons of CO2e in 2023 by cutting wafer manufacturing roughly the same as its operational CO2e footprint in 2022.45 As chiplets can be stacked in a package,they allow the creation of energy-efficient processors as an alternative to d

206、ata centers.By upgrading its server processors to AMD EPYC,STMicroelectronics slashed its electricity consumption by 33%while increasing compute performance.51Capgemini Research Institute 2025The semiconductor industry in the AI eraUpgraded to energy-efficient machinery and equipmentImplemented ener

207、gy-management systemsReduced water usageEnergy conservationImplemented water recycling and reuse systemsModified processes to reduce water consumptionAchieved zero waste by reusing all wastewaterReducing water usage and promoting water circularityReplaced hazardous chemicals with less toxic alternat

208、ivesImplemented closed-loop systems for chemical reuseRecycled or reclaimed chemicals for reuse in processesManaging hazardous chemicals Implemented waste-minimization programsPartnered with organizations on waste-to-resource initiativesDesigned products and processes to reduce material useUtilized

209、end-of-life products as raw materials(closed-loop recycling)Waste reductionFigure 25.Top three initiatives currently being deployed in each categorySource:Capgemini Research Institute,Semiconductor survey,November 2024,N=182 IDMs,foundries,OSAT firms and semiconductor capital equipment firms.52Capge

210、mini Research Institute 202452Capgemini Research Institute 2025The semiconductor industry in the AI eraThy Phan Senior Director,Synopsys“EDA tools are leveraging AI to enhance PPA(performance,power,area)and development time by automating iterative design processes.These tools can concurrently optimi

211、ze performance,power,and area,which are key factors of chip design,to find the best balance among them so that the specifications can be met or exceeded.”53Capgemini Research Institute 2025The semiconductor industry in the AI eraExploring big techs shift to in-house chip designAlphabet As part of a

212、broader strategy to reduce reliance on external suppliers,Alphabet used AlphaChip technology to design its own TPU AI accelerators,which has been used in several initiatives:46 TPU AI accelerator has helped Google to drive large-scale AI models and cloud services,including Gemini and Imagen.This AI

213、accelerator has been used to improve the design of each successive processor chips,the latest one being Trillium 6th Generation chips being used extensively in data centers worldwide,which offers 4.7X enhanced computational performance while being 67%more energy efficient as compared to the previous

214、 version.47 TPU AI accelerator has also played a significant role in development of Dimensity 5G system-on-chips for MediaTek,which are used extensively in smartphones.AlphaChip AI has also been used in chip design recently,where reinforcement learning(RL)has been used to optimize chip layouts for c

215、omputers,cutting down design times and boosting performance.Googles custom-built Tensor G4 chips are used to power Google AI in Pixel smartphones,and enhance user experience.48 Google has integrated Tensor A1 chips into Pixel Buds Pro 2 earbuds to enhance its noise cancellation capabilities.49 In 20

216、24,Googles parent company,Alphabet,allocated a considerable portion of its$13 billion capital expenditure to expanding its technical infrastructure,including data centers and servers,to support production of its custom Tensor Processing Unit(TPU)AI chips.5054Capgemini Research Institute 2025The semi

217、conductor industry in the AI eraAmazon Amazon has launched its 4th generation general-purpose processor Graviton4,which offers 30%higher computational performance and 75%increase in memory bandwidth compared to its previous version.It is planned to be used in Amazon Elastic Compute Cloud(Amazon EC2)

218、R8g to manage databases,in-memory caches,and large-scale data analytics.51 The company has also launched Trainium 2 for AI training.It offers 4X faster model training speed,3X memory capacity and 2X performance per watt as compared to the previous version.52 Amazon has developed Inferentia chips,whi

219、ch power Amazon EC2 Inf1 instances for deep learning and generative AI applications.It offers 2.3X higher throughput at up to 70%lower cost per inference.53 Intel and AWS are seeking to expand their collaboration,including a co-investment in custom chip design,under a multi-year,multi-billion-dollar

220、 framework.54 These chips serve more than 50,000 customers,including Formula 1,Pinterest,and SAP.Graviton4 also enhances security and energy efficiency,playing a critical role in AWSs strategy to lower costs for its cloud clients.In 2024,Amazons planned capital expenditure rose sharply to$75 billion

221、,driven largely by Gen AI investments,and up from$48.4 billion in 2023.55 The organization expended$30.5 billion in CapEx in H1,principally on infrastructure upgrades to accommodate AI and ML demand.56Apple In November 2020,Apple announced Apple Silicon,in-house-designed chips for iPhone,iPads and A

222、pple watches.57 Apple launched the M1 chip in 2020,following with the M2 and,recently,the M3 chip,offering substantial improvements in performance and energy efficiency with each iteration.The M3 chip,used in devices such as the MacBook Air,marks a leap in chip design,integrating 3-nanometer technol

223、ogy for better computational power and efficiency.58 TSMC is producing Apples custom-designed chips,including the A16 and A17 series,at the formers new Arizona plant,as part of Apples broader strategy to localize its US semiconductor manufacturing.59 In 2023,Apple confirmed that Amkor would package

224、Apple silicon chips produced at a nearby TSMC facility,as part of a US manufacturing push.60Meta In order to avoid excessive reliance on external vendors,Meta is working on building Artemis chips in-house for its data centers customized to support the AI development programs,especially the inferenci

225、ng part.61 55Capgemini Research Institute 2025The semiconductor industry in the AI era Metas first-generation custom AI chip,MTIA Meta Training and Inference Accelerator is tailored for in-house AI workloads,and offers greater efficiency and computational power than other CPUs in the market.62 It is

226、 focused on reducing power consumption while handling large-scale AI workloads,using TSMCs 7-nm process.It is current working on second-generation MTIA to be deployed across its data centers,and will use high-bandwidth memory to boost performance.63 Meta has deployed 3D chip integration technology i

227、n AR devices,which increases computational power without increase size of the devices.64In 2024,Meta announced plans to spend$38 billion$40 billion on AI infrastructure,marking a significant increase from previous guidance.This investment is earmarked for enhancing its LLM capabilities,including tra

228、ining Llama 3,and readying its infrastructure for more advanced AI workloads.Metas chip-development efforts are central to its independence strategy.65Microsoft Microsoft has developed Azure Maia 100,Microsofts first in-house AI accelerator,to support its cloud-based AI workloads and optimize Azure

229、AI infrastructure.66 Microsoft added a rack-level,closed-loop liquid cooling system,which smooths integration of these Maia 100 systems into Microsofts existing data center infrastructure,without increasing their footprint.67 It has also built Arm-powered Azure Cobalt CPU,a 128-core chip,to support

230、general cloud services on Azure,such as Teams,SQL servers,and plan to offer virtual machines for different kinds of workloads.68 Microsoft has signed a$15 billion deal with Intel to produce custom chips.69Sanjiv AgarwalSemiconductor Industry Leader,Capgemini“Semiconductors are enabling the deploymen

231、t of Artificial Intelligence(AI)across enterprises.However,unlocking AIs full potential requires collaboration among companies throughout the semiconductor value chain,as well as partnerships with downstream industries.This collective effort will accelerate innovation,reduce risks,and eliminate inef

232、ficiencies,ultimately lowering costs and enhancing overall performance for businesses.56Capgemini Research Institute 2025The semiconductor industry in the AI eraPartnerships are taking the industry forwardPartnerships allow semiconductor organizations to share costs,mitigate reducing financial risks

233、,and more effectively meet growing demand for specialized components.In our research,we identify several industry collaborations and highlight five themes emerging from these partnerships(see figure 26).Sanjiv Agarwal,Semiconductor Industry Leader from Capgemini,adds,“Semiconductors are enabling the

234、 deployment of Artificial Intelligence(AI)across enterprises.However,unlocking AIs full potential requires collaboration among companies throughout the semiconductor value chain,as well as partnerships with downstream industries.This collective effort will accelerate innovation,reduce risks,and elim

235、inate inefficiencies,ultimately lowering costs and enhancing overall performance for businesses.57Capgemini Research Institute 2025The semiconductor industry in the AI eraFigure 26.Five major themes of partnerships in the semiconductor industry Source:Capgemini Research Institute analysis.TSMC,Bosch

236、,Infineon,and NXP established the European Semiconductor Manufacturing Company(ESMC)joint-venture in Dresden,Germany,to bring an advanced semiconductor manufacturing presence to Europe.GlobalFoundries and STMicroelectronics announced the creation of a new,jointly-operated,high-volume semiconductor m

237、anufacturing facility in Crolles(France).TSMC will build its second factory in Japan with backing from Sony and Toyota.Supply chainIntermolecular is working with Merck to prototype and test new materials,accelerating innovation and commercial production.Synopsys and Microsoft formed a multi-year col

238、laboration to apply hyperscale cloud and AI to electronic design automation(EDA),including enhanced chip development.Samsung Semiconductor Europe showcased its latest foundry innovations for high-performance,low-power semiconductors targeting the AI and automotive industries.These include advanced G

239、ate-All-Around(GAA)technology,Bipolar-CMOS-DMOS(BCD)solutions,and eMRAM,enabling industries to fully leverage AI potential.InnovationPartnershipsTo be cont.58Capgemini Research Institute 2025The semiconductor industry in the AI eraIntel and Siemens have collaborated to advance factory efficiency and

240、 sustainability across Scopes 1 to 3 of the value chain,while supporting global industry ecosystems.Infineon and Amkor have partnered to engage with common suppliers to help them drive decarbonization strategies and boost greener supplier engagement.Merck in partnership with Intel funded a new three

241、-year academic research program in Europe to leverage AI and machine learning technologies to drive breakthroughs in innovative and more sustainable semiconductor manufacturing processes and technologies.SustainabilityIntel and Advanced Micro Devices,among others,will partner to ensure compatibility

242、 across their chips.Partners will contribute technical insights to identify“essential functions and features”for AMD and Intel chips.SoftwarizationTSMC has named TXOne,a cybersecurity provider to the semiconductor industry,as a collaborator on OT cybersecurity.AMD has teamed with Microsoft and OEMs

243、to implement layered PC defences with AMD Ryzen processors,helping to improve security from the endpoint and the edge to the cloud.NXP leverages Scalys as a cyber security specialist,providing the organization with the right support for cybersecurity innovations.CybersecurityPartnerships59Capgemini

244、Research Institute 2025The semiconductor industry in the AI era60Capgemini Research Institute 2024How the semiconductor industry can capitalize on emerging opportunities0560Capgemini Research Institute 2025The semiconductor industry in the AI eraTo navigate the bountiful but complex landscape of sem

245、iconductor opportunities,we propose the following recommendations for the semiconductor industry,based on our research,in-depth interviews,and experience.Figure 27.Focus areas for the semiconductor industrySource:Capgemini Research Institute analysis.Align with government policies and foster open in

246、novation123456Galvanize semiconductor supply chains through diversification and adoption of eco-friendly processesDiversify and secure supplier networksImplement sustainable manufacturing practicesMinimize ecological footprintHarness AI and Gen AI for advanced chip design and manufacturingAutomate d

247、esign processes using AIUse intelligent systems to improve production efficiency and yieldsDevelop AI-optimized semiconductor productsInvest in advanced manufacturing technologies and innovationAdopt next-generation fabrication methodsAccelerate research and development in emerging technologiesAdopt

248、 open standards to foster innovationPromote industry-wide collaboration for cross-platform interoperabilityEncourage open-source design and development in semiconductor innovationProtect intellectual property and enhance security measuresStrengthen cyber defenses and protect data integritySafeguard

249、proprietary technologiesCoordinate strategies with regulatory frameworksCollaborate and share innovation26%Percentage of downstream organizations which agree that the semiconductor industry is supplying chips at a rate sufficient for their needs61Capgemini Research Institute 2025The semiconductor in

250、dustry in the AI eraInvest in advanced manufacturing technologies and innovation Adopt next-generation fabrication methodsTo remain competitive in a changing market,semiconductor manufacturers must innovate continuously to create more powerful,efficient chips at lower costs.The following technology-

251、driven manufacturing techniques can support this drive:3D chip stacking:Vertical integration allows semiconductor organizations to increase the density and performance of chips without increasing the physical footprint.3D chip stacking supports heterogeneous integration,allowing the combination of m

252、any chip types in a single versatile and powerful package.The method enhances performance and bandwidth,creating new opportunities for advanced computing applications.Our research suggests that 64%of semiconductor organizations will prioritize the development of 3D IC design techniques and multi-die

253、 integration in the next two years.Automated fabrication processes:AI can help businesses reduce the incidence of human error by automating complex and repetitive tasks,resulting in more dependable,higher-quality semiconductor products.Automation also allows faster production cycles to help manufact

254、urers meet rising demand with enhanced accuracy.Almost half(49%)of semiconductor manufacturers say they are already using advanced AI and ML algorithms for process optimization and predictive maintenance.Advanced EUV lithography:This technique supports the creation of incredibly fine patterns on sil

255、icon wafers and the production of smaller,more complex semiconductor chips and is consequently crucial to the development of technologies such as 5G networks,IoT,and AI.Higher transistor densities greatly improve electronic device performance and energy efficiency.Additionally,by reducing the requir

256、ed number of patterning steps,EUV lithography reduces complexity and cost.Major semiconductor manufacturers such as Intel70 and TSMC71 have invested heavily in EUV technology.Accelerate R&D in emerging technologiesShowcasing a strong commitment to R&D ensures long-term growth and resilience.Semicond

257、uctor organizations should invest in the following R&D initiatives:Neuromorphic computing hardware:By imitating the structure of the human brain,neuromorphic computing hardware can process complex,unstructured data more effectively than conventional hardware.Parallel computing with neuromorphic circ

258、uits may open up opportunities in fields such as robotics,driverless cars,and healthcare.Additionally,because neuromorphic computing can handle data more effectively and requires less processing power,it drastically lowers AIs carbon footprint.Semiconductor organizations can obtain a competitive adv

259、antage through early adoption of this technology.Intel has recently announced its neuromorphic system,code-named Hala Point.This utilizes Intels Loihi 2 processor to support research for future brain-inspired AI and tackles challenges related to the efficiency and sustainability of todays AI.72Advan

260、ced silicon photonics integration:Advanced silicon photonics integration can help modern computing and communication systems meet the increasing demand for quicker and more effective data transfer.Photonic devices can offer greater bandwidths and reduced latencies,boosting operations at 5G networks,

261、data centers,and AI applications.Process node miniaturization:Miniaturization allows more transistors to be packed onto a single chip,resulting in more62Capgemini Research Institute 2025The semiconductor industry in the AI erafunctional and powerful integrated circuits.Businesses can lower the cost

262、per transistor and increase total cost efficiency by streamlining the process.Miniaturization of high-density circuits that can execute complex tasks helps address the growing demand for mobile devices,AI,and high-performance computing.Our research reveals that 31%of semiconductor manufacturers are

263、prioritizing semiconductor performance and miniaturization by embracing advanced lithography techniques.Galvanize semiconductor supply chains through diversification and adoption of eco-friendly processesDiversify supplier networksBelow are a few strategies for semiconductor organizations to diversi

264、fy their relationships and invest in technologies to enhance and secure the supply chain:Foster international partnerships:Global collaboration is an effective supply chain risk-mitigation strategy.Joint ventures with local firms can allow semiconductor organizations to access new markets and naviga

265、te unfamiliar regulatory systems more easily,as well as increase the customer base and revenue.Through partnerships,organizations can access a wider range of technologies and expertise and mitigate potential supply chain bottlenecks.For instance,TSMC,Robert Bosch,Infineon,and NXP Semiconductors anno

266、unced their European Semiconductor Manufacturing Company(ESMC)joint venture in Germany to provide advanced semiconductor manufacturing services within the EU.According to Jochen Hanebeck,CEO of Infineon Technologies,“Our joint investment is an important milestone to bolster the European semiconducto

267、r ecosystem.Dresden is strengthening its position as one of the worlds most important semiconductor hubs and is already home to Infineons largest front-end site.”73 Collaboration also facilitates cost-sharing for ambitious projects and fosters resilience through diversification.Develop alternative m

268、aterials:The manufacturing of various high-performance components requires rare earth elements,the supply of which is often concentrated in a limited number of countries,leading to geopolitical and economic vulnerabilities.Our research suggests that the limited availability of critical materials suc

269、h as silicon wafers,inert gases,and rare earth elements affects 67%of semiconductor supply chains.Investment in research to find viable alternative materials can mitigate such risks.Furthermore,diversifying material sources can enhance resilience against market fluctuations and trade restrictions.63

270、Capgemini Research Institute 2025The semiconductor industry in the AI eraImplement digital twin technology:Digital twins enable proactive simulation of various scenarios,allowing the anticipatory identification of vulnerabilities.By analyzing these simulations,organizations can develop mitigation st

271、rategies and optimize their operations,ensuring continuity of supply chain operations.Intel collaborated with supply chain software organization AIMMS to optimize its supply chain using digital twin.This has improved product flow,reduced delivery times,increased chip availability,and managed cost.74

272、 Digital twins offer a competitive edge by boosting agility and responsiveness in an industry where precision,quality,and timely delivery are critical.Implement sustainable manufacturing practicesLowering the environmental impact of the chip manufacturing process is of the utmost importance to the s

273、emiconductor industry.The following actions address urgent market demand and regulatory requirements:Using sustainable materials and packaging:Traditional packaging materials are often non-biodegradable and environmentally harmful.TSMC has collaborated with suppliers to optimize the materials used i

274、n drum packaging,proactively replacing raw materials that contain substances of very high concern(SVHC),such as phosphite ester(TNPP),and refining drum-bottom design to reduce chemical residue by 60%while maintaining quality and safety standards.Following full implementation by 2025,TSMC expects to

275、reduce chemical residue by 3.2 million liters annually and cut carbon emissions by around 570 tons,increasing environmental sustainability and green manufacturing capabilities.75Implementing green chemistry practices:The semiconductor industry uses a variety of potentially harmful chemicals.By shift

276、ing to environmentally friendly alternatives,organizations can reduce their ecological footprints.Furthermore,adopting environmentally responsible practices helps to ensure compliance with increasingly stringent regulations,avoiding costly fines and legal issues.Our research reveals that 63%of organ

277、izations are either already deploying such practices across some regions/products or are actively scaling up the use of less toxic alternatives.Minimize ecological footprintCommitment to sustainability is not just a trend but an essential step in ensuring the industrys long-term viability.This motiv

278、ates organizations to balance technological advancement with environmental responsibility by switching to renewable energy sources and designing energy-efficient chips.Utilize renewable energy sources:Reducing reliance on fossil fuels and mitigating the impact of fluctuating energy prices can lead t

279、o long-term cost savings.Switching to 31%Percentage of semiconductor manufacturers prioritizing semiconductor performance and miniaturization by embracing advanced lithography techniques64Capgemini Research Institute 2025The semiconductor industry in the AI eraHarness AI and Gen AI for advanced chip

280、 design and manufacturingAutomate design processes using AI and Gen AIAI-led automation accelerates development cycles.ML algorithms and Gen AI can optimize various stages of the chip design workflow,from initial architecture to physical layout.For instance:Googles AlphaChip uses RL to optimize chip

281、 layouts in hours rather than the weeks or months it would otherwise require and has been used to design the last three generations of TPUs.78 Design Space Optimization AI(DSO.ai)from Synopsys is a generative optimization paradigm that uses RL technology to autonomously search design spaces,optimizi

282、ng for power,performance,and area(PPA).Samsungs SoC designers used DSO.ai to fine-tune the design and maximize yields for its first 3-nm smartphone SoC.79 Electronic design automation(EDA)firms employ ML algorithms to optimize chip design.Cadence introduced Cadence Cerebrus to improve productivity t

283、hrough a renewable energy sources can also ensure a more stable and reliable energy supply,supporting the continuous operation of manufacturing processes.To enable this,semiconductor organizations are already collaborating with partners to source renewable energy.Micron,for instance,signed a virtual

284、 power purchase agreement for its US operations with Terra-Gen.Under the agreement,Micron will receive around 178 megawatts(MW)of wind electricity capacity and associated renewable energy certificates annually.This contract will support Microns US target to reach 100%renewable energy by the end of 2

285、025 and marks the latest milestone in Microns journey to achieve net-zero GHG emissions by 2050.76 Design energy-efficient chips:According to AMDs CEO,Lisa Su,”Driving performance gains over the next decade requires relentless focus on energy efficiency.”77 Semiconductor organizations can reduce ene

286、rgy consumption by focusing on energy-efficient chips.This also extends the battery life of portable devices,enhancing user experience and convenience.Additionally,energy efficiency brings a competitive edge by meeting regulatory standards and consumer expectations for eco-friendly products.unique R

287、L engine.In one instance,the ML model reduced overall optimization time from 18 to less than nine days.80Use intelligent systems to improve production efficiency and yields Through intelligent operations such as AI and ML,semiconductor manufacturers can enhance production efficiency,optimize complex

288、 fabrication processes,reduce the incidence of defects,and accelerate time-to-market for advanced chips.Examples include:TSMC deployed AI and ML to develop systems for precise fault detection and classification,and intelligent advanced equipment and process control,to ensure the consistency of tool

289、matching and process stability.This allows the organization to meet 5Gs stricter quality requirements for mobile,high-performance computing(HPC),automotive,and IoT.81 ASML relies on AI to improve the performance of its lithography machines.The organization uses ML software to predict deviations in t

290、he lithography process.82 Intel is relying on AI to analyze bug reports and categorize bugs to allow engineers to respond immediately to issues with new designs.It also uses AI to investigate minute variations in chips coming off production lines,including clock speed,cache memory size,and power con

291、sumption.The AI system scrutinizes this complex collection of details 65Capgemini Research Institute 2025The semiconductor industry in the AI eraBrett BonthronIndustry Leader for Tech and Digital,CapgeminiSoftwarization is transforming the semiconductor industry,enabling unprecedented flexibility an

292、d scalability.By decoupling hardware from software,companies can rapidly adapt to evolving demands,optimize performance across diverse applications,and unlock new capabilities through updates.This shift not only accelerates innovation but also fosters collaboration across the ecosystem,as software b

293、ecomes the bridge connecting hardware excellence with cutting-edge functionality.66Capgemini Research Institute 2025The semiconductor industry in the AI eraand finds the best fit for these processors in one of Intels products.Intel has further begun testing Gen AI tools that process the detailed spe

294、cs for a chip and create tests to ensure the chip design meets them.83 Develop AI-optimized semiconductor productsAs downstream industries integrate AI into their products and services,the semiconductor industry must focus on products optimized for AI workloads.These specialized chips are designed t

295、o deliver enhanced performance,energy efficiency,and scalability.For instance:NVIDIAs H100 Tensor Core GPU offers significant advances in computational capabilities for training and inference of large-scale AI models,including Gen AI applications such as GPT-3 and GPT-4.84 Major cloud service provid

296、ers,including Google Cloud,Microsoft Azure,and AWS,have integrated H100 GPUs into their data centers.NVIDIA is also collaborating with AWS to offer a low-cost inference for Gen AI with Amazon SageMaker integrated into NVIDIAs NIM inference microservices.This allows customers to deploy pre-compiled a

297、nd optimized FMs at speed,reducing time to market for Gen AI applications.85 AMD launched its MI300X accelerator in 2023,a GPU designed for Gen AI workloads and HPC applications,featuring a large memory capacity to handle extensive language models and complex neural networks.86Intels AI accelerator

298、Gaudi 2 offers a benchmarked alternative to NVIDIAs H100 by driving enhanced performance for deep learning training workloads.87Align with government policies and foster open innovationCoordinate strategies with regulatory frameworksGovernment policy can have a significant impact on the operations o

299、f a semiconductor organization within its territory.Organizations should take the following steps to ensure they are aligned with local frameworks and ready to take advantage of the benefits they offer.Stay informed and adaptive:Government policy has a strong influence on availability of raw materia

300、ls for semiconductor manufacturing,including critical resources and minerals.These policies also impact trade agreements and tariffs,influencing cost and supply chain dynamics.Organizations should monitor government policies carefully,adjusting business strategies to align with regulatory changes,co

301、mpliance,market access,and opportunities.Collaborate with government initiatives:National programs such as tax credits or subsidies for R&D can offer significant benefits,helping semiconductor organizations to offset the extreme costs associated with chip manufacturing.Tax benefits against R&D allow

302、 organizations to deploy more resources on innovation,as well as ensuring a reliable and resilient supply chain.TSMC Arizona and the US Department of Commerce announced up to$6.6 billion in direct funding under the CHIPS and Science Act,fulfilling the goal of bringing advanced chip manufacturing to

303、the US.88 Collaborate and share innovationStrategic collaborations with organizations in similar fields,research institutions,academic bodies,etc.,help semiconductor organizations to pool resources and expertise to address complex challenges and accelerate innovation.Such collaborations facilitate t

304、he creation of cutting-edge solutions,from advanced chip design to new manufacturing techniques,and lay the foundation for a more interconnected,resilient,and innovative semiconductor ecosystem.Areas where semiconductor organizations can participate are:Industry consortia:Partnerships can address th

305、e growing complexity of technology development.Pooling knowledge and expertise and harnessing complementary strengths and advantages(e.g.,production capacity)can accelerate innovation,promote industry standardization,help streamline manufacturing operations,and improve interoperability.Such relation

306、ships also strengthen collective resilience to supply chain disruption.67Capgemini Research Institute 2025The semiconductor industry in the AI eraEcosystem partnerships:Strategic alliances with suppliers,customers,and technology partners enhance innovation and extend market reach.Suppliers can help

307、semiconductor organizations access cutting-edge raw materials or components.Customers can help organizations understand market demand and direct curation of products specific to requirements.Technology partners can boost the integration of cutting-edge technologies,such as AI,into their operations o

308、r product offerings.Collectively addressing growth challenges:New technologies and other innovations bring with them new challenges in terms of talent and sustainability requirements.Working together,industry,government,academia,and civil society can meet these requirements in a way that benefits al

309、l parties.Industry can drive innovation and enrich the talent pool through rigorous training programs,while governments can boost investment and innovation with sympathetic policy formulation.Academia helps develop talent at the tertiary educational level,and civil society imposes checks and balance

310、s by pressuring for healthy sustainability practices.Protect IP and enhance security measures Strengthen cyber defenses and protect data integrityThe complexity of semiconductor supply chains attracts a critical threat landscape.Implementing robust security measures to every link of the semiconducto

311、r supply chain is important for enhancing cybersecurity and data integrity.The threat of more sophisticated cyberattacks means organizations should continue to innovate to safeguard the integrity of both their products and data.To bolster cybersecurity and maintain data integrity organizations shoul

312、d:Ensure trusted supply chain components:Maintaining the integrity and reliability of products using chips requires a trusted semiconductor supply chain.TSMC upgraded the security of its factories by requiring new equipment to comply with standards set out in procurement contracts,strengthening the

313、cybersecurity of its supply chain.89 Sourcing parts from verified,secure suppliers helps prevent the infiltration of counterfeit parts,which can jeopardize product functionality and safety.Organizations should implement robust supplier-verification processes,including regular audits and certificatio

314、ns,such that they adhere to quality standards and provide traceability to meet industry specifications.Technologies such as blockchain can further secure the supply chain by providing transparent,immutable records of component origins.Implement robust cybersecurity measures:Organizations can prevent

315、 unauthorized access and data breaches by using advanced security protocols,which ensure that sensitive information remains confidential while avoiding financial losses associated with IP theft.Invest in hardware-based security:Semiconductor organizations should prioritize investment in hardware-bas

316、ed security to boost the overall resilience and trustworthiness of products.Incorporation of encryption and hardware security into chip design can provide robust protection against a wide range of cyber threats.Our research reveals that 58%of semiconductor organizations use cryptographic protection

317、when designing hardware-based security for chips.Additionally,integration of security features at the hardware level can enhance performance and reduce latency compared with software-only solutions.As cyber threats evolve,establishing a strong hardware security foundation becomes essential to mainta

318、ining trust and safeguarding critical information.Harness AIs threat-detection potential:Semiconductor organizations can proactively monitor and respond to security threats using AI and ML.AI-driven systems can analyze vast amounts of data in real time,identifying patterns 68Capgemini Research Insti

319、tute 2025The semiconductor industry in the AI eraand anomalies that could indicate potential threats in a dynamic defense mechanism that avoids costly security incidents.Safeguard proprietary technologiesImplement advanced security measures:A comprehensive approach will be most effective in safeguar

320、ding IP.Organizations should deploy end-to-end encryption for both data at rest and in transit to secure designs,fabrication processes,and trade secrets.Multi-factor authentication and strict access controls can help limit internal threats.Similarly,AI-powered intrusion detection systems can help mo

321、nitor networks in real time.Regular security audits and vulnerability assessments can identify weaknesses in security operations.A few organizations also conduct“red-team”exercises to simulate cyberattacks,assisting the preparation for real offensives.Implementing secure boot processes and code sign

322、ing can further protect firmware and software layers.Employ technological solutions for IP management:Advanced technologies can further safeguard industry IP.Blockchain technology can create immutable and transparent records to store patent registrations,licensing agreements,and transaction details,

323、helping them to resolve disputes.IPwe created a blockchain-based platform that helps organizations manage their patent assets.90 Organizations can deploy AI and ML to monitor and detect potential IP infringements,and quantum-resistant cryptography(QRC)can withstand the advanced hacking capabilities

324、of quantum computers.KSE,a security IP platform,has integrated QRC,allowing semiconductor organizations to safeguard SoC products(and the devices that use them)from emerging quantum computing threats.91Enhance legal protection for IP security:IP fuels innovation and secures competitive advantage.Enh

325、ancing legal measures for IP security is essential to combat the increasing sophistication of IP theft,cyber espionage,and the global nature of semiconductor supply chains.Strengthened legal frameworks protect proprietary technologies from infringement,unauthorized use,and counterfeiting,as well as

326、protecting the financial and resource-based R&D investment and the return on that investment.The semiconductor industry should advocate for stronger IP laws and enforcement mechanisms,both domestically and internationally.It should collaborate with government agencies to update existing legislation,

327、ensuring it keeps pace with technological advancements in nanotechnology and quantum computing.Organizations could also push for stricter penalties for IP violations and faster patent approval.Given the global nature of semiconductor supply chains and operations,international treaties and agreements

328、 would benefit all organizations.The ongoing legal battle between Arm and Qualcomm over Qualcomms interpretation of the licensing agreement Arm had with Nuvia(which Qualcomm acquired)92 highlights the complex nature of intra-industry litigation,a complication that standardized industry regulation co

329、uld help avoid.Organizations should also clearly outline IP ownership,licensing rights,and confidentiality obligations in their contractual agreements with partners,suppliers,and customers.Adopt open standards to foster innovation Open standards are essential for interoperability,innovation,and redu

330、cing proprietary lock-in as technology evolves rapidly.Despite challenges like market demands and integration complexities,adopting widely recognized standards can streamline semiconductor development,foster collaboration,and enhance performance across the supply chain.Promote industry-wide collabor

331、ation for cross-platform interoperabilityorming consortia with manufacturers,developers,and end-users is vital to ensure compliance with open standards.These groups can drive cross-platform interoperability by 69Capgemini Research Institute 2025The semiconductor industry in the AI erastandardizing i

332、nterfaces,protocols,and communication methods.Unified frameworks,like common instruction sets and physical layer standards,reduce fragmentation,cut costs,and enhance scalability across the industry.Notable examples include a consortium that contributes to cooperation for open architecture and design

333、,such as the RISC-V93 and the Open Compute Project(OCP).94Encourage open-source design and development in semiconductor innovationThe semiconductor industry should embrace open-source models for chip design and development,enabling collaboration through shared design files,tools,and processes.This approach fosters innovation,accelerates product development,and ensures broader testing and validatio