《麦肯锡全球研究院：2024押注时刻：欧洲的人工智能机遇发展报告-提升欧洲在AI价值链中的竞争力（英文版）（15页）.pdf》由会员分享，可在线阅读，更多相关《麦肯锡全球研究院：2024押注时刻：欧洲的人工智能机遇发展报告-提升欧洲在AI价值链中的竞争力（英文版）（15页）.pdf（15页珍藏版）》请在三个皮匠报告上搜索。

1、October 2024QuantumBlack,AI by McKinseyTime to place our bets:Europes AI opportunityBoosting Europes competitiveness across the AI value chain.by Alexander Sukharevsky,Eric Hazan,Sven Smit,Marc-Antoine de la Chevasnerie,Marc de Jong,Solveigh Hieronimus,Jan Mischke,and Guillaume DagorretAt a glance A

2、 three-lens approachon adoption,creation,and energyis required to assess Europes competitiveness in the emerging generative AI(gen AI)economy.While much of the current discourse centers around large language models(LLMs),European policy makers and business leaders must look beyond LLMs.Adopting a ho

3、listic approach to capitalize fully on gen AIs potential could boost European labor productivity by up to 3 percent annually through 2030.On adoption,European organizations lag behind their US counterparts by 45 to 70percent.Yet this is where most of gen AIs economic potential lies.With the technolo

4、gy still in its early stages and much of its productivity gains yet to be unlocked,the window of opportunity for Europe remains wide open.On creation,Europe leads in only one of the eight segments of a simplified gen AI value chain:AI semiconductor equipment.Europe is a challenger in three other seg

5、ments:foundation models,AI applications,and AI services.But it has below 5 percent market share in the remaining four:raw materials,AI semiconductor design,AI semiconductor manufacturing,and cloud infrastructure and supercomputers.On energy,gen AI is expected to accelerate data center power demand,p

6、otentially accounting for more than 5 percent of Europes total electricity consumption by 2030.Without competitive electricity prices,it becomes less likely that European data centers will host gen AI applications and services.1 In this article,unless specified otherwise,“generative AI”(gen AI)encom

7、passes all AI technologies,including the latest advancements in genAI.2 By January 2023,the company had already gained 100 million users and was valued at$29 billion($80 billion today).This triggered massive investments to fund gen AI companies($25 billion in worldwide private investments in 2023)an

8、d spurred the release of multiple breakthrough innovations and competing models(for example,Googles Gemini and Metas Llama).Artificial Intelligence Index report 2024,Stanford University,2024;DigitalRank,Similarweb,accessed September 2024;Julia Boorstin,“Why OpenAI is the first company to be No.1 on

9、the CNBC Disruptor 50 list two years in a row,”CNBC,May 14,2024.3 In this article,“Europe”refers to the 27 member states of the European Union plus Norway,Switzerland,and the United Kingdom.4 Per academic research,innovators historically capture less than 5 to 10 percent of broader economic returns

10、generated by their inventions.Adopters of the technology and society at large generate the remaining returns.For more,see William D.Nordhaus,Schumpeterian profits in the American economy:Theory and measurement,National Bureau of Economic Research working paper,number 10433,April 2004.5 For example,f

11、rom 2016 to 2022,annual growth was 0.5 percent in Western Europe and 1.2 percent in North America.From 2002 to 2007,it was 1.1 percent and 1.9 percent,respectively.Chris Bradley,Jan Mischke,Marc Canal,Olivia White,Sven Smit,and Denitsa Georgieva,“Investing in productivity growth,”McKinsey Global Ins

12、titute(MGI),March 27,2024.6 Eric Hazan,Anu Madgavkar,Michael Chui,Sven Smit,Dana Maor,Gurneet Singh Dandona,and Roland Huyghues-Despointes,“A new future of work:The race to deploy AI and raise skills in Europe and beyond,”MGI,May 21,2024.Europe has made major progress in raising AI awareness and set

13、ting commitments,but major bottlenecks persist.Policy makers and business leaders could explore several levers,including increasing investments(such as a public innovation procurement in AI applications for healthcare and defense sectors),leapfrogging in emerging semiconductor technologies(such as q

14、uantum and neuromorphic computing),and addressing talent retention.Additionally,preparing the workforce through reskilling and mobility programs will be crucial in fully leveraging thebenefits of gen AI adoption.A holistic approach to help Europe realize generative AIs full potentialFor generative A

15、I(gen AI),1 the blockbuster release of OpenAIs ChatGPT in November 2022 marked the beginning of a boom.2 Since then,much of the conversation around the technology has focused on foundation models,particularly large language models(LLMs).In this field,Europe3 appears to be lagging behind its counterp

16、arts.However,LLMs are just one part of the gen AI landscape.Engaging on gen AI adoption,creation,and energy requirements can help capture a more complete picture of where the region stands in the emerging gen AI economy.Most of the value generated by gen AI will stem from organizations adoption and

17、scaling of gen AI solutions4an important consideration in Europe,where labor productivity has been slowing.5 McKinsey Global Institute(MGI)research estimates that gen AI could help Europe achieve an annual productivity growth rate of up to 3 percent through 2030(Exhibit 1).6 This potential additiona

18、l growth 2Time to place our bets:Europes AI opportunityExhibit 1Web Exhibit of Generative AI productivity potential in Western Europe in 2030,by sector,$billion1Generative AI could add$575.1 billion to the European economy by 2030.McKinsey&Company1Western Europe:Austria,Belgium,Denmark,Finland,Franc

19、e,Germany,Greece,Ireland,Italy,Netherlands,Norway,Portugal,Spain,Sweden,Switzerland,and UK.Potential value add from 2019 base period.Source:“The economic potential of generative AI:The next productivity frontier,”McKinsey,June 14,2023Consumer goods and retail101.9Total potential valueof potential pr

20、oductivitygains are from sectorswith high spending gaps andhigh productivity potentialHealthcare and pharma57.2Telecommunications 11.6Insurance 9.8Agriculture 5.1Banking andcapital markets44.8High techand software44.0Chemicals and materials29.2Energy and utilities27.2Media and entertainment27.0575.1

21、56%Constructionand real estate55.7Professionalservices54.3Transportation53.9Advancedmanufacturing53.4will be critical for financing the European model,particularly in navigating the energy transition,solving the empowerment gap,and supporting an aging population.7 It could also drive breakthrough in

22、novations that transform daily life,such as accelerated drug discovery,improved patient care,and personalized education.7 For more,see Kweilin Ellingrud,Marco Piccitto,Tilman Tacke,Rebecca J.Anderson,Ishaa Sandhu,and Kevin Russell,“A better life everyone can afford:Lifting a quarter billion people t

23、o economic empowerment,”MGI,May 20,2024;Mekala Krishnan,Chris Bradley,Humayun Tai,Tiago Devesa,Sven Smit,and Daniel Pacthod,“The hard stuff:Navigating the physical realities of the energy transition,”MGI,August 14,2024.8 Artificial Intelligence Index report 2024,Stanford University,2024.In terms of

24、creation of gen AI,since 2022,more than 90 percent of LLM-related funding has taken place outside of Europe.8 Moreover,European companies represent only 25 of the 101 AI models considered notable by the Stanford University AI Index,far behind US companies(which boast 61 notable models).But the oppor

25、tunities for capturing the economic value resulting from the creation of gen AI technologies extend well beyond LLMs.3Time to place our bets:Europes AI opportunityThey are spread across an eight-segment value chain:raw materials,AI semiconductor equipment,AI semiconductor design,AI semiconductor man

26、ufacturing,cloud infrastructure and supercomputers,foundation models(including LLMs),AI applications,and AI services.9Finally,to power the creation and adoption of gen AI,Europe also needs to consider its energy capacity.This is a key consideration,given that Europes energy system will be forced by

27、2030 to manage a rise in consumption of more than 5 percent,triggered by the demand for data center power(accelerated by gen AI).10To realize the full potential of gen AI,Europes business leaders and policy makers must embrace a holistic view of the technology that encompasses the challenges and opp

28、ortunities posed by creation,adoption,and energy(Exhibit 2).In this article,we describe those challenges,detailing where Europe 9 Simplified value chain of the most important segments(excludes other AI elements,such as distribution platforms and vector databases).10 Electricity Data Explorer,Ember,a

29、ccessed September 2024;McKinsey research and analysis.11 For more,see Zach Meyers and John Springford,“How Europe can make the most of AI,”Centre for European Reform,September 14,2023.stands relative to other regions,and provide a series of steps that leaders in Europe might consider if they are to

30、fully participate inand tap into the value created bythis impressive new technology.Adoption of gen AI:Opportunity remains wide open,but Europe is starting from a disadvantageThe vast majority of the economic value of gen AI is expected to come from its adoption by European organizations.The technol

31、ogy is still in its early stages,and most productivity potential has yet to be captured,so the opportunities here remain wide open.Yet European corporations are moving much more slowly than those in other countries.11How much is Europe lagging behind?The information here is incomplete,so we sought t

32、o quantify it by examining three indicators.First,we Exhibit 2Web Exhibit of To fully capture the value of generative AI,European leaders can embrace a holistic approach that encompasses creation,adoption,and energy.McKinsey&CompanyCreationAdoptionCreation of new technologies and applications across

33、simplifed 8-step generative AI(gen AI)value chainEnergyPower required to run gen AI applications,with low carbon emissions and competitive pricesDeployment of gen AI technologies acrossdiferent use cases to increase labor productivity 1.Raw materials2.AI semiconductor equipment3.AI semiconductor des

34、ign4.AI semiconductor manufacturing5.Cloud infrastructure and supercomputers6.Foundation models7.AI applications8.AI servicesPotential high-impact use cases:Chatbots for customer service in retailAI-driven drug discovery in pharmaceuticalsSupply chain optimization in logistics4Time to place our bets

35、:Europes AI opportunitylooked at external AI spending of corporations,such as the purchase of AI software-as-a-service(SaaS)solutions.Since not all AI spending is externalsome,such as hiring AI engineers,is internalwe also examined general IT spending,of which AI is a component,as an indicator of IT

36、 readiness and a crucial foundation for AI adoption.Finally,we factored in the responses of European executives to the McKinsey Global Survey on the state of AI.12We analyzed the first two metrics both in absolute terms and relative to company sales,comparing them with US figures when possible.This

37、relative comparison helps account for differences in sector size,which would otherwise skew the data because of economies of scale.For instance,the high-tech and software sector is 4.9 times larger in the United States than in Western Europe,13 so we find an AI external spend-to-sales ratio of 0.4 p

38、ercent for the United States versus 0.7 percent for Western Europe.But in AI external spend absolute value,we find$8.7 billion versus$2.6 billion,respectively,leading to a 70 percent gap.Additionally,with the two first metrics,figures show that companies in Western Europe lag behind their US counter

39、parts by 45 to 70 percent.This gap exists across all sectors.When evaluating sectors of similar size14 in Western Europe and the United States(for example,advanced manufacturing,12 The online survey was in the field from April 11 to April 21,2023,and garnered responses from 1,684 participants repres

40、enting the full range of regions,industries,company sizes,functional specialties,and tenures.13 Austria,Belgium,Denmark,Finland,France,Germany,Greece,Ireland,Italy,the Netherlands,Norway,Portugal,Spain,Sweden,Switzerland,and the United Kingdom.14 Sectors with a size ratio between Western Europe and

41、the United States below 2:1.15 Survey question,with 1,363 responses:Has the organization adopted AI in at least one business function?chemicals and materials,and construction and real estate),we find that those in Europe lag behind by 45 to 55 percent.For sectors that are significantly larger in the

42、 United States than in Western Europe(for example,healthcare and pharma,high tech and software,and media and entertainment),the gap was even more pronounced,ranging from 50 to 70percent(Exhibit 3).When looking at external spending on AI infrastructure,software,and services,Western Europe lags behind

43、 the United States by an average of 61 percent for sectors of similar size and 71percent for sectors that are two or more times larger in the United States than in Western Europe.Looking at internal IT spend,we see that for sectors of similar size,Western Europe lags behind the United States by an a

44、verage of 43 percent,and by 46 percent when sectors differ in size by at least twotimes.Per the 2023 McKinsey Global Survey on the state of AI,Europe lags behind North America in gen AI adoption by 30 percent,with 40 percent of surveyed North American companies reporting having adopted gen AI in at

45、least one business function,compared with about 30 percent for surveyed European companies.15Western Europe lags behind the United States on external spending on AI by an average of 61 percent for sectors of similar size.5Time to place our bets:Europes AI opportunityExhibit 3Western Europe lags behi

46、nd the United States in AI and IT spending across sectors,with an average gap of 4570 percent.Note:IT spending used as proxy for AI internal spending.Sectors ordered from most similar in size to least similar.1Austria,Belgium,Denmark,Finland,France,Germany,Greece,Ireland,Italy,Netherlands,Norway,Por

47、tugal,Spain,Sweden,Switzerland,and UK.2Sectors with size ratio 2:1 between US and Western Europe or Western Europe and US.3Sectors with size ratio 2:1 between US and Western Europe or Western Europe and US.Source:Worldwide AI and Generative AI Spending Guide,IDC,February 2024;McKinsey analysisMcKins

48、ey&CompanyIT internal spending gap between Western Europe1 and US in 2022,by sector Construction and real estateChemicals and materialsAdvanced manufacturingInsuranceEnergy and utilitiesTelecommunicationsTransportationAgricultureProfessional servicesWestern EuropeUS05101505010015020025030020Spending

49、 as share of sales in sectors of similar size,2%Consumer goods and retailBanking and capital marketsHealthcare and pharmaHigh tech and softwareMedia and entertainmentAbsolute spending in sectors of difering size,3$billionWestern EuropeUSAverage gap in relative value to sales45%Average gap in absolut

50、e value 50%AI external spending gap between Western Europe1 and US in 2022,by sectorConstruction and real estateChemicals and materialsAdvanced manufacturingInsuranceEnergy and utilitiesTelecommunicationsTransportationAgricultureProfessional servicesWestern EuropeUS0024681012140.20.40.60.81.0Spendin

51、g as share of sales in sectors of similar size,2%Consumer goods and retailBanking and capital marketsHealthcare and pharmaHigh tech and softwareMedia and entertainmentAbsolute spending in sectors of difering size,3$billionWestern EuropeUSAverage gap in relative value to sales55%Average gap in absolu

52、te value 70%6Time to place our bets:Europes AI opportunityCreation of gen AI tech:Europe leads in one segment,is a challenger in three,but is almost absent in fourBeyond adoption,Europes ability to capitalize on gen AI will depend on its ability to spur the creation of gen AI technologies that sprea

53、d across the simplified eight-segment value chain:raw materials(for example,germanium and silicon),AI semiconductor equipment(for example,lithography systems),AI semiconductor design(for example,development of high-end GPUs),AI semiconductor manufacturing(for example,foundries),cloud infrastructure

54、and supercomputers(for example,infrastructure as a service and platform as a service),foundation models(for example,LLMs),AI applications(for example,AI-powered software),and AI services(for example,advisory services and implementation).Europe is currently competitive in four of the eight segments o

55、f the value chain:AI semiconductor equipment,foundation models,AI applications,and AI services.However,the region has less than 5 percent of global market share in the remaining four segments:raw materials,AI semiconductor design,AI semiconductor manufacturing,and cloud infrastructure and supercompu

56、ters(table):Europes ability to capitalize on gen AI will depend on its ability to spur the creation of gen AI technologies that spread across the value chain.7Time to place our bets:Europes AI opportunityEurope is strong in four segments of a simplified generative AI value chain and lags in the rema

57、ining four.TableSegmentDescriptionEuropean market share in 2023Historical European market share,directionalKey dataRaw materialsMaterials needed to produce semiconductors and their machinery(eg,gallium to make lithography tools)StableEurope supplies 5%of critical,strategic1 raw materials needed for

58、chip manufacturing and semiconductorsAI semiconductor equipmentGoods needed for AI semiconductor production(eg,silicon wafers,lithography tools)IncreasingEurope has 8090%market share for extreme ultraviolet lithography(allows for finer patterns on semiconductor wafers,essential for high-end AI chips

59、)AI semiconductor designDesign,including intellectual property,of semiconductors for AIDecreasingEurope has 2%share of design of logic semiconductors used for AI(eg,GPUs)AI semiconductor manufacturingProduction of semiconductors for AIStableEurope has 1%of worlds production capacity of 7-nanometer l

60、ogic semiconductors used for AICloud infrastructure and supercomputersInfrastructure,including basic software layer,needed for computing power and data hostingStableEuropean cloud companies have 40%Negligible(15%)1“Critical”is based on economic importance and supply risk,and“strategic”is defined as

61、important for the green and digital transition,defense,and aerospace.8Time to place our bets:Europes AI opportunity Raw materials.The chip-manufacturing and semiconductor industries require more than 40raw materials,16 of which(for example,gallium,magnesium,and silicon)the European Union classifies

62、as both critical and strategic.16 About 5 percent of these materials are supplied by European companies.As a result,the region relies heavily on imports from countries such as China,which supplies about 75 percent of the European Unions needs in silicon and 90 percent of its needs in gallium and mag

63、nesium.17 The Critical Raw Materials Act(CRMA)supports local production,streamlining permitting processes and boosting the recycling of key materials.18 AI semiconductor equipment.The Netherlandsbased ASML is the market leader for the lithography machines required to produce high-end semiconductors(

64、up to seven-nanometer logic)suitable for AI.19 European companies also lead in other equipment segments,such as atomic layer deposition(ASMInternational,also based in the Netherlands,with about a 50 percent market share)and metalorganic chemical-vapor deposition(Germany-based company AIXTRON,with 70

65、to 80 percent market share).20 Yet,in other key niches,like dry etchers and dicing machines,European companies are less present.AI semiconductor design.European companies like Infineon Technologies,NXP Semiconductors,and STMicroelectronics play a global role in the semiconductor-integrated-design-ma

66、nufacturing space,with about 15 percent market share in 2023.21 But Europe has less of a presence in the design 16“Critical”is based on economic importance and supply risk,and“strategic”is defined as important for the green and digital transition,defense,and aerospace.The 16 materials include galliu

67、m,germanium,rare earths,and silicon.17 Study on the critical raw materials for the EU 2023,European Commission,March 16,2023.18 Emma Watkins,Emma Bergeling,and Eline Blot,“Circularity gaps of the European Critical Raw Materials Act,”Institute for European Environmental Policy,October 30,2023.19“Fitc

68、h affirms ASML at A;outlook stable,”Fitch Ratings,April 5,2023.20 AIXTRON annual reports;ASM annual reports;DataTrack,TrendForce,accessed September 2024.21 Omdia,Informa,accessed September 2024.22 Kif Leswing,“Nvidia dominates the AI chip market,but theres more competition than ever,”CNBC,June 2,202

69、4.23 Masayuki Shikata and Akira Yamashita,“SoftBanks Arm plans to launch AI chips in 2025,”Nikkei Asia,May 23,2024.24 World Fab Forecast,SEMI(including discrete,analog,and memory semiconductors),accessed September 2024.25“Emerging resilience in the semiconductor supply chain,”Semiconductor Industry

70、Association,May 8,2024.26 Florian Dbes,“Il faut donner envie dinvestir en Europe,plaide le patron dASML(“We need to make people want to invest in Europe,argues the boss of ASML”),Les Echos,June 6,2024.of AI-suitable semiconductors,a space led by Nvidia.22 Nonetheless,some European players are taking

71、 steps to bridge the gap.Britain-based ARM has ambitions to launch AI semiconductors in 2025.23 Europe also plays an important,if indirect,role in AI semiconductor design through its strong position in the design and manufacturing of power semiconductors(for example,through Infineon and STMicroelect

72、ronics).AI semiconductor manufacturing.Europe produces only about 8 percent of the worlds semiconductors and fewer than 1 percent of the logic capacity semiconductors of up to seven nanometers suitable for AI.24 Beyond that,Europe has no capacity for high-bandwidth memory(HBM)and advanced packaging.

73、Looking ahead,global capacity for advanced semiconductor manufacturing is expected to continue to be fully owned by non-European players,such as TSMC.25 In large part,thats because fab payback time in Europe is higher than that of Southeast Asia,notably due to higher labor and energy costs.In additi

74、on to higher costs,European companies also face complex administrative processes.It can take up to four years to get a semiconductor plant up and running in Europe,compared to one year inTaiwan.26 Cloud infrastructure and supercomputers.Europe lags behind the United States in computing power.Europe

75、is home to 18percent of global data-center-installed capacity,compared with 37 percent in the United States(while European and US GDPs are comparable,with around$23 trillion and$27 trillion,respectively)and in most cases,9Time to place our bets:Europes AI opportunitythese European data centers are o

76、wned by US companies.27 In 2023,European cloud companies(for example,OVH and UpCloud)had about 5 percent market share globally(about 15 percent in Europe),while US players(for example,Amazon Web Services,Google,and Microsoft)had more than 70 percent global market share.28 Furthermore,Europe has only

77、 half the supercomputing capacity in flop/s,29 which is increasingly necessary in basic and applied research.30 This is partially because the United States has seen the emergence of private players specializing in this segment(for example,CoreWeave),while Europe supercomputers mostly lie in research

78、 centers.Whats more,the operating costs of European data centers are typically more than 50 percent higher than those in the United States,largely driven by Europes higher energy costs.31 Foundation models.In 2023,61 notable AI models32 originated from US-based organizations,far outpacing Europes 25

79、.33 A few of the European models are competing globally.One such is France-based Mistral 27 IDC Global data,accessed September 2024;International Monetary Fund data,accessed September 2024;McKinsey analysis and research.28 IDC Global data,accessed September 2024;McKinsey analysis and research.29 Mea

80、sured by total computing power of supercomputers in floating point operations per second.30 TOP500 release,62nd edition,TOP500,November 2023.31 Jonathan Atkin et al.,“RBC Datacenter download,”RBC Capital Markets,September 20,2021.32“A notable model meets any of the following criteria:(i)state-of-the

81、-art improvement on a recognized benchmark;(ii)highly cited(over 1000 citations);(iii)historical relevance;(iv)significant use.“What is a notable model,”Epoch AI,accessed September 2024.33 Artificial Intelligence Index report 2024,Stanford University,2024.34 Crunchbase data,accessed September 2024.3

82、5 PitchBook data,accessed September 2024.36“SAP advances vision of business AI with investments in Aleph Alpha,Anthropic and Cohere to complement$1+billion AI commitment from Sapphire Ventures,”PR Newswire,July 18,2023.37 PitchBook data,accessed September 2024.38 Crunchbase data,accessed September 2

83、024.AI,a leading open-source model provider,with$1 billion raised since 2023.34 Yet in the technological race to constantly improve models performances,the company remains underfunded compared with its US competitors.For example,OpenAI has raised$11.3 billion,and Anthropic has raised$8.7 billion.35

84、AI applications.Europe has several emerging AI unicorns(for example,DeepL,Synthesia,and Wayve).The region also is home to leading global software companies(such as Dassault Systmes,Hexagon,and SAP)that are increasingly building gen AI technologies into their solutions.For example,in 2023,SAP pledged

85、 to invest more than$1 billion in gen AI companies.36 But Europe lags behind the United States,garnering only 12 percent of the global pool of private equity and venture capital funding for SaaS AI companies as of 2023.37 Whats more,several leading AI start-ups and scale-ups of European origin(for e

86、xample,Hugging Face,with a$4.5 billion valuation,and Dataiku,with a$3.7 billion valuation38),have The operating costs of European data centers are typically more than 50 percent higher than those in the United States,largely driven by Europes higher energy costs.10Time to place our bets:Europes AI o

87、pportunitymoved their headquarters from Europe to the United States.AI services.Europe holds about a 15 percent share of the global AI services market,positioning it just behind the United States,which leads with approximately 40 percent.39 This significant market presence provides Europe with a fou

88、ndation for expanding AI-related services.The near absence of European companies in four of the eight segments of the simplified value chain could result in missed opportunities for the regions economy.The global market of gen AI technologies is expected to boom,with high double-digit annual growth

89、anticipated over the next ten years.40 This situation could be a challenge to the regions strategic autonomy,ultimately jeopardizing gen AI adoption and productivity gains.A semiconductor shortage in 2022,for example,hit the European auto industry especially hard,resulting in an estimated 100 billio

90、n GDP loss.41 Similarly,insufficient access to cloud infrastructure and supercomputers could limit development and operations of gen AI technologies.Energy for gen AI:Expected to drive increased electricity demand in Europe amid already-high pricesMcKinsey estimates that rising data center power dem

91、and could increase Europes electricity consumption by at least 180 terawatt-hours by 2030equivalent to more than 5 percent of total European electricity annual consumption in 2023.42 This is driven by demand for data center computing power in Europe,which McKinsey expects to more than triple by 2030

92、 to reach 35 gigawatts of 39 Riccardo Righi et al.,“EU in the global artificial intelligence landscape,”European Commission,2021.40“Generative AI to become a$1.3 trillion market by 2032,research finds,”Bloomberg Intelligence,June 1,2023.41“Missing chips cost EUR100bn to the European auto sector,”All

93、ianz,September 13,2022;International Organization of Motor Vehicle Manufacturers data,accessed September 2024.42 Electricity Data Explorer,Ember,accessed September 2024;McKinsey research and analysis.43“Investing in the rising data center economy,”McKinsey,January 17,2023.44 Electricity Data Explore

94、r,Ember,accessed September 2024;Patrick Chen,Tamara Grnewald,Jesse Noffsinger,and Eivind Samseth,“Global Energy Perspective 2023:Power outlook,”McKinsey,January 16,2024.45 Enerdata,Ember,US Energy Information Administration,Eurostat.46“Winds of change,”Nexans 2021 Capital Markets Day.47 Total electr

95、icity generation mix with low-carbon energy,including biofuel and wastes,hydro,wind and solar,and other renewable sources.International Energy Agency data,accessed September 2024.installed capacity.Indeed,data centers are major energy consumers:a hyperscalers data center can use as much power as 80,

96、000 households.43These new demands will place additional pressure on a European power grid thats already undergoing significant stresses.First,electricity demand is expected to escalate in the region on the back of growing decarbonization efforts and electrification throughout various sectors,with a

97、bsolute electricity demand expected to increase by 20 to 25 percent by 2030(from 3,200 terawatt-hours in 2023 to around 4,000 terawatt-hours in 2030,including demand from data centers).44 Also,energy price competitiveness in Europe is low,with industrial-electricity prices some 70 percent higher in

98、Europe than in the United States in May 2024.45 Finally,Europe has the oldest power grid in the world(45 to 50 years,on average,versus 35 to 40 years in North America and 15 to 20 years in China).46 This can lead to inefficiencies in electricity distribution.On the bright side,this significant incre

99、ase in electricity consumption could serve as a positive incentive for energy operators to invest in new capacities.Additionally,Europe has an edge in clean energy,with 61 percent of low-carbon sources in its electricity mix,compared with 40 percent in the United States and 34 percent in China.47How

100、 to boost Europes competitiveness in gen AIEurope clearly faces a host of challenges with gen AI,but they arent insurmountable.Policy makers and business leaders in Europe can consider several activities to increase the regions ability to fully realize the potential economic gains of AI when it come

101、s to adoption,creation,and energy.11Time to place our bets:Europes AI opportunityAdoption of gen AI in EuropeTo facilitate gen AI adoption,European leaders might consider the following actions:Reskill and upskill the workforce.Research from MGI indicates that to reap the full productivity dividends

102、of gen AI,Europe would need to double its current pace of job transitionfrom the 0.4 percent per year seen in 201619,prior to the COVID-19 pandemic,to an unprecedented 0.8 percent by 2030.48(The effort for the United States would be lower,as transition rates have already been at such levels.)That co

103、uld require the reskilling of 12million workers,roughly 6.5 percent of Europes current workforce.These occupational shifts herald a substantial evolution in workforce skills:by 2030,demand for technological skills is expected to increase by 25 percent compared with 2022,while demand for basic cognit

104、ive skills could decline by 14 percent.To support this change,policy makers could direct students and workers toward programs that prepare them for high-demand jobs.49 They could also encourage partnerships among universities and businesses.In doing so,policy makers must ensure that the right traini

105、ngs are developed and offered for reskilling and upskilling and set up dedicated mechanisms to fund the trainings.Attract and retain AI talent.Europe has slightly more AI professionals than the United States does(in 2023,120,000 versus 112,000).50 However,while 22 percent of the worlds leading AI re

106、searchers51 studied in Europe,only 14 percent continue to work in the region.52 Compensation disparities are a significant factor:in 2023,salaries for software developers in the United States 48 Eric Hazan,Anu Madgavkar,Michael Chui,Sven Smit,Dana Maor,Gurneet Singh Dandona,and Roland Huyghues-Despo

107、intes,“A new future of work:The race to deploy AI and raise skills in Europe and beyond,”MGI,May 21,2024.49 For more,see“Netherlands advanced:Building a future labor market that works,”McKinsey,June 18,2024.50 Measured by active AI roles per region.State of European Tech 2023,Atomico,November 28,202

108、3.51 Defined as top 20 percent of AI researchers,based on Annual Conference on Neural Information Processing Systems acceptance rate of papers.52 Global AI Talent Tracker 2.0,MacroPolo,accessed September 2024.53 2023 Developer Survey,Stack Exchange,June 13,2023;McKinsey analysis.54 Global AI Talent

109、Tracker 2.0,MacroPolo.55“The state of AI in early 2024:Gen AI adoption spikes and starts to generate value,”QuantumBlack,AI by McKinsey,May 30,2024.56 Refers to the concept of creating a harmonized framework across European countries that could target the tech sector,including AI-related scale-ups.T

110、his would involve a coalition of the willing across Europe to align tax policies,fiscal measures,and regulatory standards to create a unified environment that supports the growth and scaling of tech companieseffectively acting as a 28th regime alongside existing national frameworks.57 Matthieu Quire

111、t,“IA:les entreprises vont dans le mur prvient McKinsey”(“AI:Businesses are heading for disaster,warns McKinsey”),Les Echos,April 29,2024.were two to four times higher than those of their European counterparts.53 This disparity is likely attributable to the greater financial resources of US companie

112、s,which benefit from larger economies of scale and higher levels of venture capital and private equity funding.Europe also lags behind the United States in AI-related research,with only two universities in a key ranking of top institutions for AI research in 2022,compared with 15 for the United Stat

113、es.54 European workers also fall behind in gen AI awareness and use:the latest McKinsey Global Survey on AI finds that employees in North America are 15 percent more likely than their European counterparts to use AI regularly.55 To address these gaps,policy makers could implement measures to enhance

114、 Europes attractiveness to top-tier talent(for example,premiums or tax breaks for returning or incoming top talent,support for research institutions and private labs,and public grants),and business leaders could focus on upskilling workers on gen AI use.More broadly,addressing the differences in sca

115、le and resources among European companiesthrough measures such as establishing a 28th tech regime,56 simplifying business regulations,consolidating pension funds,and enhancing support for scale-up fundingcould also be beneficial.Fully embrace gen AI via end-to-end transformation.Research has found t

116、hat 90 percent of AI projects are stuck in the experimentation phase.57 Many companies stumble in their gen AI transformation by either starting too small or spreading resources too thin on a few use cases.Both tactics yield minimal value.In contrast,successful companies concentrate on fully transfo

117、rming a few critical 12Time to place our bets:Europes AI opportunitybusiness domains,such as innovation and R&D,logistics,and procurement,from end to end.To truly unlock the full potential of gen AI,companies must master crucial enterprise capabilities,from planning to scaling.58 Above all,they need

118、 to supercharge their operating model,bringing together business,technology,and operations into a powerhouse of innovation.On data,the focus should be on continually enriching both proprietary and real-world data and ensuring its organization-wide accessibility to enable smarter decisions at every l

119、evel.Help companies navigate regulation.Recent McKinsey research shows that eight of ten European companies report that they dont fully understand the obligations introduced by the EU AI Act,and 70 percent find them to be complex.That confusion has consequences.Meta,for example,recently stopped the

120、rollout of its multimodal model in the EU,reportedly because of a lack of readability and predictability of the regulatory environment.59 This situation could challenge the competitiveness of European companies by reducing their ability to access the worlds most high-performing AI models.Creation of

121、 gen AI in EuropeRegarding creating gen AI,winning in every segment isnt a realistic strategy for Europe.A differentiated approach,based on current strengths,is crucial for the region to stay relevant.Potential steps include the following:Increase investment.In 2023,US private investments in AI reac

122、hed$67 billion,compared with just$11 billion in Europe.60 This gap is even more striking when looking specifically at investments in gen AI.In 2023,US private 58 Eric Lamarre,Kate Smaje,and Rodney Zemmel,“Rewired to outcompete,”McKinsey Quarterly,June 20,2023.59 Dan Milmo,“Meta pulls plug on release

123、 of advanced AI model in EU,”Guardian,July 18,2024.60 Artificial Intelligence Index report 2024,Stanford University,2024;Quid data,accessed September 2024.61 PitchBook data,accessed September 2024.62 Massimo Giordano,Sven Smit,Jan Mischke,Guillaume Dagorret,Fredrik Dahlqvist,Sylvain Johansson,Marc-A

124、ntoine de la Chevasnerie,Solveigh Hieronimus,and Pieter Ottink,“Investment:Taking the pulse of European competitiveness,”MGI,June 20,2024.investments in gen AI amounted to$23 billion,compared with less than$2 billion in Europe.61 MGI published an article on strategies to enhance European investment.

125、62 Among the proposed solutions:public investment and precommercial innovation procurement vehicles at the EU levelsay,by dedicating public funding for AI-based diagnostics and treatment in the European Unions overall healthcare budget.Aims like these could be achieved by leveraging existing investm

126、ent vehicles,such as the InvestEU Fund,with an overall aspiration to dedicate 0.1 percent of European GDP a year in public investment to build gen AI infrastructure.European policy makers could also boost private capital investment through higher private equity and venture capital allocations by pen

127、sion funds and insurers.(This could require consolidation of these funds and regulation changes to allow the shifting of fund allocations.)Leapfrog in semiconductors.Per McKinsey research,Europe has negligible market share in the design and manufacturing of chips smaller than seven nanometers for AI

128、.A first step to addressing this issue could be to boost local expertise by attracting more R&D centers in the region.A second step could consist of leapfrogging in other promising semiconductor design fields,such as analog-,neuromorphic-,optical-,and quantum-computing semiconductors.European policy

129、 makers could also consider measures to attract advanced-semiconductor-manufacturing capacity in the region(for example,financial incentives and a fast-track permitting process).They could leverage recent momentum illustrated by multibillion-dollar investments across Europe in semiconductor fab manu

130、facturing 13Time to place our bets:Europes AI opportunity(including a state-of-the art fab in Germany).63 The European Chips Act,signed in 2023 with a funding commitment exceeding 40billion,represents an important initial step in strengthening Europes semiconductor industry,although it addresses a b

131、road spectrum of chip technologies rather than exclusively AI-specific chips.Improve local computing power and resources.In June 2024,Mistral AI executives reported a lack of hosting and computing capacity in Europe to develop and scale AI.64 Indeed,as of 2023,Europe hosts 18 percent of global data

132、center capacity(less than 5 percent owned by European companies),compared with 37percent for the United States.65 This deficit could deepen in the future,as McKinsey expects global data center demand to grow 22 percent per annum by 2030.To level the playing field for computing power,EU policy makers

133、 may want to explore ways to make Europe more competitive at hosting cloud facilities on the continent.This might include targeted incentives and support to enhance the mix of local companies and global players,as well as modernizing the underlying energy infrastructure.Foster AI models and applicat

134、ions that cater to local specificities.LLMs are currently predominantly trained on English data.66 However,Europe is a diverse market with different languages and cultures.By harnessing region-specific data,local players can create differentiated models and applications suited for a regions specific

135、 needs.Additionally,local expertise provides a significant advantage in navigating the specificities of EU regulations.63 For more,see“TSMC,Bosch,Infineon,and NXP establish joint venture to bring advanced semiconductor manufacturing to Europe,”Taiwan Semiconductor Manufacturing press release,August

136、8,2023.64 Cynthia Kroet,“Mistral AI warns of lack of data centres and training capacity in Europe,”Euronews,June 14,2024.65 Global Worldwide Semiannual Public Cloud Services Tracker,IDC Global,updated second half of 2023;McKinsey research and analysis.66 As of September 2024,around 50 percent of web

137、sites are in English,compared with only 5 percent in German and 4 percent in French.“Usage statistics of content languages for websites,”accessed September 2024.67 For more,see Markus Schlde,Xavier Veillard,and Alexander Weiss,“Four themes shaping the future of the stormy European power market,”McKi

138、nsey,January 27,2023.68 For more,see Frank Neffke,Martin Henning,and Ron Boschma,“How do regions diversify over time?Industry relatedness and the development of new growth paths in regions,”Economic Geography,July 2011,Volume 87,Number 3;Jing Xiao,Ron Boschma,and Martin Andersson,“Industrial diversi

139、fication in Europe:The differentiated role of relatedness,”Economic Geography,2018,Volume 94,Number 5.Energy capacity for gen AI in EuropeRegarding gen AI energy,policy makers can strive to ensure sufficient and affordable dispatchable power for data centers while staying committed to Europes climat

140、e-driven decarbonization goals.Addressing these challenges requires considering local variations in electricity demand and supply,such as the presence of energy-intensive industries and levels of energy independence.In addition to expanding low-carbon electricity infrastructure and streamlining perm

141、itting,Europe might also explore redesigning its power market,potentially mutualizing electricity purchases through a single EU or regulatory agency and establishing separate markets for zero marginal cost(wind and solar)and marginal cost resources.67When it comes to unlocking the full potential of

142、gen AI,Europe sits at a crossroads.Given the technologys novelty,the adoption race remains wide open.Europe has numerous opportunities to tactically reinforce its positions along the value chain while ensuring that it guides gen AI development by ethical considerations.Policy makers must understand

143、that the stakes here are considerable and extend beyond immediate economic impacts.Europes participation in the current AI boom is important not merely for todays gains but also to secure a foothold in future technological advances.History,after all,can point to examples of the snowball effect of te

144、chnology,in which pioneering innovations typically emerge from existing industries with related capabilities.6814Time to place our bets:Europes AI opportunityDesigned by McKinsey Global Publishing Copyright 2024 McKinsey&Company.All rights reserved.Alexander Sukharevsky,a global leader of QuantumBla

145、ck,AI by McKinsey,is a senior partner in McKinseys London office;Eric Hazan is a senior partner in the Paris office,where Marc-Antoine de la Chevasnerie is a partner and Guillaume Dagorret is a McKinsey Global Institute(MGI)senior fellow;Sven Smit,the chair of MGI,is a senior partner in the Amsterda

146、m office,where Marc de Jong is a senior partner;Solveigh Hieronimus is a senior partner in the Munich office;and Jan Mischke is an MGI partner in the Zurich office.The authors wish to thank Adrien Fresko,Arjita Bhan,Arnaud Tournesac,Christoph Sohns,Gardar Bjrnsson Rova,Klaus Pototzky,Michael Chui,Os

147、kar Harmsen,and Pieter Ottink for their contributions to this article.This article was edited by Larry Kanter,a senior editor in the New York office.Scan Download PersonalizeFind more content like this on the McKinsey Insights AppEuropean leaders can rely on the continents strong economic fundamenta

148、ls to elevate their gen AI ambitions.The region boasts a vast market of 500 million people,a strong industrial ecosystem comprising world-leading companies,69 a world-69 In 2023,Europe is home to 118 of the 500 largest companies around the world,including LVMH,Nestl,and Volkswagen.Fortune Global 500,Fortune,accessed September 2024.class talent pool,and an edge in clean energy.All of this provides a strong foundation on which to build an AI infrastructure and turn a strong present position into a leading role in the future.15Time to place our bets:Europes AI opportunity