《凯雷(Carlyle):2023勇敢的新世界:人工智能及其下游产业的影响力研究报告(英文版)(16页).pdf》由会员分享,可在线阅读,更多相关《凯雷(Carlyle):2023勇敢的新世界:人工智能及其下游产业的影响力研究报告(英文版)(16页).pdf(16页珍藏版)》请在三个皮匠报告上搜索。
1、October 2023Global InsightsBRAVE NEW WORLDAI and its Downstream Implications2 The advent of Artificial Intelligence may represent a watershed in human history,with the potential to transform daily lives to an extent that may be difficult to appreciate fully at this moment in time.But as unprecedente
2、d as the technological shock from Generative AI may prove to be,the capital market response to it already follows familiar patterns.Rather than simply separate reality from hype,successful investors must be able to map that reality onto company fundamentals.This rewards second-and-third order thinki
3、ng,as the most salient feature of the technological revolution escalating revenue growth at companies at the epicenter of the technological quake may ultimately prove to be a small fraction of the total economic value it delivers.As with the advent of electrification a turning point to which the dev
4、elopment of AI systems has been compared the main risk for investors today may be viewing the AI revolution too narrowly.The productivity gains from investment in software development and life sciences,content generation,and CRM systems already suggest that the assets best positioned to benefit from
5、 AI may have not yet landed on the broader markets radar.EXECUTIVE SUMMARY3Figure 1.Source:IDC,Tractica,Grand View Research,Statista,GlobeNewswire,Jefferies Equity Research.It is difficult to overstate the transformation potential of Artificial Intelligence(AI).We may soon live in a world where comp
6、uter systems can generate new scientific knowledge and perform virtually any human task.As unprecedented as the technological shock may prove to be,the capital market response to it already follows familiar patterns.When a foundational technology enters the publics consciousness,investors naturally
7、focus on the technology itself and companies thought to be operating at its frontier.Generative AI has been no exception.Asset prices quickly reach levels difficult to rationalize using conventional financial metrics;“value”comes to be associated with subjective impressions of the technologys potent
8、ial,barriers to entry,and ultimate scalability.Debates regarding the valuation of nascent technology often degrade on two axes.Enthusiasts,typically from the tech sector itself,recast investor skepticism as ignorance;an unwillingness to deploy aggressively into the space reveals a lack of technical
9、understanding.Detractors,for their part,often dismiss novel valuation methods and optimistic“total addressable market”forecasts(Figure 1)as tell-tale signs of a hype campaign designed to separate credulous investors from their capital.Portfolios can be derided as uninformed or nave,depending on pers
10、pective.Such discussions elide a crucial point.While dismissing AIs transformational potential could prove to be a very expensive mistake,returns ultimately depend on how new technology gets adopted and monetized.And this process can occur over long horizons and manifest on income statements some di
11、stance away from the initial shock.As with the advent of electrification a turning point to which the development of AI systems has been compared the main risk for investors today may be viewing the AI revolution too narrowly and failing to perceive all of the downstream opportunities(and risks)it c
12、reates.Figure 1.AI Market Size Expectations($Billions)$0$200$100$300$500$700$900$1,100$1,300$1,500$400$600$800$1,000$1,200$1,400$1,600200224202520262027202820292030IDC(February 2021)Tractica(March 2020)Grand View Research(June 2020)Figure 1:AI Market Size Expectations($Trillion
13、s)Globe Newswire(June 2022)4GROUNDBREAKING CAPABILITIES&ADOPTION RATES Investor interest in“artificial intelligence”has spiked over the past year thanks to the release of Generative AI tools capable of producing content and analyses of unprecedented sophistication and breadth in response to natural
14、language prompts.Most notable has been OpenAIs release of ChatGPT,which reached 100 million users in just two months,a small fraction of the time it took Facebook and other social media platforms to achieve similar scale(Figure 2).These models can reason probabilistically,have been trained on virtua
15、lly the entire internet corpus,and can be directed to process that information through conventional text that one might otherwise put into an email(not arcane code).Generative AI already represents an historic technological leap,at least as meaningful as internet-based search engines displacement of
16、 reference libraries.But whereas that revolution liberated information from the physical constraints of the analog world,AI liberates information flows from human intermediation.Machine Learning algorithms demonstrated softwares capacity to identify patterns in data and anticipate sequences faster a
17、nd more precisely than humans.Generative AI represents the next step in this evolution,with software now able to synthesize data and curate responses beyond those directly intended by the programmer(Figure 3,p.5).And there is still ample opportunity to reinvent the language tools that help engineers
18、 develop new generations of software even more efficiently.1One notable subset of Generative AI is large language models(LLMs).Impressive as this class of deep-learning algorithm is,it represents but one step on a longer road to“Artificial General Intelligence”autonomous computer systems that can le
19、arn to perform virtually any task of scientific or economic value.While many AI researchers would argue that were on the cusp of this world-historical turning point,others contend that AGI may be decades away if its ever achieved at all.Much of the disagreement centers on arcane Cartesian questions
20、of self-awareness and mysteries surrounding the biochemistry of human consciousness and cognition.2 The more practical and economically relevant the definition,the closer to AGI we may be.Figure 2.Source:Visual Capitalist,February 2023.1.Developer Tools 2.0,”Sequoia Capital,March 2023.2.C.f.Landgreb
21、e,J.and B.Smith.Why Machines Will Never Rule the World.Routledge,2022.Figure 2:Time to 100 Million Users200811Years10Years8Years5Years4.5Years4Years3.5Years2.5Years9Months2Months2006200882009201020162022THE TIME IT TOOK FOR SELECTED ONLINE SERVICES TO REACH 100 MILLION USERSFigure 2.Time
22、to 100 Million Users53TRADE SECRET AND STRICTLY CONFIDENTIALProgrammer-readable codeInterpreter/CompilerData+NN architectureProgrammer-readable codeOptimizer/CompilerNatural-languagelike instructionsDataMachine ProgramStatistically-basedMachine ProgramProgrammer-readable codeNN architectureSOFTWARE
23、1.0SOFTWARE 2.0SOFTWARE 3.0AI agentMachine ProgramStatistically-basedMachine ProgramMachine ProgramProgrammer-readable codeInterpreter/CompilerFigure 3.Next Step in Evolution of SoftwareIMMEDIATE APPLICATIONS All major technology and software vendors are currently embedding Generative AI into their
24、stack.Desktop applications(email,word-processing,etc.),e-commerce,internet search,social media,and content consumption will all integrate AI functionality.Such efforts remain in a beta stage with limited visibility into monetization.But the user experience is likely to improve immeasurably across ea
25、ch of these dimensions,with significant scope for labor productivity gains from accelerated information gathering and idea and text generation(Figure 4,p.6).More consequential may be the evolution of business models and corporate strategy.Management teams could increasingly rely on AI to formulate m
26、arketing strategies and pricing decisions and diligence potential acquisition targets.Digital marketing is likely to become even more precisely tailored,both in terms of the content of advertising campaigns and the targeting of audiences most likely to act on them.AI will revolutionize customer rela
27、tions management(CRM)across industries,generating upselling proposals in real time based on text from the conversation cross-referenced with internal customer data,external market trends,and other relevant information.Chatbots may soon account for the bulk of consumer-facing interactions in travel,f
28、inance,and e-commerce and eventually guide customers entire shopping experience.The applications for media and education are obvious.Generative AI applications can produce new music,fictional narratives,poetry,visual artwork,and digital imagery.The recent Screen Actors Guild(SAG)and Writers Guild of
29、 America(WGA)strikes have been fomented,in part,by concerns about AIs displacement potential.AI-generated content raises novel copyright issues since existing works are accessed to produce Figure 3.Source:Itamar Friedman,Software 3.0 the era of intelligent software development,May 2022.6“substantial
30、ly similar”outputs.3 Technologically,the horse is out of the barn;the question is whether owners of existing copyrights will be the only ones legally sanctioned to employ AI to assist in the formulation,production,and marketing of cinematic,televisual,and audio works.ChatGPT easily passed the Unifor
31、m Bar Examination taken by U.S.law school graduates and would earn a respectable 3.4 grade point average(on a 4-point scale)if enrolled as a freshman at Harvard College.4 Generative AIs prowess writing essays and taking tests raise thorny issues about the future of educational integrity,but also ope
32、n the door to a new generation of digital tutors,autodidacts,and more flexible educational arrangements.Huge productivity gains are already evident in software development,where Generative AI has halved the time necessary to write and test new code(Figure 5,p.7).LLMs can predict the next lines of co
33、de based on the code already written and generate new code in response to tailored prompts from software engineers who are skilled in natural language describing software structures.As LLMs become familiar with the functionality and structure of programming languages,prompts can become less precise,
34、allowing neophytes to code like seasoned professionals.5 While guidance from experienced engineers is fundamental to enable LLMs to write code,LLMs create significant efficiencies by filling in coding gaps in simplified prompts.Eventual gains from such automation may be especially pronounced among v
35、ideo game makers operating at the intersection of AI-generated content and software.Companies will increasingly rely on Generative AI to clean existing data and produce prototype designs and accelerate product development.Life sciences companies,for instance,already use AI to generate sequences of a
36、mino acids and DNA nucleotides to shorten the drug design phase from months to weeks.Existing development programs require researchers to sort through millions of potential chemical reactions to synthesize a target molecule.AI models trained on existing chemical reactions data have already yielded a
37、 15%reduction in development costs.6 We should expect to see comparable productivity gains wherever R&D depends on time-consuming,iterative processes based on complex interactions between variables or inputs.Manufacturers can not only use Generative AI to design new products,but also optimize supply
38、 chains and automate shipping and production processes.The automotive industry has been especially aggressive in its adoption of AI and antecedent algorithmic technologies to these ends.Figure 4.Source:Carlyle Analysis,2023.3.ABA Journal,March 2023.“ChatGPT goes to Harvard,”Substack,July 2023.4.“Bey
39、ond The Hype:How Generative AI Is Transforming Software Development,”Towards Data Science,May 2023.5.G2Retro as a two-step graph generative models for retrosynthesis prediction,Communications Chemistry,May 2023.6.G2Retro as a two-step graph generative models for retrosynthesis prediction,Communicati
40、ons Chemistry,May 2023.Figure 4.AI Use Cases4TRADE SECRET AND STRICTLY CONFIDENTIALCONTENT GENERATIONSOFTWARE DEVELOPMENTIMAGEGENERATIONNEW PRODUCTDEVELOPMENTSALES&MARKETINGQ&AINTERFACES Content SEO Primary research Synthesis Alert generation Support ticketing systems Language translation Create web
41、site drafts Automatic code generation CoPilots Regenerative code Test script generation Bug fixes Custom generated photos Image touch up Banner creation Medical imaging Product detail page image generation“Try it on”AR Interactive data products Conversational interface&querying Whitelabeled 1st part
42、y trained models UX design Translation from design to code Content creation Lead generation Sales forecasting Personalized ads Org specific sales collateral Customer support Conversion rate optimization A/B testing R&D idea generation Identity verification Order taking Advanced Chatbots Disaster pla
43、nning&recovery Strategy development Competitor research7Figure 5.Accelerated Software DevelopmentRISKS&JOB LOSS The Janus face of new technology is obsolescence.It is estimated that Generative AI applications could eventually automate 60%to 70%of employee workloads,7 and this naturally arouses fear
44、of job loss.It is important to note that this estimate refers to employee tasks not the employees themselves.For most occupations,we subscribe to the view that AI wont take your job;someone using AI will.This will result in dynamic adjustments in labor demand across occupations and activities rather
45、 than job loss(Figure 6,p.8).Workload automation should increase throughput volumes,naturally increasing productivity levels(output per hour of work);and by freeing managers finite time and attention and speeding more junior employees progression up the learning curve,AI also could facilitate a sust
46、ained increase in productivity growth rates as human capital gets deployed more creatively(Figure 7,p.9).Obsolescence may be of greater concern for businesses and business models,as competition increasingly depends on the speed with which companies adopt AI capabilities to cut costs and increase sca
47、lability.Competitive pressure this great naturally opens the door to charlatanism.Companies will market themselves opportunistically and,occasionally,deceptively.Mentions of“AI”on corporate earnings calls has risen exponentially(Figure 8,p.9),and the more“AI”is invoked by competitors,the more suscep
48、tible laggard management teams become to imprudent budgeting and fairy-tale solutions.We must also be mindful of the“hallucination problem”with LLMs,or their tendency to generate factually incorrect text that may seem semantically or syntactically plausible based on the corpus of data on which it ha
49、s been trained.These statistical models predict the next word based on massive volumes of data and past context.They are built for fluency rather than reason,which means human verification of their outputs will still be required in many cases,and their use in mission critical applications like aeron
50、autics or defense could lay very far in the future.Figure 5.Source:McKinsey,2023.7.“Economic Potential of Generative AI,”McKinsey&Co.June 2023.With generative AIWithout generative AI01020303545Code documentationCode generationCode refactoringHigh-complexity tasks4550TASK COMPLETION TIME U
51、SING GENERATIVE AI,%8Figure 6.Source:https:/ labor demand change and generative AIautomation acceleration by occupation,US,202230 Increase in automation adoption driven by generative AI acceleration,percentage points Changein labordemand,%152525355M10M3540Midpoint automationadoption by 2030,%Employm
52、ent,absoluteAgricultureBuildersBusiness and legal professionalsEducation and workforce training2020253035Mechanical installation and repairTransportation servicesFood servicesProduction workHealth aides,technicians,and wellnessHealth professionalsProperty maintenanceOfcesupportCustomer se
53、rvice and salesCreatives and arts managementSTEM professionalsManagersCommunity services101520Increasing labor demandand high change of work activities Decreasing labor demandand modest change of work activities Increasing labor demandand modest change of work activities Figure 6.Dynamic Adjustment
54、in Labor Demand9Figure 7.Economy-Wide Positive Productivity ShockFigure 7.Source:Carlyle Analysis,Brookings Institution,2023.Figure 8.Note:Includes mentions of“AI”in analyst/journalist questions.Source:Company data,Statista,Goldman Sachs Global Investment Research.Figure 8.Mentions of AI on Company
55、Earnings Calls7TRADE SECRET AND STRICTLY CONFIDENTIAL75%95%115%135%155%175%195%215%202320242025202620272028202920302033420352036203720382039204020412042Productivity Relative to 2022 BaselineBaselineOne-Time Generative AI ShockPersistent Acceleration8TRADE SECRET AND STRICTLY CONFIDENTIAL3
56、87265352425060708090NvidiaAlphabetMetaMicrosoftSalesforceAMDAmazonQ1 2022Q1 202310BARRIERS TO ENTRY At this stage,most of the market discourse has focused on those companies directly responsible for the development of LLMs.And,given the enormous costs involved,this has been and
57、 is likely to continue to be dominated by massive,cash-rich incumbents.Developing a state-of-the-art Generative AI model requires massive computational resources,specialized hardware like Graphics Processing Units(GPUs)and Tensor Processing Units(TPUs),and vast datasets that must be collected,stored
58、,and curated.A single training run for a model comparable to ChatGPT requires millions of dollars.8 Rather than compete with better funded and more sophisticated incumbents,enterprises seeking to integrate AI into their products and services are more likely to partner with them.This has led to a boo
59、m in the market values of industry-leading hardware,software,and data cloud platforms(Figure 9)including a$700 billion increase in Nvidias market capitalization since ChatGPTs release and creates significant headwinds for new entrants and small companies across much of the value chain.This has not s
60、topped capital from flowing to newer and younger companies,however.Over the past year,virtually any asset with known“AI upside”has become very richly valued,especially on a relative basis(Figure 10,p.11).While all industries have been affected by the decline in venture and growth capital over the pa
61、st year,AI companies have captured a larger share of that funding,especially those focused on novel approaches to AGI.In the U.S.,AIs share of funding rounds reached 23%in Q2-2023,more than tripling over the past 10 years and now the highest among all industry verticals(Figure 11,p 11).In terms of i
62、nvested capital,AIs share has increased even more over the past year thanks,in large part,to Microsofts$10 billion investment in OpenAI and Stripes$6.3 billion Series I round.9Figure 9.Source:Carlyle Analysis of Bloomberg Data,July 21,2023.8.“Can You Build Large Language Models Like ChatGPT At Half
63、Cost?”UniteAI,May 2023.9.Global Private Markets Quarterly Q2-2023,Carlyle Global Investment Solutions,July 2023.Figure 9.MegaCap AI Companies Share of Total Returns9TRADE SECRET AND STRICTLY CONFIDENTIALBREAKDOWN OF RETURN BY COMPANYSHARE OF RETURN BY COMPANY15.7%12.7%6.3%7.9%7.2%7.2%12.1%69.2%30.8%
64、0%10%20%30%40%50%60%70%80%90%100%AppleMicrosoftAlphabetAmazonTeslaMetaNVIDIAS&P 500Share of S&P 500 ReturnTop 7 Stocks Total Share2.9%2.3%1.1%1.4%1.3%1.3%2.2%12.6%5.6%0%2%4%6%8%10%12%14%16%18%20%AppleMicrosoftAlphabetAmazonTeslaMetaNVIDIAS&P 500Contribution to S&P 500 Return in Percentage PointsTop
65、7 Stocks PP Contribution11Figure 10.Rise in AI Attention&ValuationsFigure 10.SG AI Newsflow Indicator Continue to Surge Source:Factiva,SG Cross Asset Research/Equity Strategy.Data as of 08/05/2023.AI-Related Stocks Drove Virtually All the Returns of the S&P 500 This Year Source:Datastream,SG Cross A
66、sset/Research/Equity Strategy.Data as of 11/05/2023.Figure 11.Source:Carlyle Global Investment Solutions,Global Private Markets Quarterly,Q3-2023.Jan/15Aug/15Mar/16Oct/16May/17Dec/17Jul/18Feb/19Sept/19Apr/20Nov/20Jun/21Jan/22Aug/22Mar/234500400035003000250020000450040003500300025002000150
67、010005000Jan/23Feb/23Mar/23Apr/23May/234200438003700360042004380037003600S&P ex-Al Boom stocksS&P 500SG AI NEWSFLOW INDICATOR CONTINUE TO SURGEAI-RELATED STOCKS DROVE VIRTUALLY ALL THE RETURNS OF THE S&P 500 THIS YEAR Figure 11.AIs Increasing Share of VC Funding2014 H12014 H220
68、15 HI2015 H22016 H12016 H22017 H12017 H22018 H12018 H22019 H12019 H22020 H12020 H22021 H12021 H22022 H12022 H22013 H16,0005,0004,0003,0002,0001,0000Number of VC&Growth Capital Funding Rounds30%25%20%15%10%5%0%AI Funding Rounds in%of Total RoundsNon-AIAIAI Share12LESSONS FROM ELECTRIFICATION One wond
69、ers if by focusing narrowly on the assets closest to the epicenter of this technological quake,investors may be repeating the mistakes of the past.Generative AI has been analogized to the advent of electricity,and this comparison may be apt for reasons that extend well beyond its technological signi
70、ficance.Though discovered in the 1880s,electric current only began to transform society in the 1920s when mass electrification was made possible by high-pressure steam power plants and centralized generation,distribution,and system management.In just a few years,electric companies revenues grew by m
71、ore than 3.4x(35%CAGR)during a period of consumer price deflation.The valuations assigned to those fundamentals doubled during this time(Figure 12,p.13),as investors aggressively bid up the market values of companies operating at the frontier of this technological revolution.As it turned out,far mor
72、e economic value was being created by the companies buying that power.Electrification allowed manufacturers to use a large number of complex machines simultaneously,which made mass production processes possible and sharply reduced the cost of producing consumer durables like refrigerators,washing ma
73、chines,and radios(Figure 13,p.13).And since these products had to be plugged in to operate,mass electrification not only drove down manufacturers production costs,but also stimulated demand for their products.In the ten years from the start of the sustained boom in electricity generation,durable goo
74、ds manufacturers generated a 200%total return,on average,in the depths of the Great Depression(!),which was more than 2x the average total return to electric companies over the same period(Figure 14,p.14).No sane person could contend that mass electrification was mere“hype,”as eventual market demand
75、 for electricity met or exceeded the most optimistic forecasts.But the displacement of kerosene-fired illumination was but the tip of the iceberg,as the vast majority of the economic value accrued to the downstream users of the new technology rather than the companies responsible for its introductio
76、n.The same dynamics are likely at play today with Generative AI.Specialized semiconductor sales may indeed go through the roof,just as demand for the most advanced boilers rose exponentially during the period of mass electrification.A step-function increase in the volume of data generated,stored,and
77、 analyzed by companies will almost surely benefit cloud platforms just as a comparable jump in the regional transmission of electric current benefited electric utilities.Future growth in the utility sector will require significant investment in Generative AI to support power grid development.And com
78、panies at the forefront of the design of advanced AI systems today will likely be as influential to economic development as those responsible for developing the latest iteration of high-pressure steam turbines then.But the bulk of the economic value may,once again,be created by the companies most ad
79、ept at capitalizing on these trends by slashing production costs and developing the new products and services made possible by these new technologies.This is likely to be especially true in software,pharmaceuticals,and other sectors where Generative AI can reduce the enormous sums spent developing i
80、ntangible assets that can be infinitely reproduced at nearly zero marginal cost.But the bulk of the economic value may,once again,be created by the companies most adept at capitalizing on these trends by slashing production costs and developing the new products and services made possible by these ne
81、w technologies.13Figure 13.Two-Year Decline in Production Costs by ItemFigure 12.Source:Carlyle Analysis;CRSP Database,December 2021.Figure 13.Source:Ronald C.Tobey,1997,“Technology as Freedom:The New Deal and the Electrical Modernization of the American Home.”0%-10%-20%-30%-40%-50%-60%-70%-80%Cofee
82、Maker-69%ElectricBlanketRadioFanCookerWasherToasterRefrigeratorFlatironRangePeak Two Year Price Decline,1926-1936-58%-54%-52%-50%-44%-41%-34%-28%-17%RetailOil&GasDurable GoodsRailroad/OtherConsumer ProdTelecomIndustrialsTechHealth CareUtilities150.0%100.0%50.0%0.0%-50.0%95%30%Figure 12.Rise in Valua
83、tion Ratios,1925-2914Jul-26Oct-26Jan-27Apr-27Jul-27Oct-27Jan-28Apr-28Jul-28Oct-28Jan-29Apr-29Jul-29Oct-29Jan-30Apr-30Jul-30Oct-30Jan-31Apr-31Jul-31Oct-31Jan-32Apr-32Jul-32Oct-32Jan-33Apr-33Jul-33Oct-33Jan-34Apr-34Jul-34Oct-34Jan-35Apr-35Jul-35Oct-35Jan-36Apr-364.5x4.0 x3.5x3.0 x2.5x2.0 x1.5x1.0 x0.5
84、x0.0 xCumulative MOICMarket AverageUtilitiesDurable Goods3.02x1.96x1.37xFigure 14.Total Stock Market Returns by SectorFigure 14.Source:Carlyle Analysis;CRSP Database,December 2021.PLAYING THE LONG(ER)GAME Rather than simply separating reality from hype,successful investors must be able to map that r
85、eality onto company fundamentals.This rewards second-and-third order thinking,as the most salient feature of the technological revolution escalating revenue growth at companies at the epicenter of the technological quake may ultimately prove to be a small fraction of the total economic value it deli
86、vers.One companys revenue is anothers investment.And the productivity gains from investment in software development and life sciences,content generation,and CRM systems already suggest that the assets best positioned to benefit from AI may have not yet landed on the broader markets radar.15Jason Tho
87、masJason Thomas is the Head of Global Research&Investment Strategy at Carlyle,focusing on economic and statistical analysis of Carlyle portfolio data,asset prices,and broader trends in the global economy.Prior to joining Carlyle,Mr.Thomas served on the White House staff as Special Assistant to the P
88、resident and Director for Policy Development at the National Economic Council.In this capacity,Mr.Thomas acted as the primary adviser to the President for public finance.HEAD OF GLOBAL RESEARCH&INVESTMENT STRATEGY received a BA from Claremont McKenna College and an MS and PhD in finance from George
89、Washington University,where he studied as a Bank of America Foundation,Leo and Lillian Goodwin Foundation,and School of Business Fellow.Mr.Thomas has earned the chartered financial analyst designation and is a Financial Risk Manager certified by the Global Association of Risk Professionals.Economic
90、and market views and forecasts reflect our judgment as of the date of this presentation and are subject to change without notice.In particular,forecasts are estimated,based on assumptions,and may change materially as economic and market conditions change.The Carlyle Group has no obligation to provid
91、e updates or changes to these forecasts.Certain information contained herein has been obtained from sources prepared by other parties,which in certain cases have not been updated through the date hereof.While such information is believed to be reliable for the purpose used herein,The Carlyle Group a
92、nd its affiliates assume no responsibility for the accuracy,completeness or fairness of such information.References to particular portfolio companies are not intended as,and should not be construed as,recommendations for any particular company,investment,or security.The investments described herein
93、were not made by a single investment fund or other product and do not represent all of the investments purchased or sold by any fund or product.This material should not be construed as an offer to sell or the solicitation of an offer to buy any security in any jurisdiction where such an offer or sol
94、icitation would be illegal.We are not soliciting any action based on this material.It is for the general information of clients of The Carlyle Group.It does not constitute a personal recommendation or take into account the particular investment objectives,financial situations,or needs of individual
95、investors.16Economic and market views and forecasts reflect our judgment as of the date of this presentation and are subject to change without notice.In particular,forecasts are estimated,based on assumptions,and may change materially as economic and market conditions change.The Carlyle Group has no
96、 obligation to provide updates or changes to these forecasts.Certain information contained herein has been obtained from sources prepared by other parties,which in certain cases have not been updated through the date hereof.While such information is believed to be reliable for the purpose used herei
97、n,The Carlyle Group and its affiliates assume no responsibility for the accuracy,completeness or fairness of such information.References to particular portfolio companies are not intended as,and should not be construed as,recommendations for any particular company,investment,or security.The investme
98、nts described herein were not made by a single investment fund or other product and do not represent all of the investments purchased or sold by any fund or product.This material should not be construed as an offer to sell or the solicitation of an offer to buy any security in any jurisdiction where
99、 such an offer or solicitation would be illegal.We are not soliciting any action based on this material.It is for the general information of clients of The Carlyle Group.It does not constitute a personal recommendation or take into account the particular investment objectives,financial situations,or
100、 needs of individual investors.Michael WandMichael Wand is a Managing Director and Co-Head of the Carlyle Europe Technology Partners investment advisory team.His main investment focus lies in enterprise and infrastructure software,media technology,IT security,data analytics and digital services.He i
101、s based in London.Mr.Wand has led numerous technology investments,notably P&I AG,FRS Global,UC4/Automic,Foundry,eggplant and Dept,and advised on a dozen completed realizations.Mr.Wand is currently a member of the Boards of Directors of Shopware,SER Group,HSO,Dept,Jagex,Disguise,LiveU and Incubeta.In
102、 addition,Mr.Wand is Co-Head of Europe Private Equity,sitting on Carlyles European Private Equity Executive MANAGING DIRECTOR AND CO-HEAD OF CARLYLE EUROPE TECHNOLOGY PARTNERS INVESTMENT ADVISORY TEAMCommittee which was formed in 2022 to drive deeper collaboration across all the firms Corporate Priv
103、ate Equity activities in the region.Prior to joining Carlyle in 2001,Mr.Wand worked for approximately 10 years in investment banking,the last seven years of which were in the technology sector.As a Managing Director,he was responsible for the European software and Internet research team at Deutsche Bank in London.Prior to that,he worked as a European Software Analyst for Paribas in London and BHF-Bank in Frankfurt,Germany.During his banking career,Mr.Wand was the underwriting analyst for European technology innovators such as SAP,Autonomy,SurfControl,Utimaco and nCipher.