全球化研究报告-PDF版

华利安：2025可持续发展软件行业更新报告（英文版）（45页）.pdf

October 2025Sustainability SoftwareSector Update010203Houlihan Lokey in Sustainability SoftwareDefin.

2025-11-05 45页 5星级

IBV：2025代理型AI的战略崛起：从渐进式收益到全新价值创造的运营范式转变报告（英文版）（24页）.pdf

IBM Institute for Business Value|Research InsightsAgentic AIs strategic ascentShifting operations fr.

2025-11-05 24页 5星级

奥纬咨询：2025 21世纪的“古登堡革命”：生成式AI如何重塑出版业研究报告（英文版）（14页）.pdf

THE 21ST CENTURY GUTENBERGHow Generative AI is rewriting publishings playbookCharles De PommerolRaul.

2025-11-05 14页 5星级

IMD：2025年全球数字竞争力排名报告（英文版）（122页）.pdf

2025IMDWorld Digital Competitiveness RankingReconfjguring digital strategies amid trade fragmentatio.

2025-11-05 122页 5星级

麦肯锡：2025欧洲深度科技引擎或可推动万亿美元经济增长研究报告（英文版）（17页）.pdf

Europes deep-tech engine could spur$1 trillion in economic growthOctober 2025Copyright 2025 McKinsey.

2025-11-05 17页 5星级

SlashData：2025年第三季度编程语言社区规模评估报告（英文版）（26页）.pdf

SIZING PROGRAMMING LANGUAGE COMMUNITIESHow many developers are using each programming language in Q3.

2025-11-04 26页 5星级

国际电信联盟（ITU）：2025人工智能向善：创新扩大影响案例集报告（英文版）（874页）.pdf

ITUPublicationsInternational Telecommunication UnionTelecommunication Standardization SectorAI for G.

2025-11-04 874页 5星级

SlashData：2025软件项目中代理型AI应用状况研究报告：开发者架构采用及目标用例（英文版）（25页）.pdf

THE STATE OF AGENTIC AI ADOPTION IN SOFTWARE PROJECTSThe architectures developers use,and the use ca.

2025-11-04 25页 5星级

SlashData：2025生成式AI应用集成障碍研究报告（英文版）（19页）.pdf

BARRIERS TO INTEGRATINGGENERATIVE AI IN APPLICATIONSUnderstanding what prevents developers from buil.

2025-11-04 19页 5星级

Cyber London：2025英国网络安全与人工智能全景报告：协同创新（英文版）（22页）.pdf

Innovation through CollaborationN O T-F O R-P R O F I T C Y B E R C L U S T E R R E C O G N I S E DB.

2025-11-04 22页 5星级

SlashData：2025后端与云技术策略基准研究报告（英文版）（28页）.pdf

BENCHMARKING BACKEND AND CLOUD TECHNOLOGY STRATEGIESHow company size and industry shape developers d.

2025-11-04 28页 5星级

Linux基金会&Canonical：2025年全球开源状况报告（英文版）（43页）.pdf

Marco Gerosa,PhD,Northern Arizona UniversityAdrienn Lawson,The Linux FoundationForeword by Jon Seage.

2025-11-04 43页 5星级

SlashData：2025 AI编程工具团队选型：专业开发者使用情况、满意度与信任度研究报告（英文版）（25页）.pdf

CHOOSING THE RIGHT AI CODING TOOL FOR YOUR TEAMUsage,satisfaction,and trust among professional devel.

2025-11-04 25页 5星级

防特网（Fortinet）：2025年全球网络安全技能差距研究报告（英文版）（42页）.pdf

2025 Cybersecurity Skills GapGlobal Research ReportContents 3Methodology 4ExecutiveSummary 7AIIsSeen.

2025-11-04 42页 5星级

安永：2025从AI到AX（AI体验）：下一代AI赋能服务的共情需求研究报告（英文版）（46页）.pdf

From AI to AX:Empathy Demand in the next generation of AI-powered servicesHuman SignalsOctober 2025C.

2025-11-04 46页 5星级

麻省理工学院：2025 ChatGPT对大脑的影响：AI助手辅助写作会累积“认知债务”研究报告（英文版）（206页）.pdf

Your Brain on ChatGPT:Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Ta.

2025-11-04 206页 5星级

SlashData：2025年FinOps状况报告：不同区域与企业规模下的开发者团队云成本管理（英文版）（21页.pdf

THE STATE OF FINOPS IN 2025How developer teams manage cloud costs across regions and company sizesOC.

2025-11-04 21页 5星级

Hiscox：2025年网络安全就绪度报告：中小企业采取行动应对当前及未来网络安全威胁（英文版）（16页）.pdf

CYBER READINESS REPORT 2025SMEs take action against current and future cyber security threatsContent.

2025-11-04 16页 5星级

波士顿咨询（BCG）：2025释放生成式AI在汽车行业的价值研究报告（英文版）（22页）.pdf

Executive PerspectivesOctober 2025AutomotiveUnlocking the Value of GenAI in Automotive2Copyright 202.

2025-11-04 22页 5星级

CB Insights：2025年15大技术趋势研究报告（英文版）（93页）.pdf

2025 Tech TrendsOUTLINE2Your all-in-oneAI Super Analystfor market intelligenceInstant insights to bet on the rightmarkets,stay ahead of competitors,and source the right companies.IIIIGet a demo3Tech trendsFinancial servicesThe cyborg wealth advisor04AI agents are given money to spend 12Crypto takes more baby steps toward the mainstream 19Compressed fintech valuations create opportunities for acquirers24Healthcare&life sciencesDisease management enters a new phase with AI30Investment floodgates open for RNA therapeutics35Autonomous robots eye caregiving40AIAI M&A fuels the next wave of corporate strategy45LLMs explainability moment51Open-source cedes top LLMs but dominates smaller models58The US is leading the AI arms race for now65EnterpriseSpatial computing for enterprises is here to stay71Retail&consumerRetails personalization imperative76IndustrialsThe future data center arrives81Cheaper access to space sparks investor rush864The cyborg wealth advisorFinancial services5Drowning in admin work,financial advisors spend less than a fifth of their time with clientsFinancial servicesHigher-earning advisors(those driving$1M in annual revenue)spend 26%of their time with clients.Thats over 50%more than the benchmark,or nearly4 hours more each week.Top performers eke out an additional 9%of their time for clients driving 200 additional 1-hour meetings per advisor per yearPercentage of time senior financial advisors dedicate to routine tasksExtremely valuable.Source:2022 Kitces study of US-based financial advisors,n=767.17%Time with clientsProspecting/marketing22ck-office tasks(admin,client servicing,investment mgmt.)48%Middle-office tasks(meeting prep,investment research,compliance,etc.)70%of advisors time is spent awayfrom clients/prospects6To get that time back,wealth giants like Morgan Stanley double down on AI for advisor productivityFinancial servicesMorgan Stanleys wealth management business saw record net revenue in Q324 on the heels of deeper AI integrationSource:CB Insights Morgan Stanley Q324 earnings insights;Morgan Stanley quarterly financial reporting.$6.6B$6.7B$6.4B$6.6B$6.9B$6.8BNet revenue$7.3B00.20.40.60.811.2$0.0B$1.0B$2.0B$3.0B$4.0B$5.0B$6.0B$7.0B$8.0BQ1Q2Q3Q4Q1Q2Q320232024 1 14%4%YoYMorgan Stanleys wealth management revenuesIGet insights on Morgan Stanleys recent earnings calls7But outside of the largest and most well-resourced firms,wealth managers are cautious about AIFinancial servicesWealth firms AI plans have been hamstrung by concerns around regulation and data privacy,as well as a lack of knowledge and technical skillsSource:Wipro;Financial Planning.34E%0%5 %05EP%Incremental/small-scaleimplementationStill learning&collectinginfo about AIEarly days:Nearly half of wealth managers are still learning about AI,while another third are implementing it incrementallySource:Financial Planning study,October 2024,n=270of firms see the lack of regulatory guidelines as a top obstacle to adopting AI62%Compan y ASource:Wipro study,September 2024,n=1008One of wealth management firms biggest opportunities with AI is to bring scale and speed to the human element of advisoryFinancial servicesAI can augment financial advisors capabilities at every point of the wealth value chainSource:CB Insights research The new wealth tech value chain report.The visual shows 15 technology markets gaining momentum across wealth management.Companies in these markets either develop AI solutions or enable wealth tech platform integration via APIs.The new wealth tech value chain:Open integration orchestrates software used across products and services AI personalizes how clients and advisors work together9Personalized client engagement is one of the most mature AI use cases in wealth managementFinancial servicesGenAI startups are automating marketing tasks like copywriting and video creationSource:CB Insights The generative AI in financial services market map.To identify players for this market map,we reviewed generative AI companies targeting horizontal applications and shortlisted those with funding momentum(received funding since the start of 2022 or with total funding of$30M ).We then filtered based on the industries the companies serve,selecting companies targeting financial services applications.IIIExploreDive into the data on marketing-focused genAI companies in financial services10For firms looking to build AI into client engagement,integration with existing tech stack is keyFinancial servicesEase of integration is a selling point,as shown by positive customer reactions to Nitrogens AI-powered client engagement and prospecting toolSource:CB Insights Wealth management prospecting and lead generation market report and Nitrogen buyer interview.IICompare companies in this market11Wealth managers can select from 50 AI-focused wealth tech partners that have already validated their productsSource:CB Insights Commercial Maturity scores(measuring a companys ability to compete for customers or serve as a partner).Analysis includes B2B startups with Commercial Maturity scores that appear in both the Wealth Tech and AI Expert Collections.Data as of 11/7/2024.The majority of B2B AI wealth techs are growing client distribution,signaling commercial readinessFinancial servicesCommercial Maturity of B2B wealth tech startups Commercial Maturity of B2B wealth tech startups with AI offeringswith AI offerings00S%3%0 0P%5:Established4:Scaling3:Deploying2:Validating1:EmergingPotential partners in the Deploying stage or later with Potential partners in the Deploying stage or later with high Mosaic scores(measuring business health)high Mosaic scores(measuring business health)IIIExplore the data12AI agents are given money to spendFinancial services13AI agents are emerging across the software landscape,but theyre utility is limited until they can make transactions seamlesslyNote:AI agents refer to LLM-based systems that can independently reason and execute tasks.Private companies that are primarily focused on AI agent infrastructure applications only visualized.Financial servicesPrivate companies building the infrastructure for AI agents proliferate InfrastructureIIIIIExplore AI agent startups Developer frameworks Multi-agent&orchestrationAuthentication&trustVoiceLow-code or no-code developmentPaymentsNascent space with few private players14Current payments infrastructure is designed for humans and keeping bots outSource:StripeFinancial servicesOnline payments flow todayCardholderMerchantAcquirerIssuing bankPayment gatewayCard network15Crypto emerges as the first AI payment railSource:CB Insights Skyfire Scouting Report,key people data;media mentions.Early movers Skyfire and Coinbase target agent-to-agent transactions to circumvent human identify verification required for bank accounts and credit cardsFinancial servicesSelect Skyfire investorsSelect Skyfire investorsKey peoplePrevious experienceAmir SarhangiCEO&co-founder VP of Product and Services Ripple Founder Jibe,acquired by Google(2015)pioneer of RCS messaging protocol now standard on AndroidCraig DeWittCo-founder Senior Director of Product RippleOctober 2024 Coinbase launches capability to create AI agents with crypto wallet Skyfire announces integration with Base blockchain,enabling agents to use Coinbase wallets16Stripe launches developer tool to enable agents to spend funds Source:X;Stripe.Financial servicesStripes and Coinbases entry signals a strategic push from established players to lead the emerging agent-enabled payments ecosystemStripe Issuing allows developers to generate single-use virtual cards with spend controls that agents can use to make purchases17Source:Company releases;media mentions.Financial servicesAI agents are on the roadmap including autonomous AI shopping agentsEventually,“you could maybe give it a budget.”Trishul Chilimbi,VP&distinguished scientist at Amazon|WIRED(October 2024)At Google I/O 2024,CEO Sundar Pichai showcased Gemini performing a shopping return end-to-end(May 2024)Exploring computer using-agents that can handle personal data like credit card info to accomplish tasks|The Information(October 2024)Big tech envisions the future of agent commerce,but earning user trust will be key18Agent trust layer will emerge as the first big market opportunitySource:CB Insights company profiles Okta,Skyfire,Anon,Ping Identity;company releases.Before autonomous spending can scale,watch for identity verification and access management players to build the“drivers license”for AI agentsFinancial servicesCategoryCompanyStatusApplicationAuthenticationSeed(December 2023)Developer toolkit enabling agents to authenticate on sites without APIs while protecting the users credentialsIdentityPublicly tradedAuth for GenAI product for developers to provide identity to agents,enabling agents to call APIs on a users behalf and receive async authentication from the user IdentityTaken private(October 2022)Helix product provides AI agents with identities,enabling organizations to authorize and authenticate themIdentity/payments networkSeed(October 2024)Building”AgentID”that assigns agents a digital wallet with unique identifier&verification service“For instance,if someone directs an AI agent to book flightswhen the price goes below$200,Async Authentication can send a push notification to the user on the travel app to authenticate on demand and approve the transaction.”Okta press release(October 2024)19Crypto takes more baby steps toward the mainstreamFinancial services20Crypto has returned to the conversation for executivesSource:CB Insights Earnings transcript analytics.Quarterly mentions of crypto and related terms on earnings calls(as of Q324)Financial services1,271IIIExplore the trend with CBI earnings transcript analytics21Payments leaders remain crypto-curiousSource:CB Insights research Crypto is showing signs of life in payments.Note:Per category,payments leaders selected based on market capitalization/valuation and transaction volume.*Block announced in November 2024 it would shut down TBD its Bitcoin-focused business to focus on Bitcoin mining and its crypto wallet.Financial servicesIndicates at least one new strategic partnership,acquisition,or startup investment in digital currencies or blockchain in 2024 YTDPayment NetworksPartnered with Transak to enable crypto-to-fiat conversions(January 2024)Partnered with MetaMask and Baanx to launch crypto-to-fiat card(August 2024)Partnered with Ripple to integrate$XRP digital currency(August 2024)PayPal Ventures invested in crypto services company Mesh(January 2024)Invested in Africa-focused crypto exchange Yellow Cards Series C(October 2024)*Financial InstitutionsInvested in HQLAx,a blockchain-based platform for collateral mobility(April 2024)Tech&Regional PlayersAcquired stablecoin platform Bridge for$1.1B(October 2024)Invested in$15M Series A for Chainbase,a Web3 data infrastructure company(July 2024)Samsung NEXT backed blockchain company Pi Squared(July 2024)Merchant Acquirers22$1,100M$706M$400M$320M$300M$250M$250M$200M$200M$197MStablecoins are a bright spot,highlighted by Stripes$1.1B acquisition of stablecoin payments platform BridgeBridges exit is the largest crypto acquisition(by exit valuation)everFinancial servicesAcquired byAcquired by*BK Global ConsortiumDigitalAssetsAcquired by Acquired by October 2024October 2018February 2018March 2022October 2021Acquired by*BRIDGEAcquired by May 2023Acquired by May 2021Acquired by March 2021Acquired by*November 2022Acquired byJanuary 2022Largest crypto acquisitions by exit valuationLargest crypto acquisitions by exit valuationSource:CB Insights acquisition data(as of 11/1/2024);Visa.Note:First exits only.Excludes:SPAC mergers,bitcoin mining exits,and companies acquired in tokens.*Corporate majority.*Stock purchase.*As of April 2024,Apex has not met the financial targets required to achieve the full$200M deal value.23Despite the activity,the broader sector remains immature Source:CB Insights Commercial Maturity scores(measuring a companys ability to compete for customers or serve as a partner).Data as of 10/31/2024.Companies raising equity deals only.Blockchain startups raising funding are behind AI startups in deploying their products commercially indicating enterprises should vet not prioritize the marketFinancial servicesCommercial Maturity of Commercial Maturity of blockchaincompanies funded companies funded in 2024 YTDin 2024 YTD822%0%5 %05E%5:Established4:Scaling3:Deploying2:Validating1:Emerging16C!%0%5 %05E%5:Established4:Scaling3:Deploying2:Validating1:EmergingCommercial Maturity of Commercial Maturity of AIcompanies funded in companies funded in 2024 YTD2024 YTDVS.VS.IIIExplore the dataIIIExplore the data24Compressed fintech valuations create opportunities for acquirersFinancial services25Across VC,exit valuations on a per employee basis have deflated since 2021Financial servicesMedian VC-backed M&A valuation per employee down to$1.3M in 2024 from$1.7M in 2021Source:CB Insights data for nearly 1,500 VC-backed tech companies with disclosed valuations and employee headcounts at time of M&A exit.Data as of 10/29/2024.$1.2M$1.6M$1.7M$1.4M$1.4M$1.3M$0.0M$0.2M$0.4M$0.6M$0.8M$1.0M$1.2M$1.4M$1.6M$1.8M201920202021202220232024 YTD26$2.3M$1.9M$1.3M$1.1M$0.0M$0.5M$1.0M$1.5M$2.0M$2.5MCybersecurityAIDigital HealthFintechThis isnt just a market correction its a shakeout of different types of techFinancial servicesFintech valuations in particular havent kept up with employee headcount amid a tough fundraising environmentSource:CB Insights valuation and headcount data.“VC”represents the baseline across technologies.Includes exits from 2023 and 2024 YTD(as of 10/29/2024).$1.4M$0.0M$0.5M$1.0M$1.5M$2.0M$2.5MVCMedian M&A exit price per employee by technology,2023-2024 YTD27Fintech exit valuations have fallen dramatically since 2021 peakFinancial servicesThough 2024 shows signs of a rebound,with strong exits like Stripes purchase of BridgeSource:CB Insights valuation data as of 11/12/2024.Only includes fintech companies with disclosed exit valuations.$80M$96M$95M$60M$109M$116M$146M$115M$65M$81M$0M$20M$40M$60M$80M$100M$120M$140M$160M2015201620172018201920202021202220232024 YTDMedian M&A exit valuations for fintech companiesIIIIPlatform search:2024 fintech M&A deals by valuation28$0M$100M$200M$300M$400M$500M$600M0100200300400500600700800900Employee headcount at exitAcquisition price2024 acquisitions point to opportunistic consolidation as startups burn cashSelect fintech acquisitions in 2024,plotted by headcount and price“despite Revel having been a cash burner since the companys inception,we expect Revel synergies to contribute approximately$15M of adjusted EBITDA in the back half of 2024.”Jared Isaacman,CEO of Shift4(which acquired Revel in May 2024)“when you think about Visible Alpha,which has performed really well this year on top of the desktop growth,these are all areas where we think we can continue to engage our clients at the highest levels.”Martina Cheung,CEO of S&P Global(which acquired Visible Alpha in February 2024)Financial servicesSource:CB Insights valuation and headcount data as of 10/29/2024.29Solid fintech teams/products that may have hit growth plateaus signal clear opportunities for acquirersFinancial servicesSource:CB Insights.We filtered for solid fintech businesses with Mosaic health scores of 600 out of 1,000,then identified those that may be facing growth headwinds or be strapped for cash(5% headcount reduction YoY and no funding over the last 2 years).That yielded 120 fintech companies.Above,we show those most likely to get acquired(30% M&A probability)over the next 2 years.CompaniesM&A Probability.Mosaic(Overall)12 Month Headcount ChangeCountryBusiness Model50.9i7-11.70%Philippines50.6g0-6.86%United States46.9a3-6.45%United States46.9c3-30.14%United States41.1a9-7.69%United States40.3d8-29.82%United States38.6t0-12.75%India35.5a4-11.30%United States33.8f3-16.94%AustraliaB2BB2GCommissionB2BSaaSB2BSaaSB2CFee-for-serviceB2BSaaSUsage-basedB2BB2CCommissionB2BB2CCommissionB2BSaaSB2BCommissionIISee 120 fintech startups that meet this criteria30Disease management enters a new phase with AIHealthcare&life sciences31Companies using AI to precisely evaluate patient symptoms are gaining tractionSource:CB Insights symptom checkers&provider discovery market report.Data as of 11/18/24.These solutions typically use a conversational,chatbot-style interface to collect information on patient symptoms through a tailored series of questions.By comparing this data against known disease patterns,these solutions can then calculate the likelihood of certain conditions and provide recommendations regarding the most appropriate type of care and provider for a given patient.Healthcare&life sciencesThese companies saw just one deal in 2023 but in 2024 so far,theyve seen a flurry of equity investments and acquisitions by industry peers2023202320242024CompaniesInvestment StageDeal DateRound InvestorsGYANTSeries A-II06/08/2023OSF Ventures,Wing Venture CapitalCompaniesInvestment StageDeal DateRound InvestorsGYANTAcquired01/31/2024FabricUbieSeries D04/12/2024JA Mitsui Lease Asset,Sumitomo Mitsui Trust BankSenselyAcquired06/04/2024MediktorLoyalSeries B07/22/2024Concord Health PartnersUbieSeries D-II10/08/2024Google VenturesIIExplore the symptom checkers&provider discovery market32AI is also enabling earlier disease detection sometimes before symptoms even appearSource:CB Insights Advanced search.Categories are not mutually exclusive or exhaustive of all conditions being targeted for early detection.Illustration is not exhaustive of all focus areas for each company.Healthcare&life sciencesAI-enabled testing/screening solutions are helping disease management become more proactive across specialtiesAI-enabled early disease detectionAlzheimers diseaseUpper gastrointestinal diseaseHeart diseaseDepressionCancer(general)Liver diseaseEye diseaseThoracic disease33In some cases,AI is helping providers analyze preexisting datasets across entire patient populations to discover at-risk individualsSource:CB Insights The Future According to Mayo Clinic,company investment data.Featured investors inclusive of incubator/accelerator arms.Healthcare&life sciencesStartups like Lucem Health which enable providers to mine health data sources and proactively identify high-risk patients without direct testing are drawing attention from healthcare and big tech powerhousesCompanyBacked byCedars-Sinai and Mayo ClinicAWSBayerSelect Select LucemLucem Health peersHealth peers34These high-momentum AI-enabled startups are building and scaling their commercial networks as they push toward more proactive patient careSource:CB Insights Advanced search.The table above contains VC-backed,AI-enabled early disease detection and prediction startups with a Commercial Maturity score of 3(Deploying,i.e.,growing commercial distribution)or 4(Scaling,i.e.,expanding to additional markets).*Among a list of 50 similar startups,these players have the highest Mosaic scores,which measure private company health and growth potential.Data as of 10/30/24.Healthcare&life sciencesHealthcare strategists should keep an eye on top-ranked*startups as they explore partnership and investment opportunitiesCompanyDisease focusTotal fundingSelect business relationshipsSelect investorsCancer831|3$43MPrinceton Radiology,Longhorn Imaging Center,RAYUS RadiologyAllianz Life Ventures,Nvidia Inception ProgramHeart and lung disease801|3$168M3M,AstraZeneca,Mayo ClinicLG Technology Ventures,National Institutes of Health,3M VenturesTuberculosis,lung cancer,stroke783|4$126MSiemens Healthineers,AstraZeneca,MedtronicMerck Global Health Innovation Fund,Novo Holdings,MassMutual VenturesEye disease,vascular disease,neurodegenerative conditions,rare diseases759|3$17MDoheny Eye Institute,Boehringer Ingelheim,NovartisSanoptisCardiac disease742|3$21MAgepha Pharma,Venturi Cardiology,Surrey Cardiovascular ClinicOxford Science Enterprises,Oxford UniversityCardiovascular,kidney,eye disease696|3$12MSeverance Hospital,Bundang Jaesaeng General HospitalAstraZeneca,Mayo Clinic,JLabs KoreaMosaic|Commercial Maturity scoreIIIDiscover other high-momentum startups35Investment floodgates open for RNA therapeuticsHealthcare&life sciences36The market for RNA therapeutics has taken off since 2016Image source:IQVIA.Data as of November 2023.Healthcare&life sciencesTherapies target specific genes to silence disease-causing mutations or produce desired proteinsNotable ASO approved in 2016 for treating spinal muscular atrophy(generated$1.8B in sales in 2022)mRNA tech from Covid-19 vaccine development is a recent driverFirst FDA-approved RNA interference(RNAi)therapeutic in 2018;treats a form of peripheral nerve disease37RNA therapeutics target traditionally“undruggable”proteins,found in genetic disorders,cancer,cardiovascular disease,and more*IQVIA estimate.Healthcare&life sciences3 leading approaches account for 80%*of the RNA therapeutics in development(as of November 2023)Synthetic DNA/RNA strands designed to bind to specific messenger RNA sequences to modify or block protein expression.Small RNA molecules that silence specific genes by binding to and destroying their messenger RNA,preventing disease-causing protein production.A single-stranded RNA molecule that delivers genetic instructions to cells to teach them toproduce specific proteins like antibodies.RNA INTERFERENCE(RNAi)ANTISENSE OLIGONUCLEOTIDES(ASOs)MESSENGER RNA(mRNA)38Leader Alnylam Pharma advances from liver to brain-targeted therapies,directed at central nervous system(CNS)diseasesSource:CB Insights Alnylam Q324 earnings call insights;Alnylam.Note:The liver is the easiest place to deliver RNAi molecules,though companies are exploring other tissues.Healthcare&life sciencesOverall,the company has 5 FDA-approved RNAi therapies and 15 in its pipelineProductDiseasePhaseMivelsiran(ALN-APP)Cerebral Amyloid Angiopathy(CAA)2Mivelsiran(ALN-APP)Alzheimers Disease1ALN-HTT02Huntingtons Disease1ALN-SODSOD1 Amyotrophic Lateral Sclerosis(ALS)1Alnylam Pharma neurologic pipeline(as of October 2024)Earnings InsightsAlnylams pipeline updates highlighted the companys increasing focus on CNS indications,a shift from its historical emphasis on liver-targeted therapies.Programs like ALN-APP for Alzheimers disease and ALN-HTT02 for Huntingtons disease were featured more prominently in the Q3 2024 call,with the company sharing details on trial designs and expected data readouts.Insights generated by combining CB Insights data and AIQ3 FY 202439Biotech startups expand where and how therapies can be delivered to target different types of diseasesSource:CB Insights Advanced search.Note:Select startups visualized out of 120 based on approach and recency of funding.Healthcare&life sciencesEvaluate investment opportunities in the RNAi/ASO market as pharma leaders back newcomersCompanyLocationLatest roundDateFocus areaSelect pharma backersCity TherapeuticsUS$135M Series AOctober 2024Novel RNAi trigger molecules and delivery systemsRegeneron VenturesLeal TherapeuticsUS$45 Series AOctober 2024ASOs targeting central nervous system(CNS)disordersChugai Venture FundJudo BioUSSeries AOctober 2024Directing small interfering RNA(siRNA)medicines to the kidney,targeting hypertension,endocrine disorders,etc.Hygieia PharmaceuticalsChinaSeries A-IISeptember 2024siRNA delivery tech to target both the CNS and peripheral nervous system(PNS)Vico TherapeuticsNetherlands$12M Series BJune 2024Novel ASOs for neurodegenerative disorders Sanegene BioChina$80M Series ADecember 2023Delivery of RNAi drugs to the liver;novel extrahepatic(outside the liver)techniques Switch TherapeuticsUS$52M Series BMarch 2023Cell-specific RNA therapy that only activates under certain conditions,combining ASO/siRNA benefitsEli Lilly&CompanyIdentify investment opportunities40Autonomous robots eye caregivingHealthcare&life sciences41The US healthcare industry faces a critical staffing shortage amid a rapidly aging populationSource:Our World in Data,Physician workforce in the United States of America:forecasting nationwide shortages(Zhang et.al),American Association of Colleges of Nursing.Healthcare&life sciencesProjected shortages of over 139,000 physicians and 63,000 nurses by 2030Under-25sUnder-15sAges 65 42Healthcare leaders look to advanced robotics for solutionsSource:CB Insights research Robots are coming for healthcare.Note:Select deals based on application novelty and healthcare stakeholder relationships.Mosaic=startup health score out of 1,000(as of 11/14/2024).Healthcare&life sciencesCategories in focus include caregiving support,hospital operations and logistics,lab automation,and remote care deliveryCaregiving supportExoskeletons for rehabilitation and mobility assistanceLatest round:$10M Series A(May 2023)|Mosaic:573Select partners:Toronto Rehabilitation Institute,Shirley Ryan AbilityLabHospital operationsService robot that assists clinical staff with non-patient-facing tasks,like supply deliveryLatest round:$44M Series C(June 2024)|Mosaic:809Select clients:Northwestern Medicine,Cedars-Sinai,and Rochester Regional HealthLab automationAutomated tissue sectioning and transfer systemLatest round:$36M Series B(July 2024)|Mosaic:751Select investors:Northwell Holdings&Ventures,Mayo Clinic,Ochsner Ventures,CU Healthcare Innovation FundRemote care deliveryMicrorobotic pill that functions as a virtual endoscope for real-time internal imagingLatest round:$1.4M convertible note(October 2023)Select investor/licensor:Mayo Clinic43Source:CB Insights 1X scouting report;Company releases.Highly capable humanoid robot developers eye caregiving applicationsHealthcare&life sciencesTesla and Figure,have mentioned caregiving as a potential use case.This includes in-home assistance with tasks like lifting heavy objects or performing household choresScouting ReportGenerated by CB Insights1X Scouting Report Core Products/Services:Develops and manufactures humanoid robots equipped with AI for various industries.Primary product is Neo,a second-generation android designed for home assistance and consumer tasks.Robots feature human-like movements and behaviors,capable of learning and improving task performance over time.The companys“master plan”highlights a long-term business opportunity:700M aging population in need of at-home careTesla Optimus RobotWhat can it do?It can basically do anything you want.It can be a teacher,babysit your kids,it can walk your dog,mow your lawn,get the groceries,just be your friend,serve drinks.”Tesla CEO,Elon Musk 44Riding wave of market interest,autonomous robots have healthcare in viewSource:CB Insights Advanced search.Note:Mosaic scores as of 11/9/2024.*Based on company disclosures from CB Insights Analyst Briefings,company websites,press articles,and earnings transcripts.Healthcare&life sciencesHealthcare leaders should explore partnerships and pilots now to prepare for next 10 yearsLeading private humanoid robotics companies by Mosaic score&public companies with targeted healthcare/caregiving applicationsCompanyMosaicCommercial MaturityManufacturingLogisticsRetailHealthcare8792:ValidatingXXXXXXX7982:ValidatingXXXXXXX6923:DeployingXXXXXXX6542:ValidatingXXXXX6512:ValidatingXXXPublic-XXXXXXXPublic-XXXXxXXIIIIIITrack humanoid robotics companiesCurrent target areas*Anticipated target areas*45AI M&A fuels the next wave of corporate strategyAI46Since 2020,AIs share of corporate tech M&A has doubled as corporations race to grab AI tech&talent AISource:CB Insights data on nearly 30,000 corporate acquisitions of tech companies since 2020.“AI”covers companies developing AI solutions,including cross-industry applications,industry-specific products,AI development tools,and more with a focus on those using AI as a core part of their products or services.0.00%1.00%2.00%3.00%4.00%5.00%6.00%7.00%8.00%3.5%4.2%4.0%6.9%AIs share of corp.tech M&A7.2%0.00%1.00%2.00%3.00%4.00%5.00%6.00%7.00%8.00 202021202220232024 YTDIIIIPlatform search:300 corporate AI acquisitions in 202447The top dogs of AI M&A right now:Nvidia,Snowflake,Accenture,DatabricksSource:CB Insights data on tech company exits.“AI”covers companies developing AI solutions,including cross-industry applications,industry-specific products,AI development tools,and more with a focus on those using AI as a core part of their products or services.Big tech used to dominate AI M&A;now its mainly AI infra&data mgmt.playersRankAcquirer#of AI acquisitionsSelect acquisitions17253434Top acquirers in 2020-21:Big tech&gamingRankAcquirer#of AI acquisitionsAcquisitions14141414Top acquirers in 2023-24 YTD:AI infrastructure,data mgmt.,&consultingAI48Some leaders are doubling down on AI acquisitions to embed AI across their offeringsNvidia,Salesforce,and Thomson Reuters have all accelerated AI M&A this year,each with a focus on offering broader AI solutions in their domainsSource:CB Insights data on tech company exits;Nvidia Q2 FY25 earnings insights.Acquirer#of AI acquisitions in 2024(vs.2023)2024 acquisitions3 in 2024(vs.1 in 2023)2 in 2024(vs.1 in 2023)2 in 2024(vs.1 in 2023)AI49222340112233445Customer self-service(CSS)Data scienceplatformsEmail securityMarketingautomationpersonalizationAI-poweredchatbotsTwo customer-facing tech markets AI chatbots and marketing personalization lead this year in corporate AI acquisitionsAI chatbots is the hottest tech market in the current corporate AI M&A environmentSource:CB Insights tech market reports AI-powered chatbots,Marketing automation personalization,Email security,Data science platforms,and Customer self-service(CSS).Data as of 10/31/2024.AICompanyAcquirerDate09/03/202408/08/202403/13/202401/10/2024Corporate M&A teams pile into AI chatbot marketAI acquisitions in 2024 YTDAI acquisitions in 2024 YTD50Which AI startups are incumbents most likely to acquire next?Source:CB Insights platform search.M&A probability data as of 11/5/24.Note:We started with a pool of 4,500 AI companies that 1)have a B2B business model and 2)have received investment from a corporation or CVC.We filtered down for those with healthy businesses(Mosaic score 600 )and strong management teams(Management Mosaic score 600 )resulting in 500 companies then zeroed in on those with 40% M&A exit probability to get the above 8 results.Eight AI startups with corporate backers lead the pack with high M&A probability,solid momentum,and strong management teamsCompanyM&A probabilityMosaic|Commercial Maturity scoreFocus areaSelect corporate/CVC backers55p1|4Security operations center(SOC)AI agent/copilot50p7|3Computer vision&digital twins for infrastructure mgmt.46i5|3Parametric insurance for climate risk45x2|4Coding AI agent/copilot45q4|3Warehouse robots for picking,packing,&sorting44c2|3Data processing for AI/ML workloads43i6|3AI copilot for complex service requests(e.g.,hardware repair)43h3|3AI-powered threat detectionAI(the venture arm of BNP Paribas)(Google Ventures)IIIExplore these 8 startups51LLMs explainability momentAI52AI has a“black box”problemSource:CB Insights research Explainable AI.Note:Interpretability is a models inherent transparency and understandability by design.Interpretability image source:AWSAIWith highly complex neural networks(like LLMs),its especially hard to understand how they arrive at their outputs and what drives their decision-making processInterpretability versus performance trade-off given common ML algorithms53Researchers are actively investigating techniques to understand why LLMs say what they say and how we might better control themSource:“From Understanding to Utilization:A Survey on Explainability for Large Language Models,”Luo and Specia(February 2024).Note:Explainability refers to methods and techniques applied after-the-fact to understand how a model makes decisions.AIExplainability techniques for pre-trained LLMsLocal analysisFeature Attribution Analysis Dissecting Transformer Blocks Gradient-Based Methods Perturbation-Based MethodsAnalyzing MHSA Sublayers Analyzing MLP SublayersVector-Based MethodsAims to explain specific model predictionsGlobal analysisProbing-Based Methods Mechanistic InterpretabilityProbing KnowledgeProbing RepresentationsCausal Tracing Vocabulary LensCircuit Discovery Methods focused on model-wide understandingNotable research area focused on“reverse engineering”neural networksLeveraging explainability(applications)Model EditingEnhancing Model PerformanceHypernetwork Knowledge Editors Locate-Then-EditImproving Utilization of Long Text Improving In-Context LearningControllable GenerationReducing Hallucination Ethical Alignment54Leading model developers drive new breakthroughs in understanding LLM behaviorSource:Company researchAIWhile the field is still in the early stages,recent advances highlight the potential of scaling interpretability“Scaling Monosemanticity:Extracting Interpretable Features from Claude 3 Sonnet”(May 2024)Researchers identified tens of millions of specific features inside the model(like a map of what it knows)&ability to manipulate these features to change how the model behaves“Scaling and evaluating sparse autoencoders”(June 2024)Researchers developed new methods using a smaller neural network to map 16M features(interpretable patterns of neural activity found within language models)of a frontier LLM55Explaining model behavior is important for earning users trust and critical for regulated industries Source:CB Insights Earnings transcripts analytics.Select quotes&industries visualized.AIExplainability advances will support further adoption of complex models“In addition to creating training datasets for this autonomous delivery vehicle customers computer vision perception models,weve expanded our scope to includefine-tuning of the clients LLMs builtfor real-time explainability of the autopilots decisions.”Telus International Q1 FY 2024,CEO Jeff Puritt“Machine learning and AI models anchored on consumer preferences and prior experiences across our provider network have been built for weight loss.We are also moving toward integrating explainability into the EMR,further empowering providers with these groundbreaking tools.”Hims Q2 FY 2024,CEO Andrew Dudum“To date,Equifax has about 70 approved AI patents supporting our AI neurodecisioning technology,which we call NDT and explainable AI,which is critical to ensuring that the correct data is used to make credit decisions that surface by AI models and scores.”Equifax Q3 FY 2023,CEO Mark BegorMEDICAL DECISIONINGAUTONOMOUS VEHICLESCREDIT DECISIONINGIIISee what leaders are saying56Companies in model validation&monitoring an already established category in ML are now targeting enterprise deployments of LLMsSource:CB Insights research The responsible AI market map(August 2023)Companies in this category focus on identifying model vulnerabilities,mitigating bias,and improving performance against adversarial attacksAI57Emerging startups focused on mechanistic interpretability go a step further to help companies steer models Watch this market for investment&partnership opportunities,though LLM developers may move to offer these tools as well beyond their research effortsAIIIITrack these companies58Open-source cedes top LLMs but dominates smaller modelsAI59Big techs dominant approach is to prioritize closed flagship models while also releasing lighter-weight open modelsSource:CB Insights company data;company releases.Note:Market cap data as of 11/15/2024.Companies selected based on market cap®ional relevance.Developers open-sourcing AI models do so on a spectrum,sharing some combination of model weights,underlying source code,and original training data.AIPublic companyMarket capCountryDominant approachNotable closed activityNotable open activity$3.60TUSOpenInvested in multiple closed model developersIntroduced NVLM 1.0 multimodal frontier-level LLM family(September 2024)$3.45TUSClosedAnnounced proprietary on-device and server foundation models(June 2024)Released OpenELM model family(April 2024)$3.17TUSClosedMulti-billion-dollar investment in OpenAI;rumored to be working on 500B parameter MAI-1 model(May 2024)Released Phi-3 small language models(April 2024)$2.22TUSClosedAmazon Titan foundation model family available on Amazon BedrockSupports open-source AI models on AWS infrastructure$2.01TUSClosedAnnounced flagship Gemini 1.5 model(February 2024)Introduced lightweight Gemma model family(February 2024)$1.46TUSOpenIntroduced Llama 3.1 frontier-level model family(July 2024)$532.9BChinaClosedAnnounced Hunyuan Turbo foundation model(September 2024)Text-to-image large model Hunyuan-DiT released(May 2024)$214.3BChinaOpenFlagship Qwen language models available via APILaunched Qwen2.5 family of over 100 models(September 2024)60In the private market,closed LLM developers lead in equity fundingSource:CB Insights Advanced search.Note:Analysis includes LLM developers sorted by total equity funding.Categorized by dominant approach.AI$19.1B$7.5B$6.1B$1.5B$1.5B$1.0B$1.0B$0.97B$0.85B$0.42BClosedOpen“Acquihired”by Microsoft(March 2024)“Acquihired”by Amazon(June 2024)IIITrack model developers61High compute costs,limited moats,and big tech competition have created a market ripe for a shake-upSource:CB InsightsAIRecent pivots and quasi-exits among foundation model players validate the trendQuasi-exits collapsing into big techAll essentially“acqui-hired”by big tech companies,with the founders and large portions of teams going to the acquirer.The deals reflect the high costs of model development,with licensing payments going to investors.Pivots to lighter-weight models,while layering paid services on topBoth moved away in 2024 from competing on general-purpose LLMs to building smaller and/or optimized models and related AI tools.Paywall frontier modelsFor open-source AI developers without a clear path to revenue,selling access to their best models while open-sourcing their lower tiers is one approach companies are taking much like big tech.62Performance gaps are converging thanks to big tech efforts,though closed models maintain their leadSource:Company releases.Note:Scores from release date.Select leading models visualized.*Claude 3.5 Sonnet(new)released October 2024 scores 78.0%on the MMLU-Pro.*As of its beta release,Grok-2 is not open-source,though Elon Musk indicated on X it will be once the next version is released.AICompanyModelMMLU-ProMMLURelease dateClaude 3.5 Sonnet*CLOSED SOURCE76.1.3%(0-shot CoT)June 2024Grok-2*CLOSED SOURCE75.5.5%(0-shot CoT)August 2024LLama3.1-405BOPEN SOURCE73.3.6%(0-shot CoT)July 2024GPT-4oCLOSED SOURCE72.6.7%(0-shot CoT)May 2024Qwen2.5-72BOPEN SOURCE71.6.1%(5-shot)September 2024Gemini-1.5-ProCLOSED SOURCE69.0.9%(0-shot CoT)May 2024MMLU-Pro and MMLU benchmarks for leading foundation models63Proprietary models become more cost-effective for enterprise useSource:CB Insights.Chart data from OpenAI.Falling inference costs and growing context windows,combined with managed infrastructure and enterprise support,are strengthening the business case for API-based solutionsAIMar 23Jun 23Sep 23Dec 23Mar 24Jun 24Sep 24GPT-4o$4per 1M tokensGPT-4$36per 1M tokensGPT-4 Turbo$14per 1M tokensGPT-4o$7per 1M tokensGPT-4o mini$0.25per 1M tokensThe cost of 1M tokens has fallen from$36 to$0.25 in the last 18 months.Comparing the more performant models(GPT-4 vs.GPT-4o),the price is still a nearly 10 x improvement.64A few frontier,closed models will dominate sophisticated enterprise workSource:CB Insights buyer interviews OpenAI,Cohere.AIThough enterprises will adopt smaller,open-source models for task-specific applications where locally hosted models are preferred“The reason that we have focused on the large closed-source solutions is one,just objectively,their models are better right now,meaning they have higher accuracy,lower hallucination.I think the second reason as well is because we get a lot of valuable enterprise support from OpenAI and GoogleWe would not get that level of support with an open-source solution.”Lead Architect,Gen AI at publicly traded communications company“At this point in time,a lot of customers come to us and say,Im not sure I can trust my sensitive data with something that is outside my service boundaries,so I want you guys to go ahead and get an open-source foundational model and then fine-tune the model so that the model resides within my service boundary.”Chief Architect,Fortune 500 companyIIIIDive into buyer interviews65The US is leading the AI arms race for nowAI66The US is running away with AI fundingSource:CB Insights.Share of annual AI equity funding by major global region.Data as of 11/8/2024.American startups now take 71 cents of every dollar of global AI equity fundingAI61bXp%US,71 %Asia,13%Europe,14%Rest of world,3 202021202220232024 YTD67AI talent is concentrated in the country,spurring a potential innovation gap between the US and everywhere elseSource:CB Insights.Analysis shows top 5 countries globally by share of AI companies.Based on 15,000 AI companies tracked by CB Insights.Over 40%of the worlds AI companies are based in the USIIIIISee 15,000 AI companies in the Expert CollectionAI68China is the only global power likely to rival the US dominance in LLMs,especially on the open-source frontSource:CB Insights;Hugging Face.*Based on results from Hugging Faces Open LLM Leaderboard v2,launched in June 2024,which measures open-source LLMs abilities across general tasks like reasoning and mathematics.Models are ranked according to performance across 6 benchmark evaluations.Chinas tech giants are simultaneously developing advanced models and investing in startups doing the same And Chinas big tech firms are backing startups that could threaten the leadership of US firms like OpenAIAlibabas Qwen2 tops Hugging Faces leaderboard*of open-source LLMs(as of June 24)RankModel1Qwen/Qwen2-72B-Instruct2meta-llama/Meta-Llama-3-70B-Instruct3microsoft/Phi-3-medium-4k-instruct401-ai/Yi-1.5-34B-Chat5CohereForAI/c4ai-command-r-plus6abacusai/Smaug-72B-v0.17Qwen/Qwen1.5-110B8Qwen/Qwen1.5-110B-Chat9microsoft/Phi-3-small-128k-instruct1001-ai/Yi-1.5-9B-ChatChinas private foundation model developers valued at$1B AI69CountryAI equity funding,Nov.22 Nov.23AI equity funding,Nov.23 Nov.24YoY change in fundingMedian YoY headcount growth*Select local AI startups gaining traction(funded in the last year headcount growth)Belgium$21M$141M571%Brazil$34M$194M471%Italy$63M$346M449%Australia$233M$639M174%9%Norway$86M$216M151%India$648M$864M33%Switzerland$204M$271M33%South Korea$641M$825M29%United States$40.4B$52.1B29%8%While the AI market will solidify around frontier models in the US and China,watch for other specialized AI hubs to emerge in specific domainsSource:CB Insights.Data as of 11/8/2024.Excludes countries that received 400)and Commercial Maturity scores of 1-4(out of 5);data as of 11/18/2024.IndustrialsEvaluate opportunities among high-momentum*space tech companies Explore 150 high-momentum space tech companies92Our most popularresearchResearchGame Changers 2025State of AI Q324 ReportThe Enterprise AI RoadmapHow AI agents will transform enterprise workflowsFuture Tech HotshotsMarket reportsLarge language model developersESG reporting,monitoring,&data analysisAI development platformsReal-world data providers93IIIIIIITry our platform for free

2025-11-03 93页 5星级