构建、支持和购买AI（英）2026

Building, Backing, andBuying AI Institutional Research Group Kaidi GaoSenior Research Analyst,Venture Capitalkaidi.gao@pitchbook.com How Big Tech and enterprise software companies arediverging and what it means for the M&A landscape Caleb WilkinsData Analystpbinstitutionalresearch@pitchbook.comPublished on June 4, 2026 PitchBook is a Morningstar company providing the most comprehensive, mostaccurate, and hard-to-find data for professionals doing business in the private markets. Contents Key takeaways Key takeaways1Introduction2Big Tech: Building and backing, not buying4NVIDIA: Acquiring to extend theAI infrastructure stack10Enterprise software incumbents:Acquiring to compete12Frontier AI labs and large AI-nativecompanies as acquirers15Regulatory and geopolitical challenges17M&A and exit implications19References20 •Big Tech has redirected AI capital toward infrastructure and strategic partnershipsrather than acquisitions. The Big Five’s global VC-backed acquisition countdeclined from a peak of 33 in 2017 and 2018 to seven in 2024. Combined capitalexpenditure for Google, Microsoft, Amazon, and Meta is projected to reach nearly$600 billion in 2026, leaving limited capacity for acquisition premiums. The grouphas shifted toward large minority investments and compute-linked partnerships tosecure AI positioning. •NVIDIA is acquiring to extend and entrench the AI infrastructure platform on whichthe rest of the market depends. Since 2022, NVIDIA has made 21 acquisitions(15 of them AI-related) spanning workload orchestration, model optimization,and inference. Its $20 billion licensing agreement with Groq in December 2025reflects a company expanding from training into inference workloads, making itsinfrastructure more capable, efficient, and harder to displace. •AI-driven disruption to per-seat SaaS economics is reshaping who acquires andwhy in the enterprise software market. AI agents can now execute workflowsthat previously required dedicated software subscriptions, compressing thelicensing economics that defined SaaS for two decades. The iShares ExpandedTech-Software ETF fell 29.6% between December 2025 and February 2026, andMorningstar downgraded six software companies from wide to narrow moat inMarch 2026. •Among enterprise software incumbents, M&A has emerged as a key response toAI disruption for companies with the financial capacity and platform logic to act.Salesforce, Snowflake, Databricks, ServiceNow, and Workday have all increased theAI share of their recent deal activity, while companies facing structural or financialheadwinds, including Cisco and Intel, have pulled back. Salesforce is the second-most-active corporate acquirer of VC-backed AI startups globally by deal count,with 18 of its 23 acquisitions since 2022 carrying an AI tag. •Frontier AI labs and large AI-native companies have become meaningfulacquirers, targeting companies with proprietary data, deep workflow integration,or infrastructure capability that cannot be quickly replicated by a model update.OpenAI and Anthropic are acquiring to build the vertical product depth that modeldevelopment alone cannot provide. The profile of targets pursued, including TorchHealth and Coefficient Bio, signals which capability gaps the market treats asstrategically urgent. •Regulatory scrutiny of talent acquisition structures and geopolitical interventionin cross-border deals represent two newly emerged risk categories for AI M&A.The Federal Trade Commission’s investigation into Microsoft’s deal with InflectionAI and the DOJ’s probe into Google’s talent acquisition of Character.AI’s teamsignal that talent-and-license deal structures face closer antitrust scrutiny. China’sorder to unwind Meta’s $2 billion Manus acquisition in April 2026 establishes thatcorporate domicile alone cannot insulate cross-border AI deals fromregulatory intervention. Data scope •Elevated private market valuations have placed the top-tier AI companies beyondconventional M&A reach, shifting the rational response for large buyers fromacquisition to investment. Recent private rounds have implied multiples that moststrategic acquirers will not meet without a highly specific capability fit or concretefinancial justification. For the broader AI startup universe, exit prospects dependon whether a target can demonstrate those metrics or align with the platform roadmaps of a concentrated group of well-capitalized acquirers. This note draws on both VC-backed M&A data and broaderM&A data that is not limited toVC-backed targets. Both datasetsare used to provide an accurateand comprehensive picture ofacquisition activity and strategyacross Big Tech, NVIDIA, andenterprise software incumbents.While the majority of AI startupacquisitions involve VC-backedcompanies, some acquired AIcompanies were PE-backed orhad no institutional investorbacking at the time of acquisition.Certain adjacent strategicdeals, including hardware andinfrastructure transactions,may also fall outside the VC-backed