AI公共报表和估值指南（英）

EMERGING TECH RESEARCH AI Public CompSheet andValuation Guide Key takeaways Institutional Research Group Dimitri ZabelinSenior Research Analyst,AI and Cybersecuritydimitri.zabelin@pitchbook.com Stock returns pbinstitutionalresearch@pitchbook.comPublished on February 24, 2026 Public markets are not applying a uniform AI premium. Capital has concentrated where revenue acceleration and operating leverageappear strongest, particularly in infrastructure and select pure plays, while diversified platforms remain more selectively priced. TheQ4 data reflects how the market exited 2025, while YTD performance as of February 20, 2026, provides context on how positioning hasevolved into early 2026, especially in light of the “SaaS-pocalypse.” Contents Key takeaways2Stock returns6Revenue7EBITDA9 AI core conglomerates:90-day performance across AI-linked megacaps was uneven. Google led at 27%, materially ahead of Amazon at4%, with Tesla and IBM each at 3%. The dispersion suggests investors differentiated based on perceived immediacy of AI monetizationwithin core business lines. Broad AI exposure alone did not drive a uniform re-rating across diversified platforms YTD; as of February 20,2026, performance has further diverged. CoreWeave is up 36%, materially outperforming peers, while Microsoft and Oracle are down 18%and 20%, respectively. The spread suggests investors are favoring concentrated AI infrastructure exposure while de-rating diversifiedplatforms amid broader multiple compression. AI core pure plays:Returns were similarly split. MongoDB gained 29%, Shopify rose 6%, and AppLovin declined 1%. Outperformanceskewed toward infrastructure-adjacent software, indicating a preference for foundational data and developer layers. Application-layer exposure was priced more selectively. Similar to AI core conglomerates, pure plays are broadly negative YTD, with Tempusroughly flat at 2%. The pullback reflects sensitivity in high-multiple software as markets reassess growth durability and timing of AImonetization at scale. PitchBook clients can access thefull Excel data pack for this reportvia theResearch Centeron thePitchBook Platform. AI semiconductors:Semiconductors were the strongest-performing segment into year-end. SK Hynix rose 61%, Micron 55%, and AMD26%. The magnitude of outperformance reflects sustained conviction in AI-driven compute and memory demand. Hardware remained thehighest-beta expression of AI capital expenditure expectations at the close of 2025. YTD, semiconductors have remained comparatively resilient. Micron is up 46% and ASML 36%, with the group outperforming the S&P 500 by approximately 6%. Relative strength indicatescontinued conviction in AI-driven infrastructure demand despite volatility in broader equities. AI autonomous machines:Automation-linked names also delivered strong Q4 gains, with Fanuc up 39%, UiPath 29%, and Intuitive Surgical28%. Performance suggests investor appetite for AI exposure tied to measurable productivity and workflow integration. Autonomousmachines are underperforming the S&P 500 by roughly 6% YTD. The reversal from Q4 strength suggests investor caution toward longer-duration automation themes as capital becomes more selective. Taken together, Q4 positioning showed capital concentrated in infrastructure and automation layers, with more selective repricing acrossdiversified platforms and application software. Early 2026 performance indicates increasing differentiation, with resilience concentrated insemiconductor names most directly exposed to sustained AI capital expenditure. Revenue AI core conglomerates:Meta led 2025 with 22% YoY revenue growth, followed by Google and Microsoft at 15% and Amazon at 12%.Growth reflects AI integration across advertising and cloud platforms, but the spread suggests varying exposure to enterprise demandand capital intensity. At scale, AI enhances growth but does not fundamentally alter the trajectory of diversified platforms. AI core pure plays:Upstart grew 63%, Palantir 56%, and Snowflake 29% YoY. These rates indicate sustained enterprise demand for AI-enabled analytics, decisioning, and data infrastructure. The data suggests the cycle is moving from experimentation toward scaledworkflow deployment, particularly within data-centric platforms. AI semiconductors:NVIDIA grew revenue 114%, Micron 49%, and AMD 34%. Infrastructure remains the most concentrated source ofAI revenue expansion, supported by sustained datacenter investment and compute intensity. Growth moderation is likely over time, butinfrastructure continues to capture a disproportionate share of the AI spend cycle. Autonomous machines:Samsara grew 33%, Symbotic 26%, and Intuitive Surgical 21%. Growth aligns with rising automation demand andproductivity-driven deployment, where AI is embedded within operational systems. Across segments, revenue data reinforces a layered cycle: Infrastructure leads, software scales behind it, and conglomerates absorb AIinto existing