2025年二季度AI公共报表和估值指南（英）2025

EMERGING TECH RESEARCH AI Public CompSheet andValuation Guide PitchBook Data, Inc. Key takeaways Nizar TarhuniExecutive Vice President of Researchand Market Intelligence Paul CondraGlobal Head of Private Markets Research Stock returns James UlanDirector of EmergingTechnology Research Public markets continue to favor scaled AI infrastructure providers.AI-driven momentum continued to shape public markets inQ2. The S&P 500 rose 10% and the Nasdaq 17%. AI core conglomerates, core pure-plays, and semiconductors returned 22%, 25%,and 27%, respectively, above the S&P 500. Outperformance was concentrated in AI datacenter chips, AI software infrastructure, andcloud hyperscalers. Markets continue to reward companies that control AI infrastructure and demonstrate monetization, signaling ashift from hype to performance-based evaluation. Institutional Research Group Dimitri ZabelinSenior Research Analyst,AI and Cybersecruitydimitri.zabelin@pitchbook.com pbinstitutionalresearch@pitchbook.comPublished on August 13, 2025 Hyperscalers lead AI core conglomerates.CoreWeave posted a 210% return, by far the top performer, driven by investorenthusiasm around neocloud infrastructure and surging demand for non-hyperscaler access to GPU clusters. Though clearly anoutlier, upside was broad-based among large-cap hyperscalers. Oracle rose 54%, supported by growing AI workloads on its cloud platform and a strengthened partnership with NVIDIA. Microsoft(up 30%) and Meta (up 26%) benefited from expanding deployment of AI copilots and productivity tools tied to generativeAI adoption. Amazon rose 14%, lagging peers despite Amazon Web Service’s foundational role in AI infrastructure. Alphabet and IBM eachreturned 12%, reflecting more cautious investor sentiment due to competitive pressure and margin concerns. Asia-basedconglomerates underperformed but had limited index impact. The cohort’s 14% median return reflects investor preference for firmswith compute control and monetization strategies. PitchBook clients can accessthefull Excel data packfor thisreport via the Details tab in thedocument viewer. AI semiconductors outperform amid persistent infrastructure demand.AI semiconductor companies outperformed all othersubsectors, returning a mean of 37% and a median of 39%. Investors favored companies enabling model training and inference,particularly in accelerators, memory, and interconnect. Broadcom (up 64%), SK hynix (up 60%), and Arm (up 51%) led. Broadcom benefited from surging demand for customer datacenterAI processors. SK hynix gained on high-bandwidth memory demand and pricing, and Arm saw licensing momentum in AI-optimizedCPU designs. NVIDIA (up 43%) and AMD (up 38%) performed strongly on GPU demand and road map visibility. NVIDIA remains central to theAI compute stack, while AMD’s MI300 series is gaining significant hyperscale traction. Astera Labs gained 46% post-IPO due toexposure to high-speed interconnects. Micron (up 39%), Marvell (up 24%) also rallied on AI datacenter demand upside. Cambricondeclined 4% amid continued skepticism over execution and scale. Valuations Valuations across all AI subsegments rose in Q2, though not materially, despite stronger-than-expected revenue growth. Cloudinfrastructure providers (hyperscalers) ended the quarter with an enterprise value (EV)/trailing 12-month (TTM) revenue multiple of6.8x, up modestly from 6.6x at year-end. AI core pure-plays companies traded at 27x, vertical conglomerates at 9.3x, vertical pure-plays at 16.8x, semiconductors at 17.3x, and autonomous machines at 6.8x. With the exception of semiconductors, EV/revenue multiples remain well below their pandemic-era peaks. Valuations for AIsemiconductor vendors, however, are nearing all-time highs. This reflects the sector’s foundational role across all aspects of AIcompute, networking, and power management. We expect further upside as macro-overhangs recede, including resolution on tariffpolicy, reduced inflation expectations, and dovish monetary policy. Revenue Across most segments, street consensus projects deceleration in revenue growth in 2026 relative to 2025. The two exceptionsare core conglomerates and autonomous machines. We agree with this view. Industrial adoption of AI is in the early innings and isadding to the cyclical recovery that begun in late 2024. Additionally, hyperscalers have signaled that demand for AI cloud computedemand continues to accelerate, and they have increased capital expenditure budgets to match. That said, we expect broader growth across the space to decelerate over the next 18 months. Growth has ramped quickly since 2023and will remain robust through 2026. We believe AI semiconductors and both AI vertical pure-plays and conglomerates are likelyto outperform revenue consensus with some seasonality built in around year-end. Agentic AI will become embedded across a widerange of enterprise workflows. Late-stage software providers stand to benefit the most as large enterprise cl