AI板块抛售：行情结束，还是反应过度？

The AI sell-off: Over or underdone? 23February 2026 Luke Templeman | Thematic Strategist |Luke.Templeman@db.com| +44 207 541 0130Galina Pozdnyakova | Research Analyst |Galina.Pozdnyakova@db.com| +44 207 547 4994 IMPORTANTRESEARCHDISCLOSURESANDANALYSTCERTIFICATIONSLOCATEDINAPPENDIX1.UNTIL19thMARCH2021INCOMPLETEDISCLOSUREINFORMATIONMAYHAVEBEENDISPLAYED,PLEASESEEAPPENDIX1FORFURTHERDETAILS. This sell-off is not the first … but the frequency of new AI tools has accelerated in recent months 2026 (a small selection) Previous years … 1)Mid-late 2022: Smaller AI firms find it difficult to attract capital duringa broader risk-off move.2)2023: The ‘first wave’ of AI disruption fears. These typically includedlabour-based businesses providing simple, repetitive labour-basedservices–such as call centres and other offshored business processoutsourcing businesses3)2024: Various small sell-offs which combined technology valuationissues with broader market concerns4)Jan 2025: The release of a low-cost AI model by China’sDeepSeekrattles US markets5)Sept 2025: Fears grow in the software industry and reverberate aroundprivate capital managers with heavy exposure.6)Oct 2025: Concerns grow for companies in the broader tech spaceincluding those that use platforms that investors fear might bereplicated or disrupted by AI. 1)Jan: Google announces new tools for various products2)Feb: Anthropic announces a variety of new AI tools3)Feb:Algorhythmannounced a new tool that helps companies scaletheir freight shipments4)Feb: OpenAI announces ‘Frontier’, an enterprise platform for AI agents5)Feb: Microsoft announces Agentic AI to automate many retail functions Points to note: •The most recent sell-off is quite distinct and includes a much broaderview about how AI will impact businesses and their markets•The sell-off reflects a view that current economic structures will changemeaningfully as a result of the latest AI products being released.•Many affected firms are network-based companies whose balance sheetassets are largely intangible, and whose services are largely intangible•There has been a rotation towards companies with real assets and goods(egdefence, commodities, banks, utilities, capital goods sectors). Software & services has been the worst performing industry group in both the US and Europesince the middle of January PE multiples: software & services used to be the 3rdmost expensive industry group, it’s now 9th(3rdto 13thin Europe), down by 5.8x (5x). No other industry group has seen a similar multiple re-rating. It’s a similar story for the PEG ratios. Although data is less available, software & services in the USslid from 7thto 17thmost expensive industry group (3rdto 15thin Europe) 02.Selected US stocks Sector focus: Software–This group had by far the largest number of battered firms, with the ten below fallingby at least 30% despite being at least $20bn market cap firms at the start of the rout. All have delivered at least10% p.a. sales growth over the last 3 years and now trade at least 25% below analysts’ target price. Sector focus: IT Services–Smaller and more specialised firms have also fallen but lessindiscriminately and upsides to analyst targets are smaller. They have also delivered decent salesand free cash flow growth recently. Sector focus: Professional services–It’s been a similar picture for this industry group, but withweaker upside potential based on analyst targets. So, despite the spillover to other servicecompanies, this implies that the software industry has been disproportionately hit. Other notable decliners in the US featured companies involved in real estate, gaming andfinancial research, showing the depth and breadth of markets’ AI disruption fears 03.Selected European stocks European software European IT services European professional services 04.What our equity analysts say What our equity analysts say: US •US software:•“With software stocks down ~19% since the beginning of the year and multiples compressing 21% given AI fears, we take this opportunity to examine the space via a quality scorecard and against GAAP & SBC adjusted FCF multiples to look for names that would appear more oversold. Based on this, we feel the market isoverly discounting the disruption to CLBT, CRM, INTU and NOW.•“While difficult to disprove the bear narrative in software given fears are more aboutgenAIimplications for the industry in the out years, we contend that anymeaningful disruption will likely play out over a much longer timeline than investors anticipate. We remain steadfast in ourbelief that vibe-coded replacements forpackaged software are unlikely to gain traction anytime soon. However, we are more concerned about switching costs in software being eroded by cheaper codegeneration, reduced integration costs, and easier change management, which could impact software economics going forward”Software: Valuing the Opportunity through AI Disruption Lenses-Deutsche Bank Research •US Payments, Pro