资产管理中的生成式人工智能：人工智能....迟早的事——量化2030年前的AI支出、效率及利润率提升

Gen AI in Asset Management: Artificial Intelligence.... Sooner orLater- Quantifying AI spend, efficiency & margin upside by 2030 This report is part of a global series in which our analysts consider what their sector will looklike by the early 2030s when AI has been fully integrated and commercialized. There arebroadly two schools of thought running through our research on AI globally. The “Sooner”camp generally believes that Artificial Intelligence is a game-changer for their sector, rightnow. The “Later” camp believes that Artificial Intelligence is a game-changer for theirindustry… just not yet. We are in the ‘Sooner’ camp as we believe AI implementation is a ‘now’problem for asset managers and the benefit of AI could start accruing from 2028 onwards. Rupal Agarwal+65 6326 7641rupal.agarwal@bernsteinsg.com Cheng Zhang, CFA, CQF+852 2123 2636cheng.zhang@bernsteinsg.com In our last report, Gen AI in Asset Management...The AI-Native Asset Manager of 2030,we presented our thoughts on AI’s impact on asset management industry, the marketstructure, alpha generation and the profitability paradox. By 2030, we expect AI-nativeasset managers to emerge where AI drives idea generation, research synthesis, portfolioconstruction, and continuous monitoring and analysts elevate into research directors andAI supervisors, while PMs evolve into AI orchestrators focused on governance, tail-riskmanagement, and conviction building. In this report, we quantify the impact of AI on techbudgets, industry efficiency, cost save and margins. We forecast AI spend in asset management to reach USD32.5bn by 2030. With globalAUM expected to reach c.USD200 trillion by 2030 (based on PWC estimates), we assumeS-curve adoption for AI in asset management, calibrated at 55% adoption rate in 2026.Consequently, Gen AI addressable AUM is expected to grow from 51trn USD in 2025 to176trn USD in 2030 ie. 28% CAGR. We expect AI spend intensity to increase for the next2yrs split by build/buy/hybrid deployment approach, before consolidating and stabilizingby 2030. Our model estimates, industry-wide Gen AI spend to grow at 29% CAGR, fromUSD9bn in 2025 to USD32.5bn by 2030 (2028 being peak AI spend year at 36bn USD). We estimate industry efficiency gain to increase from 0% in 2025 to 21% by 2030.We believe, 2028 could be an inflection year as AI led efficiency gains are likely toexceed the incremental spend on AI. So far, industry indications of AI’s potential impacton efficiency is between 20%-40%. We model 34% steady-state efficiency gain usingcost base by function; assuming highest efficiency gain (40%) for middle/back office/tech functions, 35% for central functions and 30% for front office. We expect divergentefficiency gain for AI leaders, i.e., companies that started AI implementation ~3yrs back, vs.AI laggards that are just getting started. AI leaders are likely to see efficiency gain of 8% in2026 vs. 0% for laggards and increasing to 26% vs. 17% by 2030. We expect AI led efficiency to result in margin improvement from 32.4% in 2026to 44.6% by 2030 and cost/income ratio decline from ~68.5% to 56% by 2030.Weexpect AI efficiency gains to completely offset the headwinds from fee pressure and resultin margin improvement for AI leaders to 48% by 2030 vs. laggards to 41%; with overall AIled cost save reaching c.200bn USD by 2030. Over the last decade, the asset managementindustry has been stuck at 68.5% cost/income ratio on an average. We expect AI benefitsto kick-in post 2028 with industry cost to income ratio declining to 63.2% and eventuallyimproving to c.56% by 2030. DETAILS AI’s adoption in the asset management industry has been on a steady rise, roughly moving from 35% adoption in 2024 toreaching ~60% adoption (the HF adoption is even higher). Within the industry, the divergence remains high as some globalasset managers have emerged as AI leaders, having started their AI journey back in 2023. However, bulk of the industry is stillin early phases of adoption where AI is being used to enhance day to day activities of research, writing, summarization, codingetc. We believe the industry has already reached an inflection point where AI laggards have to really catch up fast to remainrelevant; and 2026 is that last window to act. The journey towards agentic AI and automated workflows has begun, and webelieve by 2030, AI native asset managers would emerge which are either truly AI-first or the traditional asset managers whowould have completely transformed their investment processes and upskilled/re-skilled their workforce to make them AI ready.In our last report, The AI-Native Asset Manager of 2030, we presented the contours of an AI-native investment process, theexpected workforce transformation, the alpha-generation opportunities for an AI world and how AI is expected to have a hugeimpact on market structure and its participants but also act as a survival mechanism for active managers. In this note, we extendour analysis to quantify the impact