Bernstein estimates: AI spending in the asset management industry may reach US$32.5 billion by 2030, with significant upside for efficiency and margins

Overview

In this Asia Quantitative Strategy report, Bernstein extends its framework for the “AI-native asset manager by 2030,” focusing on quantifying the impact of generative AI on technology budgets, addressable AUM, efficiency gains, cost savings, and margins in the asset management industry. The report argues that the industry has entered a stage where no CIO or CTO can completely avoid investing in AI. In the short term, AI looks more like a cost center, but around 2028 efficiency benefits are expected to exceed incremental AI spending, driving significant improvements in industry margins and cost-income ratios by 2030.

Core views

The report’s core views include: first, Gen AI-addressable AUM in the asset management industry will expand rapidly, driving AI spending from about US$9 billion in 2025 to US$32.5 billion in 2030; second, AI will reshape the investment process by directly participating in idea generation, research synthesis, portfolio construction, and continuous monitoring, while human analysts and PMs shift toward research oversight, governance, tail-risk management, and conviction building; third, efficiency gains will not be evenly distributed, and large asset managers that complete enterprise-level AI deployment earlier will benefit first; fourth, AI cost savings are expected to offset downward pressure on management fees and lift industry margins from 32.4% in 2026 to 44.6% in 2030.

Analysis framework

The report uses a combination of top-down and functional decomposition approaches: it first estimates industry AI spending based on global AUM, AI adoption rates, and spending intensity per unit of AUM; then projects spending intensity across three deployment paths—buying third-party solutions, building in-house, and hybrid deployment; subsequently estimates steady-state efficiency improvements across cost pools such as research/PM, clients and distribution, IT/data, middle and back office, and central functions; and finally models the divergence in efficiency, margins, and cost-income ratios among AI leaders, expanders, and laggards.

Methodology notes

• adoption curveS-curve adoption-rate model

  Gen AI-addressable AUM = total AUM × AI adoption rate

  The report assumes that the asset management industry’s AI adoption rate rises along an S-curve and calibrates it to a 55% adoption rate in 2026; based on global AUM of about US$200 trillion in 2030, it derives addressable AUM rising from US$51 trillion in 2025 to US$176 trillion in 2030.

• spending intensitybuy/build/hybrid deployment model

  Estimate AI spending intensity per unit of AUM by deployment method

  The report divides AI deployment into three categories: buy, build, and hybrid. The initial mix is 12.5%, 24%, and 63.5%, and is assumed to move toward an equilibrium structure of 13.5%, 27%, and 59.5% by 2028; in-house build costs are more front-loaded from 2026 to 2028 and then shift to maintenance costs.

• efficiency estimationcost-function weighted efficiency model

  Derive steady-state efficiency improvement by weighting functional cost shares and AI automation potential

  The report assumes a typical asset manager cost structure of 30% research/PM, 25% clients/distribution, 15% IT/data, 20% middle and back office, and 10% central functions; efficiency improvement is highest in middle/back-office and technology functions at about 40%, followed by central functions at 35% and front office at about 30%, resulting in a weighted steady-state efficiency improvement of about 34%.

• competitive segmentationAI leaders/expanders/laggards segmentation

  Differentiate efficiency and margin paths by AI deployment maturity and AUM share

  The report divides the industry into AI leaders, expanders, and laggards, representing about 20%, 45%, and 35% of global AUM, respectively; leaders achieve efficiency benefits earlier, while laggards fall behind due to insufficient production-grade AI tools.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• global asset management companies

  direct beneficiaries of AI investment and efficiency improvement

  Strengths

  Can achieve automation and productivity gains across research, portfolio management, compliance, middle and back office, customer service, software development, and other functions.

  Weaknesses

  Short-term ROI on AI investment is unclear, and spending is relatively front-loaded around 2026 to 2028.

  Comparison

  Compared with institutions that have not deployed AI at scale, managers with enterprise-level AI capabilities are more likely to improve margins first.

  Risks

  The model assumptions depend on AI adoption rates, cost curves, governance quality, and the speed of organizational change; actual results may be lower than expected.

• AI leaders

  early beneficiaries of efficiency and margin improvement

  Strengths

  Already have production-grade Agentic AI, stronger technology infrastructure, and greater ability to amortize costs across a larger AUM base.

  Weaknesses

  Need sustained high-intensity technology investment and must bear pressures related to AI governance, systemic risk, and model reliability.

  Comparison

  The report expects leaders to achieve a 26% efficiency improvement and about a 48% margin in 2030, outperforming expanders and laggards.

  Risks

  If expanders accelerate investment and adopt mature third-party solutions, leaders’ advantages may narrow.

• AI expanders

  catch-up beneficiaries

  Strengths

  Already have AI deployment and transformation paths and may quickly replicate some leaders’ experience through hybrid deployment.

  Weaknesses

  Lag leaders by about one year and lack sufficient maturity in enterprise-level integration and governance.

  Comparison

  Efficiency improvement is expected to be about 22% in 2030, with margins of about 45%, below leaders but significantly above non-transformed institutions.

  Risks

  If investment is insufficient or organizational change progresses slowly, they may slide into the laggard group.

• AI laggards

  potentially passive and pressured group

  Strengths

  Can adopt external solutions at lower trial-and-error cost after the vendor ecosystem matures.

  Weaknesses

  Lack production-grade AI tools, experience delayed efficiency gains, and may struggle to offset fee compression.

  Comparison

  Efficiency improvement is expected to be about 17% in 2030, with margins of about 41%, below leaders and expanders.

  Risks

  May be squeezed by large platforms and more agile boutique managers, putting market share and margins under pressure.

• banks, custodians, and large financial institutions

  reference samples providing AI spending and efficiency cases

  Strengths

  Have disclosed sizable technology budgets, AI platform deployments, developer tools, and customer service automation cases.

  Weaknesses

  Most institutions have not separately disclosed AI budgets, limiting direct comparability with the asset management industry.

  Comparison

  Large universal banks treat AI as a firm-wide infrastructure investment, while pure asset managers tend to view AI more as an operating and distribution cost tool.

  Risks

  When extrapolating bank cases to asset management industry efficiency, the transferability of specific business functions may be overestimated or underestimated.

Key data

• 2030 global AUM assumptionabout US$200 trillionBased on PwC estimates and used as the basis for calculating AI-addressable AUM.
• Gen AI-addressable AUMUS$51 trillion in 2025 and US$176 trillion in 2030Corresponding to a CAGR of about 28%.
• Industry Gen AI spendingabout US$9 billion in 2025 and US$32.5 billion in 2030Expected to peak at about US$36 billion in 2028.
• Starting point for AI spending intensity in 2025about US$177 per US$1 million of AUMDerived from asset management industry revenue and financial institutions’ AI investment as a share of revenue.
• Industry efficiency improvement3% in 2026, 21% in 2030, and 34% in the long-term steady state2028 is expected to be the inflection point at which efficiency benefits exceed incremental AI costs.
• Efficiency improvement for AI leaders8% in 2026 and 26% in 2030Leaders are typically large asset managers that have undertaken about three years of enterprise-level AI deployment.
• Efficiency improvement for expanders and laggardsabout 22% and 17%, respectively, in 2030Expanders lag leaders by about one year, and laggards lag by about two years.
• Industry margin32.4% in 2026 and 44.6% in 2030AI-driven cost savings are the main source of improvement.
• Cost-income ratiohistorical average of about 68.5%, improving to about 56% in 2030AI benefits are expected to become more visible after 2028.
• AI cost savingsabout US$200 billion in 2030The report believes this can significantly offset pressure from asset management fee compression.

Impact & implications

If the report’s assumptions materialize, the competitive advantages of the asset management industry will shift further from pure scale, distribution channels, and traditional investment research capabilities toward enterprise-level AI infrastructure, data integration, governance frameworks, and human-machine collaboration capabilities. Large leaders may widen the efficiency gap in the short term through scale and infrastructure advantages; agile small boutique managers may also benefit from faster transformation; medium-sized institutions face simultaneous pressure from high operational complexity, insufficient technology investment, and fee compression.

Risks

• The actual ROI of AI investment may be lower than the report’s model assumptions, especially during the spending peak from 2026 to 2028.
• Generative AI adoption rates, deployment mix, and spending intensity per unit of AUM are all model assumptions; if industry adoption slows, 2030 spending and efficiency estimates would be revised downward.
• AI may reduce information asymmetry and blur the boundary between quantitative and fundamental investing, but it may also increase crowded algorithmic trading and systemic risk.
• The reliability, auditability, and regulatory acceptance of model automation in research, portfolio management, compliance, and customer service remain uncertain.
• Asset management fee compression may be faster or more severe than AI-driven efficiency improvement, thereby weakening margin improvement.
• Medium-sized asset managers may face greater pressure due to insufficient technology investment, higher organizational complexity, and limited ability to amortize costs through scale.
• The report extensively uses industry cases, public disclosures, and external estimates; differences in definitions and disclosure quality may lead to estimation errors.

What to watch

• Whether AI spending continues to rise from 2026 to 2028 as described in the report and peaks around 2028.
• Whether large asset managers disclose clearer AI budgets, efficiency gains, cost savings, and margin contributions.
• The speed of production-grade implementation of AI in research/PM, compliance, middle and back office, IT development, and customer service.
• Whether AUM shares among AI leaders, expanders, and laggards undergo structural changes.
• Whether management fee compression continues to intensify, and whether the ratio of AI savings to fee headwinds approaches the 6.5x forecast in the report.
• Whether regulators tighten requirements for AI investment advice, model governance, client suitability, and systemic risk.
• Whether the vendor ecosystem matures around 2028, causing the cost curves for buy, build, and hybrid deployment to stabilize.