Wave of AI Intangible Investments Will Spawn New Generation of Superstar Enterprises

Overview

This Goldman Sachs macro research report explores how AI Agent technology will drive a wave of intangible capital investment and spawn a new generation of "superstar" enterprises. The core conclusion is: productivity gains from AI require significant non-hardware investment (including data infrastructure, software, and organizational restructuring); these intangible investments are currently likely underestimated by GDP statistics; companies capable of effectively investing in and deploying AI agents will pull away from peers, gaining higher market share, lower labor costs, and higher investment returns, becoming winners in the next economic cycle.

Core views

AI investment is not just hardware spending; non-hardware intangible investment is heating up quickly. The report states US AI-related capex has reached $360 billion (1.1% of GDP), with global hyperscaler data center capex at $400 billion in 2025, expected to exceed $700 billion in 2026. However, more critical is non-hardware investment: Current IT budget allocation shows US AI transformation-related labor costs reached $153 billion/year, executive time allocation suggests organizational capital investment of about $40 billion/year. Extrapolating workforce restructuring costs, total workforce reorganization during the entire AI adoption cycle may cost $80-90 billion. Intangible investment has a historical statistical relationship with hardware investment. Goldman Sachs used EU KLEMS database regression analysis to find that every $1 of hardware investment historically drives $2 of intangible investment (where $1.30 is used for data/software, $0.50 for organizational spending, $0.20 for other intangibles). Applying this relation to recent AI hardware investment surge points to approximately $700 billion in the US and about $1 trillion globally in supporting intangible investment. This estimate may overstate current levels because the AI hardware cycle is unusually front-loaded, and investment favors hyperscale data centers over end-users. Intangible investment will lead to productivity J-Curve effects. Citing David(1989) and Brynjolfsson et al (2021) research frameworks, the report notes that rapid technical transitions (especially General Purpose Technologies GPT) often produce a "J-Curve": GDP and productivity are initially underestimated because enterprises divert resources to intangible assets required for technical transition (business process restructuring, workforce retraining, data resource accumulation, etc.). These "self-produced" investments are recorded as costs rather than investments in national accounts; only when intangible capital stock begins producing measurable output does productivity surge appear in official statistics. The report estimates that organizational investment related to AI (approx $50 billion/year) is explicitly excluded from US GDP, implying US GDP is underestimated by at least 0.2%; if the statistical relationship holds, the underestimate could reach 2%. Companies effectively investing in intangible capital will become "superstars". Citing Haskel and Westlake(2017) "Capitalism Without Capital" Four S framework (Scalability, Sunkenness, Spillovers, Synergies), noting intangible investment has high fixed cost, low marginal cost characteristics, allowing market leaders to capture larger revenue share. Data shows rise in top "superstar" enterprise revenue share synchronized with increase in intangible capital over past 40 years. Industry concentration increased continuously over past 30 years, especially in industries where intangible capital (especially organizational capital) share increased most. Regression analysis shows for every 1 percentage point increase in intangible capital share, labor income share declines 0.2-0.3 percentage points in subsequent 2-4 years.

Analysis framework

Goldman Sachs' analysis follows the logic chain "Tech Transformation -> Investment Structure -> Economic Impact -> Market Implications". First, distinguishes hardware vs non-hardware components of AI investment, noting the latter (intangible capital) is often overlooked but substantial. Second, estimates current non-hardware investment scale via four methods: IT job AI concentration, executive time allocation, workforce restructuring costs, historical statistical relationship between hardware-intangible investment. Third, places AI non-hardware investment in long-term intangible capital trend context, citing EU KLEMS data showing G10 economies' intangible investment approaching traditional capex scale. Fourth, uses productivity J-Curve theory to explain why current productivity growth may be underestimated. Finally, combines "Four S" framework and industry concentration data to derive the conclusion that companies effectively investing in AI intangible capital will become the next generation of superstars.

Methodology notes

• Industry/Industrial Analysis FrameworkSupply and Demand Framework

  Hardware and Intangible Capital Support Relationship within General Purpose Technology (GPT) Investment Cycle

  Report finds stable multiple relationship between hardware and intangible investment (1:2) through historical regression analysis, a common method for analyzing tech investment waves, helping understand full scope of AI investment not just hardware spending.

• Cycle and Prosperity FrameworkProsperity Turning Point Analysis

  Productivity J-Curve Effect

  Resources diverted to internal intangible investment (not counted in GDP) during early tech transformation stage lead to underestimated productivity and GDP; only when intangible capital starts producing measurable output does upward turning point appear. This helps explain why current macro data might underestimate real productivity improvements.

• Competition and Strategy FrameworkMoat / competitive advantage

  Intangible Capital 'Four S' Characteristics (Scalability, Sunkenness, Spillovers, Synergies)

  Intangible assets possess features such as non-rivalry (can be used by multiple people simultaneously), irrecoverability, easy spillover, and synergistic value addition; these characteristics allow pioneer investors to establish competitive advantages and expand market share.

• Industry/Industrial Analysis FrameworkIndustry Concentration Analysis

  Positive Correlation Between Intangible Capital Investment and Industry Concentration

  Report demonstrates data proving industries with highest increase in intangible capital share also saw most significant increase in market concentration. This is a common perspective for analyzing evolution of industry competitive landscape.

• Company Fundamentals and Financial FrameworkProfit Quality Analysis

  Negative Relationship Between Intangible Capital Share and Labor Income Share

  Regression analysis shows for every 1 percentage point increase in intangible capital proportion, labor income share declines 0.2-0.3 percentage points within 2-4 years; this helps understand how tech investment affects corporate cost structure and profit margins.

• Valuation methods

  Corrado-Hulten-Sichel(2005) Estimation Method for Organizational Capital Investment

  Methodology for converting executive time allocation into organizational capital investment amounts: assuming executives spend 20% time on organizational innovation, among which 35% on AI, thereby推算 AI organizational investment scale. This is academic standard method for estimating self-produced intangible investment in national accounts.

Asset mapping & comparison

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

• Snowflake

  Data management and infrastructure provider, benefiting from data infrastructure investment driven by AI transformation

  Strengths

  Revenue growth more than tripled since ChatGPT emerged in 2022

  Comparison

  Leading in data management field alongside Databricks, Palantir

• Databricks

  Data management and infrastructure provider, benefiting from data infrastructure investment driven by AI transformation

  Strengths

  Revenue growth more than tripled since ChatGPT emerged in 2022

  Comparison

  Leading in data management field alongside Snowflake, Palantir

• Palantir

  Data management and infrastructure provider, benefiting from data infrastructure investment driven by AI transformation

  Strengths

  Revenue growth more than tripled since ChatGPT emerged in 2022

  Comparison

  Leading in data management field alongside Snowflake, Databricks

• Amazon AWS

  Cloud service provider, benefiting from accelerated cloud migration driven by AI transformation

  Strengths

  Cloud revenue doubled and grew more than 100% since 2022

  Comparison

  Major cloud service provider alongside Microsoft, Alphabet, Oracle

• Microsoft

  Cloud service provider, benefiting from accelerated cloud migration driven by AI transformation

  Strengths

  Cloud revenue doubled and grew more than 100% since 2022

  Comparison

  Major cloud service provider alongside Amazon AWS, Alphabet, Oracle

• Alphabet

  Cloud service provider, benefiting from accelerated cloud migration driven by AI transformation

  Strengths

  Cloud revenue doubled and grew more than 100% since 2022

  Comparison

  Major cloud service provider alongside Amazon AWS, Microsoft, Oracle

• Oracle

  Cloud service provider, benefiting from accelerated cloud migration driven by AI transformation

  Strengths

  Cloud revenue doubled and grew more than 100% since 2022

  Comparison

  Major cloud service provider alongside Amazon AWS, Microsoft, Alphabet

Key data

• US AI-related capex$360 billion (1.1% of GDP)AI-related capital expenditure in US national accounts
• Global hyperscaler data center capex2025: $400 billion, 2026 expected >$700 billionHyperscaler capital expenditure forecast
• US AI-related labor costs$153 billion/yearAI transformation labor costs implied by current IT budget allocation
• US AI organizational capital investment$40 billion/yearOrganizational capital investment implied by executive time allocation
• Total labor restructuring cost for AI adoption cycle$80-90 billionCumulative estimate extrapolating from current restructuring costs
• Multiplier of hardware investment on intangible investment1:2Every $1 hardware investment drives $2 intangible investment ($1.30 data/software + $0.50 org + $0.20 other)
• Global AI non-hardware investment estimateApprox $1 trillionProjection applying historical statistics to recent AI hardware investment
• Intangible investment share of total investment in G10 economies48%EU KLEMS data, US and UK over 50%
• Impact of intangible capital share on labor income shareEvery 1pp increase leads to 0.2-0.3pp declineImpact within subsequent 2-4 years
• Underestimation magnitude of US GDPAt least 0.2%Could reach 2% if statistical relationship holds
• Estimated DM economy productivity improvement15% after full adoptionGoldman Sachs baseline estimate for labor productivity and GDP level
• Enterprise value growth of data management companiesFrom less than $100 billion in 2022 to over $650 billion by end of 2025Combined Snowflake, Databricks, Palantir
• Cloud service revenue growthFrom $200 billion in 2022 to over $500 billion todayCombined AWS, Microsoft, Alphabet, Oracle

Impact & implications

The report believes this has two layers of implications for investors. First, analysis supports equity analyst conclusions: data infrastructure and AI deployment/orchestration are key factors unlocking AI economic value; companies focusing on these areas will benefit from continued enterprise investment. Second, companies today investing in necessary data, labor, and organizational infrastructure to effectively deploy AI and AI agents will likely become the next generation of "superstar" enterprises enjoying excess valuations. The report simultaneously prompts two cautions: First, analysis does not cover rent distribution within AI tech stack (semiconductor or base model providers may become true AI superstars); Second, automation and business process encoding should improve productivity for most companies during AI transformation; economic fundamentals imply part of end-user enterprises will become excess winners, but AI will still bring broad-based economy-level productivity improvements.

Risks

• Market power distribution within AI tech stack highly uncertain; semiconductor or base model providers may capture most long-term value
• AI hardware investment tilts towards hyperscale data centers rather than end-users, potentially leading to smaller current intangible investment multiplier
• AI hardware cycle unusually front-loaded; non-hardware investment may heat up gradually only during adoption phase
• Many potential AI use cases currently lack cost-effectiveness
• Organizations need to resolve data security and human supervision issues before fully autonomous deployment of AI agents

What to watch

• Changes in AI-related IT budget share (currently approx 20-40%)
• Frequency of mentions of corporate AI and organizational restructuring keywords in earnings conference calls
• Revenue growth of data management and infrastructure companies (Snowflake, Databricks, Palantir, etc.)
• Revenue growth of major cloud service providers (AWS, Microsoft, Alphabet, Oracle)
• Cost of AI-driven labor restructuring and number of affected employees
• Correlation between changes in industry concentration and intangible capital investment
• Statistical updates on intangible investment in US national accounts