Goldman Sachs shifts the AI investment discussion from “how many tasks can be replaced” to “how many profit pools can be reshaped”
AI summary card
Goldman Sachs shifts the AI investment discussion from “how many tasks can be replaced” to “how many profit pools can be reshaped”
Using digital advertising as the first case study, the report argues that AI could disrupt multiple profit pools worth tens of billions of dollars, including ad budget migration, creative generation, ad-tech intermediaries, and the agency ecosystem, and that in the medium term this is more favorable for scaled platforms such as GOOGL and META that possess capital, engineering capabilities, and first-party data.
- The five major Western listed hyperscale cloud providers spent about $477 billion on AI-related capex in 2022-2024, and Goldman Sachs expects this to rise to about $1.15 trillion in 2025-2027E; investor focus is now shifting to whether these investments can generate sufficient returns.
- The report argues that AI’s impact should not be judged only by the share of automatable tasks, but by whether AI can disrupt industry profit pools that are large and profitable.
- Digital advertising is one of the sub-sectors, outside of cloud computing, where AI product innovation and end-market adoption are most advanced within the research coverage; AI is already being used in content creation, AI assistants, content recommendation, optimization engines, ad execution, and modeled conversions.
- GOOGL’s Performance Max and META’s Advantage+ are seen as the most successful current examples of integrating multiple AI advertising capabilities into a single product; AppLovin Axon 2.0 and Pinterest Performance+ are also gaining adoption and share.
- The report estimates that the main advertising profit pools AI could affect include: about $170 billion from driving ad budgets toward digital channels, about $114 billion from automating ad creative generation, about $25 billion from consolidating ad-tech intermediaries, and about $161 billion from impacting the advertising agency ecosystem.
Report interpretation
Overview
This report is the first industry deep-dive in Goldman Sachs’ “AI Profit Pools” series, using the global (ex-China) digital advertising industry as a case study to discuss how AI is gradually transmitting from infrastructure investment to the platform and application layers and reallocating existing industry profit pools. The report notes that since the launch of ChatGPT in November 2022, the market initially focused on large-model capabilities, training investment, and infrastructure expansion; but about 2.5 years later, as large cloud providers and private AI companies have deployed huge amounts of capital, investors are increasingly focused on whether this spending can generate sufficient returns over the medium to long term.
Core views
The report’s core view is that AI’s investment value should not be measured only by “which tasks can be replaced by AI,” but by assessing whether AI can disrupt existing profit pools of sufficient scale. In advertising, AI has already been rapidly deployed across creative generation, ad placement, audience targeting, performance attribution, and personalization. Goldman Sachs believes AI could lower creative production costs, improve ad delivery efficiency, continue to push budgets from traditional media toward digital channels, and allow large closed-ecosystem platforms to absorb more value from ad-tech and intermediary services. Over the medium term, large platforms such as GOOGL and META, which have capital, engineering resources, and first-party data, are more likely to benefit; smaller platforms also have long-term upside optionality, but it is harder for them to replicate the scale advantages of the large platforms.
Analysis framework
The report uses a profit-pool analysis framework, breaking the advertising value chain into ad budget migration, creative generation, ad-tech intermediaries, the agency ecosystem, and new TAM expansion opportunities, and estimating the potential scale of disruption in each area. For digital-channel migration, the report compares the pace of improvement in global (ex-China) advertising digital penetration under the current base case versus an AI scenario; for creative generation, it uses 2024 global (ex-China) advertising spend as the base and assumes different shares of ad spend attributable to creative production costs across channels and media formats; for workflow impact, it compares the current advertising process with the future AI-enabled state.
Methodology notes
Shifting from task substitution to profit-pool disruption
The report argues that if AI can only replace many low-value tasks but cannot capture sufficient revenue, the investment is unsustainable; conversely, even if AI changes only a small number of tasks, it can still have a significant industry impact as long as those tasks correspond to large profit pools.
Comparison between base case and AI scenario
Using forecasts for global (ex-China) advertising spend as the baseline, the report compares the difference between digital advertising penetration rising by about 170bps per year in 2025-2028E under the base case and about 350bps per year under the AI scenario, thereby estimating an incremental digital advertising opportunity of about $170 billion.
Estimating the share of creative production costs by channel
Based on approximately $856 billion in 2024 global (ex-China) advertising spend, the report assumes that creative generation costs account for 0%-30% of ad spend across formats such as digital and non-digital channels, search, digital video, linear TV, and connected TV, and derives an addressable profit pool of about $114 billion.
Four categories of advertising and marketing workflows
The report divides advertising and marketing workflows into campaign planning and strategy, creative development, media buying and optimization, and CRM and personalization, analyzing how AI can improve efficiency, automate execution, reduce costs, and enhance ROAS.
Asset mapping & comparison
Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).
- META PLATFORMS INC / META Advantage+Potential core beneficiary
- Strengths
- It has a massive user base, first-party data, an advertiser base, and AI advertising automation products; Advantage+ has already formed productized capabilities in audience targeting, placement selection, and creative optimization.
- Weaknesses
- It depends on existing consumer time spent and mobile behavior patterns, and it needs to continue investing in AI infrastructure and model capabilities.
- Comparison
- Similar to GOOGL, META is a scaled closed-ecosystem platform; compared with smaller platforms, its data and engineering resource advantages are stronger.
- Risks
- AI may change consumer discovery and purchasing paths, thereby disrupting ad budget allocation and ROI; regulation and privacy constraints may also affect data usage.
- Alphabet / Google Performance MaxPotential core beneficiary
- Strengths
- Performance Max can automate creative creation, bidding, targeting, and placement optimization across search, display, YouTube, and other channels, leveraging Google’s vast user data and AI capabilities.
- Weaknesses
- Advertisers may worry about the platform becoming a black box and losing control, and traditional media may also improve efficiency through AI, reducing the incremental nature of budget migration.
- Comparison
- Along with META Advantage+, it is regarded by the report as one of the most successful integrated AI advertising products at present.
- Risks
- If AI search or agents change how users discover information, Google’s existing search advertising funnel may face structural change.
- AppLovin / Axon 2.0Beneficiary among niche platforms
- Strengths
- Focused on mobile app advertising, with strong machine-learning models and ROI optimization capabilities, making it well suited to user acquisition and monetization for app developers.
- Weaknesses
- Its coverage is skewed toward app advertising, its use cases are relatively narrow, and it faces industry risks such as ad fraud.
- Comparison
- Compared with GOOGL and META, its scope is narrower, but it has differentiation in the mobile app advertising niche.
- Risks
- If large platforms further integrate advertising automation capabilities, AppLovin’s independent growth runway may be squeezed.
- Pinterest / Performance+Beneficiary among niche platforms
- Strengths
- Its visual discovery use case, users with relatively high purchase intent, and AI-driven targeting capabilities help attract advertisers in industries such as fashion, home, and food.
- Weaknesses
- Its user scale and reach are lower than those of META and Google, and its audience is more vertically concentrated.
- Comparison
- Like AppLovin, it is a case of progress in AI advertising tools among smaller platforms; compared with large platforms, its scale advantage is limited.
- Risks
- Large platforms may continue to strengthen advertiser lock-in through advantages in capital, engineering resources, and first-party data, potentially limiting Pinterest’s share expansion.
- NvidiaBeneficiary of the AI infrastructure cycle
- Strengths
- The report notes that its data center compute market share rose from about 15% in 2018 to about 85% today, and that it has earned excess profits during the AI capex cycle.
- Weaknesses
- Its benefits are more concentrated at the infrastructure layer, and whether downstream applications can ultimately generate sufficient profit-pool returns remains an investor focus.
- Comparison
- Unlike advertising platforms, Nvidia has already monetized at scale ahead of AI applications and is an upstream beneficiary in the capex transmission chain.
- Risks
- If monetization at the AI application layer is insufficient, future capex growth and ecosystem reinvestment capacity may come under pressure.
- Advertising agencies and ad-tech intermediariesPotentially pressured parties
- Strengths
- They currently possess professional capabilities and client relationships across strategy, creative, delivery, measurement, verification, and client service.
- Weaknesses
- Many services may be automated or abstracted by AI platform tools, especially campaign setup, delivery optimization, attribution, and basic creative generation.
- Comparison
- Compared with large closed-ecosystem platforms, independent intermediaries rely more on fragmented workflows and third-party data; they are in a relatively disadvantaged position as privacy restrictions increase.
- Risks
- Profit pools may migrate toward platforms with first-party data and AI automation capabilities, requiring agencies to upgrade toward strategy, brand control, complex creative work, and cross-channel consulting.
Key data
- AI-related capex of the five major listed hyperscale cloud providersAbout $477 billion (CY2022-CY2024), expected to reach about $1.15 trillion (CY2025-CY2027E)Covers AMZN, MSFT, GOOGL, META, and ORCL.
- OpenAI FY2024 operating loss and ARROperating loss of about $5 billion, ARR of about $5.5 billionThe report cites media reports to illustrate unit economics pressure in the early scaling phase of AI.
- xAI FY2025 operating loss and revenue forecastOperating loss of about $13 billion, revenue of about $500 millionThe report cites media reports for comparison with early technology transformation cases such as Tesla.
- Nvidia data center compute market shareAbout 15% (2018) increased to about 85% (current)The report believes Nvidia is a key beneficiary of the AI capex cycle.
- 2024 global (ex-China) advertising spendAbout $856 billionUsed to estimate ad creative generation and digital migration opportunities.
- 2024 global (ex-China) digital advertising penetrationAbout 69%The report says digital penetration has risen by an average of about 400bps annually since 2017.
- Opportunity from AI-driven migration of ad budgets to digital channelsAbout $170 billionDerived from the difference in the pace of digital penetration improvement between the 2025-2028E base case and AI scenario.
- Disruptable profit pool in ad creative generationAbout $114 billionBased on the assumption that creative generation costs account for about 0%-30% of ad spend across different advertising channels.
- Ad-tech intermediary consolidation opportunityAbout $25 billionAI tools could automate campaign setup, delivery, measurement, and attribution, weakening the value of some intermediaries such as DSPs, SSPs, and DMPs.
- Potential impact on the advertising agency ecosystemAbout $161 billionThe report identifies the agency ecosystem as a profit pool worth many tens of billions of dollars that AI could disrupt.
- Google Performance Max adoptionSkai sample shows that in Q4'24 about 59% of U.S. advertisers used PMax, and about 16% of total U.S. Google spend flowed through PMaxData comes from Skai’s advertiser sample and is not officially confirmed by Alphabet.
Impact & implications
From an investment perspective, AI advertising tools could abstract and platformize parts of the advertising value chain, including creative, delivery, attribution, and optimization services, causing value to concentrate further in closed-ecosystem platforms with first-party data, user scale, engineering capabilities, and the ability to invest capital. For advertisers, AI could reduce creative costs, increase personalization, improve ROAS, and enable smaller advertisers to enter channels that were previously too expensive. For advertising agencies and independent ad-tech intermediaries, automation tools could lead to profit-pool compression, client budget reallocation, and a redefinition of service value.
Risks
- AI-driven ad budget migration is not a fully incremental opportunity, because the shift of ad spending from traditional channels to digital channels is already a long-term trend.
- Traditional media owners may also adopt AI to improve efficiency, meaning the benefits brought by AI may not all flow to model builders or digital platforms.
- Advertisers may not fully reinvest creative cost savings into ad purchases and may instead retain them as margin improvement.
- Advertisers may remain cautious about adopting generative AI creative tools because of concerns about brand safety, information control, and creative quality.
- AI may change desktop and mobile consumer habits, disrupting existing search, social, and purchase funnels.
- Further strengthening of large platforms may compress the profit space of smaller platforms, ad-tech vendors, and agencies.
- AI infrastructure is highly capital intensive; if application-layer revenue and margins do not expand enough to cover the investment, industry returns may fall short of market expectations.
What to watch
- Advertiser adoption rates, share of ad spend, and ROAS improvement for GOOGL Performance Max and META Advantage+.
- Whether global (ex-China) digital advertising penetration under the AI scenario can accelerate from the base-case annual increase of about 170bps to about 350bps.
- Actual adoption of generative AI creative tools in video, images, ad copy, and A/B testing, and whether advertisers reallocate cost savings into media buying.
- Whether AI advertising products from smaller platforms such as AppLovin Axon 2.0 and Pinterest Performance+ can continue gaining share.
- Revenue growth, margins, and client budget migration for advertising agencies and independent DSPs, SSPs, DMPs, and measurement/verification platforms.
- Whether privacy constraints, third-party data limitations, and first-party data advantages continue to drive advertising value toward closed-ecosystem platforms.
- The impact of AI assistants, AI search, and agentic interfaces on consumer discovery, search, and purchase paths.