Goldman Sachs' "AI Narrative Framework": Five Key Debates about AI

```

Author: Dong Jing

Source: Hard AI

After Goldman Sachs' Communacopia+ Technology Conference in September and a series of corporate and industry announcements regarding "long-term AI capacity," debates in the market about whether we have entered an "AI bubble" have grown increasingly intense.

On October 8, according to Hard AI, Goldman Sachs' TMT research team released their latest research report, which conducts in-depth analysis on five key current AI controversies, in an attempt to provide the market with a clear narrative framework for AI. These five core controversies encompass everything from consumer and enterprise AI adoption, to AI spending expectations, power infrastructure demand, and the high-profile risk of bubbles.

According to the report, consumer AI adoption is faster than expected, as ChatGPT hit a record of 700 million weekly active users in July, but monetization is still lagging behind infrastructure investment. Goldman Sachs states that at the enterprise level, AI returns on investment are still limited; despite increasing internal deployment, MIT research shows that only 5% of companies have seen measurable P&L impact from AI.

Meanwhile, Goldman notes that AI infrastructure investment has reached historic levels this year, with total capital expenditures by the world’s top five hyperscale cloud providers (Amazon, Microsoft, Google, Meta, and Oracle) expected to hit $381 billion, up 68% year-on-year. The report points out that the rapid expansion of AI workloads will sharply increase data center demand, and global electricity demand is projected to grow by more than 165% by 2030.

Regarding the highly watched risk of an AI bubble, Goldman Sachs believes that although there are similarities between the current market environment and the late 1990s, the Nasdaq 100's P/E ratio is still 46% below the peak of the internet bubble, IPO activity is also well below levels back then, and there are no conditions yet for a major pullback.

AI Consumer Adoption: Rapid Growth but Slow Monetization

Consumer AI usage is growing faster than enterprise applications and continues to rise rapidly, but AI companies' ability to monetize consumer services and generate revenue lags behind infrastructure spending to meet demand.

According to Goldman’s analysis, consumer adoption is rising swiftly, with OpenAI reporting that ChatGPT hit 700 million weekly active users in July—a record that far surpasses its monetization growth.

Based on SensorTower data analysis:

ChatGPT dominates both global and U.S. markets, leading in both monthly and daily active users.Google Gemini has become the second largest platform due to its distribution channel advantage, achieving substantial growth in its user base, but it still noticeably trails ChatGPT.Claude and Perplexity maintain relatively modest user bases.

Notably, platform differentiation is emerging. According to usage reports from each platform:

OpenAI disclosed that ChatGPT is increasingly used for non-work purposes (about a 70/30 split between non-work and work-related tasks);Anthropic reports that Claude is used most for programming/computation tasks (36% of usage), indicating that users are starting to select models according to their specific strengths.

More notably, 90% of employees from companies report regularly using personal AI tools at work, but only 40% of companies have purchased official LLM subscription services.

The report notes that a recent wave of product launches highlights the continued maturation of AI capabilities into tangible, monetizable applications. ChatGPT's instant checkout feature, which enables direct purchases within chats together with partners like Etsy, and Google's strategic collaboration with PayPal on agent-driven commerce, both underscore the likely growth of AI-driven transactions.

Goldman Sachs anticipates 2026 will be a key year for AI monetization, with crossover between advertising and commerce driving large-scale monetization.

Enterprise AI Deployment: Internal Expansion But ROI Still Needs Improvement

Although enterprises continue to deploy generative AI both internally and externally, ROI (return on investment) visibility remains low.

Internally, companies are deploying AI to drive incremental efficiency gains and support margin improvement. Goldman has observed real efficiency improvements in complex tasks such as content and software development, ad creative generation, and inventory management and dynamic pricing. Efficiency gains are reflected in higher productivity and slower hiring growth.

However, external enterprise applications that could generate revenue growth or market share shifts are progressing slowly. The MIT report on Business AI Status shows that only 5% of companies have seen measurable P&L impact. Goldman Sachs believes this is mainly due to inconsistency in AI deployment methods and oversight systems.

In consumer internet, Goldman has identified AI’s potential to disrupt profits pools worth tens of billions of dollars. Traditional ad agencies face automation threats from AI-driven platforms like Google Performance Max and Meta Advantage+, and the global ad agency ecosystem has a profit pool of about $161 billion.

At the same time, AI is making digital advertising more effective and targeted, with the potential to accelerate the shift of ad budgets from traditional to digital channels, bringing about $170 billion in incremental digital opportunities in 2025–2028.

In the software industry, Goldman Sachs’ 2025 industry conversations show the AI application ecosystem is maturing. SaaS leaders such as Microsoft, Salesforce, ServiceNow have begun to disclose specific AI contributions; the key is when these contributions can bring additional effect to growth algorithms.

Despite this, enterprise market ROI is still pending, and the large investments in building, training, and using foundational models for AI have yet to substantially penetrate enterprises in ways that bring economic returns.

AI Expenditure Forecast: Infrastructure Investment at Unprecedented Levels

Goldman Sachs foresees both sides of the AI ecosystem increasing infrastructure spending in 2025. Since publishing their June report on AI’s impact on industry profit pools, investor attention has increased on the medium- and long-term returns of hyperscale AI spending.

According to Goldman’s estimate, the top five global hyperscale cloud providers (Amazon, Microsoft, Google, Meta, and Oracle) will see capital expenditures further rise this year, with a projected total of $381 billion in 2025, up 68% year-on-year.

Most AI investment has been announced for the second half of 2025, with hundreds of billions in partnerships announced in September, such as Oracle’s $300 billion deal with OpenAI and NVIDIA’s $100 billion investment in OpenAI.

Goldman Sachs says that this further spending, combined with growing consumer demand that may soon exceed future compute supply (despite current investment), adds tension to the ongoing debate about whether hyperscale capital expenditures will deliver long-term returns.

Goldman expects that the top five hyperscale cloud providers will continue to increase AI-related capital spending to keep up with growing consumer demand, with spending for 2025–2027 estimated to rise to about $1.4 trillion.

However, the report notes a recurring theme: there remains a huge gap between current AI service demand (from users and major platform/product development needs) and current available capacity. This mismatch is most evident in the growing backlogs at cloud computing companies and should, if realized, sustainably support revenue growth within the next 2–3 years.

Power Infrastructure Demand: 165% Growth Presents Construction Challenges

Goldman Sachs’ global data center and utilities team has previously outlined that AI and non-AI workloads are having a significant impact on data center power demand.

The rapid expansion of AI workloads will sharply increase data center demand, with global electricity demand projected to grow by more than 165% by 2030. Goldman’s data center team’s previous analysis shows that global data center demand in Q2 2025 will be around 62GW, with AI accounting for 13%; by 2027, total demand is expected to reach about 92GW, with AI workloads rising to 28%.

Meeting this demand requires large-scale construction of power generating capacity. Goldman projects that by 2030, compared to 2023, data center power demand will rise by about 165%. In the US, 60% of future demand will need new generation facilities, with an incremental 72GW of capacity needed, mainly from natural gas (60%), solar (25–30%), and wind (10–15%).

Grid investment expectations have been raised from $720 billion in July to $780 billion (by 2030), an increase of $60 billion, with a focus on distribution infrastructure, and faster growth in transmission capital expenditures to sustain data center supply growth.

Bubble Risk Assessment: Similar but Not Identical to the 1990s

Goldman Sachs believes there are some similarities between the current market and the late 1990s, but it has not reached the stage that would trigger a sharp re-pricing of public market equities.

First, current valuation levels are much lower than in the late 1990s: The Nasdaq 100 is currently trading at about a 46% discount relative to the dotcom bubble era. At the end of 1999, the LTM P/E ratio was 68.4, while on October 3, 2025, it is about 37.0.

Second, IPO activity shows differences: There were 892 US IPOs from 1998–2000 compared to significantly fewer in 2023–2025, although the average transaction size has increased from $176 million in 1998–2000 to $254 million in 2023–2025. This indicates a more cautious market, with only larger, more mature companies now opting to go public.

Finally, macro conditions are also much stricter: From March 1999 to March 2000, the average yield on 10-year US Treasuries was about 6.0%, compared with about 4.3% from September 2024 to September 2025. Goldman’s macro team expects further 25bps rate cuts in October and December, which may improve capital inflows in the near term.

This article is from WeChat public account "Hard AI". For more cutting-edge AI news, click here

Risk Warning and DisclaimerThe market has risks and investment requires caution. This article does not constitute personal investment advice, nor has it taken into account individual users' special investment objectives, financial situations, or needs. Users should consider whether any opinions, views, or conclusions in this article are suitable for their specific circumstances. Investing based on them is at your own risk. ```