The weather is good today The weather is good today.

In the past year, AI has become one of the hottest topics in the market. However, this wave of enthusiasm is now under more scrutiny. Although stocks related to AI infrastructure have performed strongly, investors still have concerns about the return on investment in AI.

Goldman Sachs pointed out in their latest report that looking at the four-stage framework of AI trading, Nvidia is an obvious beneficiary. Infrastructure stocks in stage 2 performed well, while software and IT service company stocks fluctuated in stage 3, and there are doubts about AI application stocks in stage 4.

Goldman Sachs mentioned in their report:

In the past four quarters, the capital expenditure related to AI of super-large-scale cloud computing service providers (such as Amazon, Meta, Microsoft, and Google) reached $357 billion, accounting for 23% of the total expenditure of the S&P 500.

Capital expenditures and research and development investments in the TMT industry still lag behind those in the technology bubble era relative to cash flow, but analysts' revenue expectations for these super-large-scale enterprises have not correspondingly increased with the increase in investment.

Goldman Sachs believes that the second quarter earnings season will be an important test, and investors should pay attention to the revision of revenue forecasts, which will be a key factor in assessing the sustainability of AI investment trends. Goldman Sachs expects to achieve an investment return rate similar to that in recent history, and super-large-scale enterprises need to generate approximately $335 billion in revenue by 2025, otherwise they may face the risk of valuation downgrade.

Exceptional performance in stage 2, fluctuation in stage 3, doubts about stage 4.

Goldman Sachs previously proposed the four-stage framework for AI trading:

Stage 1: Nvidia is the most obvious beneficiary;

Stage 2: Focused on AI infrastructure, including semiconductor companies beyond Nvidia, cloud service providers, data center REITs, hardware and equipment companies, security software stocks, and utilities;

Stage 3: Companies that are expected to generate incremental revenue through AI, mainly software and IT service companies;

Stage 4: Companies that have the potential to significantly increase profits through widespread adoption of AI to improve productivity;

Specifically, Goldman Sachs said:

Driven by Nvidia, stage 2 ushered in a round of gains, which rose sharply by 26% this year so far; while there has been a recent decline in stage 3 stocks due to investors' doubts about the monetization of AI, which fell 19% between February and May; Stage 4 has hardly changed in terms of revenue, valuation, and performance.

The second quarter earnings season will be an important test.

Goldman Sachs said that the second quarter earnings season will be an important test of investors' optimistic expectations:

Although analysts expect that Nvidia's revenue growth will slow down from 265% in the fourth quarter of 2023 to 25% in 2025, the stock's valuation is still higher than its 10-year average level, which is a common trend among large technology stocks.

Our sales expectations are higher than the general expectations. Many AI-related companies will release their financial reports at the end of July, Nvidia is expected to release its report by the end of August, and many software companies in stage 3 will release their financial reports before the end of August.

Investors are increasingly concerned about the problem of 'overinvestment' in the AI field, especially in super-large-scale enterprises. Goldman Sachs pointed out that compared with the technology bubble, AI capital expenditures are still relatively small.

Compared with the technology bubble era, current capital expenditures and research and development investments are still lower relative to the company's cash flow, and the profitability of TMT stocks is stronger.

In the past four quarters, the four largest super-large-scale companies (Amazon, META, Microsoft, and Google) spent $357 billion on capital expenditures and R&D, accounting for 23% of the total expenditure of the S&P 500, and a large part of the incremental investment was due to AI.

During the most serious technology bubble, due to excessive investment in many telecommunications stocks, TMT stocks used more than 100% of their operating cash flow (CFO) for capital expenditures and research and development.

In contrast, today's leading TMT stocks are very profitable. Although the proportion of capital expenditures and R&D to sales has increased, the proportion of cash flow is lower at 72%, and with the recovery of profits, this proportion continues to slightly decrease, compared to a median of 67% over the past 40 years.

However, investors remain skeptical of the potential "return on investment" for AI companies, questioning whether massive investments can bring enough sales growth and revenue. Goldman Sachs said:

The consensus forecast for capital expenditures and R&D for mega-cap companies has increased by 7% in 2024, and by 9% from 2025 to the present. In US dollars, analysts expect these mega-cap companies to spend an additional $27 billion and $38 billion for capital expenditures and R&D in 2024 and 2025, respectively, compared to early years. However, analysts only increased sales forecasts for 2025 and 2026 by $17 billion and $19 billion, respectively.

In the past five years, mega-cap companies have on average converted 31% of their past three-year capital expenditures and R&D expenditures into revenue. This total sum implies that these companies need to generate $335 billion in revenue in 2025 to achieve similar investment returns as in the recent past, and revenue levels need to grow by 16% from 2024.

Large-scale enterprises will ultimately be required to demonstrate that their investments will generate revenue and profits, failing to meet these expectations can cause significant declines in valuation and stock prices.