Goldman Sachs also came to cool AI: AI adoption trend is slowing down.

While the supergiants are still investing hundreds of billions of dollars in the field of AI hardware, the latest data from Goldman Sachs, Apollo and MIT jointly reveal an unsettling trend: the adoption of AI in enterprises is decelerating, and the adoption rate of large enterprises has even declined.

According to a report released by Jan Hatzius, chief economist of Goldman Sachs, on September 8th, although investment in AI hardware is still accelerating, the growth rate of AI adoption by US enterprises has slowed down in the third quarter.

Data shows that in the third quarter of 2025, the proportion of US companies using AI rose only slightly from 9.2% in the second quarter to 9.7%. Although it is still growing, the growth rate has slowed down. This stands in sharp contrast to the intense investment in the AI infrastructure sector. The report points out that the supergiants have already invested hundreds of billions of dollars in the construction of AI data centers, and these huge capital expenditures urgently need to prove their return on investment.

The report also shows that the adoption rate of AI in industries such as finance and real estate has increased the most, while that in the education service sector has declined. In terms of the labor market, Goldman Sachs believes that the impact remains "mild", but job substitution has already emerged in areas such as technology, design and customer service. Since the last update, layoffs caused by AI have affected 10,375 workers.

A more serious signal: The adoption rate of AI in large enterprises is declining

More worrying than the "slowdown in growth" revealed in the Goldman Sachs report is that the adoption rate of AI by large enterprises may have peaked and started to decline.

Wall Street Journal wrote that according to the analysis of Torsten Sløk, chief economist of Apollo Global Management, the adoption of AI by large enterprises is decelerating. This analysis is based on the survey data of 1.2 million enterprises by the US Census Bureau. The results show that the adoption rate of AI in enterprises that employ more than 250 people is on a downward trend.

This trend may indicate that after an initial period of enthusiastic attempts, large enterprises are entering a "period of technological disillusionment", beginning to reevaluate the practical value and integration challenges of AI tools.

Is the return on 95% of AI investments zero?

The fundamental reason why enterprises encounter obstacles in adopting AI might be found in the MIT report that triggered the market crash earlier. A report titled "The Generative AI Gap: The State of Business AI in 2025" found that as many as 95% of enterprises have a return of zero on their generative AI investments.

The lead author of the report, Aditya Challapally, pointed out that the core of the problem is not the AI model itself, but rather the "learning gap" and deficiencies in integration strategies within enterprises. For instance, many general-purpose tools designed for individual users, such as ChatGPT, perform poorly in enterprise environments where specific workflows need to be adapted. Furthermore, the report found that the success rate of "purchasing" mature AI tools (about 67%) is much higher than that of enterprises "building their own" systems (about one-third), which poses a direct challenge to those companies that have invested heavily in developing proprietary AI systems.

Market panic may spread

These negative data have had a substantial impact on market sentiment. After the MIT report was released in August, it triggered a significant sell-off in technology stocks. The Nasdaq index recorded its biggest single-day drop of the month (-1.4%), and Nvidia's share price fell by 3.5%. Wall Street Journal wrote that a trader said at the time, "This story is causing people to panic."

For investors, a series of signals, from Goldman Sachs '" slowdown in growth "to Apollo's" decline in adoption rate ", and then to MIT's revelation of the "zero return" predicament, clearly indicate that the commercialization path of AI is far more complex and lengthy than expected. Against the backdrop that the expected price-earnings ratio of the Nasdaq 100 index is still nearly one-third higher than the long-term average, investors need to shift their focus from the frenzied pursuit of technological breakthroughs to a cautious assessment of the actual implementation capabilities and profitability of enterprises.

Risk Warning and Disclaimer

The market involves risks. Please invest with caution. This article does not constitute personal investment advice and has not taken into account the individual user's specific investment objectives, financial situation or needs. Users should consider whether any opinions, views or conclusions in this article are suitable for their specific circumstances. Any investment made based on this is at your own risk.