Apollo's chief economist has found that companies with more than 250 employees are delaying their adoption of AI, based on a survey of 1.2 million companies conducted by the U.S. Census Bureau. A report from MIT three weeks ago shows that up to 95% of companies are seeing zero returns from investments in generated AI.
A recent study caused an alarm to sound at the same time. Enterprise-level artificial intelligence (AI) applications face serious challenges.
Three weeks ago, a survey released by the Massachusetts Institute of Technology (MIT) said 95% of companies receive zero returns from investments in generated AI. Last Sunday, Apollo Global Management's chief economist Torsten Sløk noted that AI adoption rates in large American companies were showing a downward trend.
Both studies reveal significant obstacles to the conversion of AI technology from hype to practical applications. Sløk cited official US survey data showing that companies with more than 250 employees are slowing down AI adoption. This may indicate a reassessment of the actual value of AI technology by a company.
The MIT report further analyzes the reasons behind this phenomenon, stating that the problem lies not in the AI model itself but in the flaws of the corporate integration strategy. The findings of these two studies have led to a strong market response, leading to the NASDAQ index experiencing its biggest day drop since August 1, leading to AI-related stocks such as NVIDIA facing significant sales.
AI adoption rates of large companies decline
The Torsten Sløk analysis is based on a large business survey conducted by the US Census Bureau. This happens every other week, covering 1.2 million businesses and asks if they will use AI tools such as machine learning, natural language processing, virtual agents, and speech recognition to help produce products or provide services over the past two weeks.
The chart above shows the moving averages from six surveys conducted by the US Census Bureau. Research data shows that AI adoption rates are declining for large companies with over 250 employees. This trend suggests that despite market enthusiasm for AI, large companies may be experiencing a “technological disillusionment stage” and are beginning to reassess the actual value and return on investment of AI tools.
The decline in adoption rates may reflect the integration challenges companies face after initial attempts and the difficulties in transforming AI tools into tangible business value. For investors, this data shows that the commercialization path for AI technology can be more complicated than previously expected.
MIT research reveals the dilemma of AI investment
Released on August 18 by MIT's Nanda Project, the Generated AI Gap: Business AI Status in 2025 Report provides a deeper analysis. The study was based on interviews with 150 business leaders, a survey of 350 employees, and an analysis of 300 publicly available AI deployment cases, and found that only about 5% of AI pilot projects achieved rapid revenue growth.
Aditya Challapally, the report's leading author, noted that the core issues lie in the “learning gap” within the organization and flaws in integration strategies. Many business leaders mistakenly attribute it to the regulatory environment or model performance, while overlooking the challenges of internal adaptation and integration.
For example, general purpose tools designed for individual users, such as CHATGPT, are popular for their flexibility, but performance is degraded in a corporate environment as they cannot effectively learn from a specific workflow or adapt to the specific business needs. This “fits one size” application approach has resulted in a considerable number of AI projects that cannot have a measurable financial impact on businesses.
Key differences in successful AI implementations
MIT research also delves into the significant differences between successful and failed AI deployment cases. Some successful companies, especially some startups, have adopted a strategy of “focusing on a single problem, implementing accurately and establishing intelligent partnerships.” Challapally noted that certain startups led by young entrepreneurs achieved “revenue growth from zero to $20 million within a year” through this approach.
The survey found that over half of the generative AI budget was allocated to sales and marketing tools. However, the biggest return on investment actually comes from back-office automation scenarios, such as outsourcing business processes and reducing costs for external agents. This suggests that companies may have misdirected the direction of their AI investment.
Another important finding is that “purchases” outweigh “buildings.” Buying AI tools from professional suppliers and establishing partnerships is around 67%, while the success rate for internal building systems is only one-tenth. This data poses a direct challenge for companies that have invested heavily in their attempts to establish their own AI systems.
Market response and the impact of investment
The findings from MIT had a major impact on the market last month. The day after the report was released, US tech stocks fell sharply on August 20th, with the NASDAQ Composite Index falling 1.4%. Nvidia, the core beneficiary of the AI boom, saw a 3.5% decline, while Palantir and ARM saw a 9.4% and 5% decline, respectively.
The report said traders close to the multi-billion-dollar US technology fund “this story is causing panic among people.”
This shift in sentiment resonates with recent warnings from Openai CEO Sam Altman about “investments are overly enthusiastic” and the potential formation of the AI bubble, resonating with even worse market skepticism regarding the prospects for commercialization of AI technology.
The release of the MIT report coincides with growing market concerns about overvaluing high-tech stocks, with the NASDAQ 100 index forecast price-to-return ratio at 27x, nearly a third higher than its long-term average.
For investors, these two studies provide important risk signals and demonstrate the need for a careful evaluation of AI-related companies and practical implementation capabilities, rather than focusing on technical breakthroughs and market enthusiasm. The AI revolution may still be ongoing, but the path to commercialization will be more complicated and longer than expected.
Edited by Stephen
