Despite global spending on artificial intelligence infrastructure reaching nearly $2 trillion, which is about 2% of the world’s GDP, only 5% of companies using AI report a noticeable increase in productivity. A new study suggests that this apparent inefficiency might actually be a driver of long-term progress. The study analyzed nearly 50 years of data on most American public companies. It found that approximately one in five firms actively invests, even if their current productivity is below average. For this analysis, data from companies trading on US exchanges was used. These companies stand out from others: they are younger, more innovative, and about four times more likely to show a significant jump in productivity-doubling sales and increasing productivity approximately by 50% over several years. Additionally, they file twice as many patents, which are cited three times more often. They are generally in the process of developing new products and technologies.
The study also shows that increased investments by these companies lead to overall productivity growth over the next five years. Researchers constructed a model demonstrating that some companies have a small but significant chance of a transformational performance leap. Investments in new technologies, research, or brand development increase the likelihood of such breakthroughs. The probability of a performance jump for a typical firm is 1.6% per year, but for actively investing firms with low profitability, it rises to 4%. The analysis showed that companies making big jumps in productivity often peak during major technological transitions, especially in the late 1990s and in recent years with digital technology developments. These waves of active and risky investments often precede significant productivity growth.
The study authors argue that the current gap-where some firms actively invest in AI while others wait-is not necessarily a sign of irrational optimism. It is part of the natural process of innovation. When new technology has the potential to transform production, broad and uneven investments are not a mistake but a feature. Redirecting capital from low-productivity companies could inadvertently stifle experiments leading to future breakthroughs. History shows that some degree of «irrationality» is not only inevitable but necessary for technological progress. Recent advances in autonomous systems, natural language processing, and healthcare innovations underscore how increased investments may lead to these transformative changes.
