Jan Hatzius, chief economist at Goldman Sachs, recently stated that despite the significant amount of money American companies have invested in artificial intelligence (AI) over the past year, the actual contribution of AI to U.S. GDP growth in 2025 is essentially zero.
Goldman Sachs' research analysis suggests that this "input-output" mismatch is mainly constrained by macroeconomic accounting logic. Since a large number of high-performance chips and hardware equipment required by U.S. companies depend on imports, these substantial capital expenditures are offset by import items in GDP calculations, leading to almost no direct impact on gross domestic product in the data.
At the same time, feedback at the micro level is also not optimistic. A recent survey of nearly 6,000 business executives in the U.S., Europe, and Australia found that although 70% of companies have integrated AI into their operations, as many as 80% of respondents admitted that AI has had no substantial impact on their company's employment structure or productivity levels so far.
This finding has raised deep concerns in the market regarding the "AI bubble" and "investment conversion rate." Analysts point out that while AI technology has transformative potential, converting hardware investment into actual productivity spillovers still requires overcoming long-term challenges such as technological adaptation and business restructuring.
