The tendency of talent allocation Shifting from Real to Fictitious” is an important issue faced by the current Chinese economy (Li, et al., 2017; Huang, et al., 2017; Liu, et al., 2018). According to the one-percent national sample census in the year of 2005 and 2015, the average years of education for manufacturing employees were 9.37 years and 10.26 years respectively, with those with junior middle school education accounting for 55.83% and 52.42% respectively. At the same period, the average years of education for those employed in the financial industry are 13.45 years and 14.27 years respectively, and those with higher education account for 55.11% and 69.63% respectively. The existing literature discusses talent allocation mainly from the perspective of government-enterprise (Murphy, et al., 1991; Li and Yin, 2014, 2017; Li and Nan, 2019), but there has been a new change in the allocation of talents in China. In addition to public administration departments (government), financial industry, real estate industry and other virtual economy that generate money with money” are also preferred occupation of the human capital groups, which is exactly the problem that the academic world cannot ignore. Given this background, this paper tries to answer the following questions: Has talent allocation been excessively biased towards the financial industry? If so, how will talent allocation affect Total Factor Productivity (TFP)? Before conducting the empirical regression, we construct a sample theoretical model to demonstrate the nonlinear effect of talent allocation between finance and manufacturing industry on TFP and its mechanism. Then, in the empirical part, we use the one-percent national sample census data and China Industrial Enterprise Database to establish the prefecture-level talent allocation indicator and empirically test the impact of talent allocation on the TFP of enterprises. We find that talent allocation and TFP show an inverted U-shaped relationship on the premise of controlling city and enterprise variables. We calculate the inflection point of talent allocation between financial industry and manufacturing is 1.10. For 283 prefecture-level cities in China, there are 273 cities whose talent allocation has been excessive biased towards the financial industry. And this phenomenon has significantly reduced the TFP of manufacturing. Considering the problem of omitted variables, reverse causality and measurement error of explanatory variables, and using the real estate-manufacturing talent allocation for regression and placebo testing, we find that the main conclusions remain established. If talents are efficiently allocated to manufacturing, Chinese manufacturing TFP can continue to grow by 2.7%, which is economically significant. In addition, heterogeneity analysis shows that there is a significant difference in the impact of over-allocation of talents on the high-tech manufacturing and non-high-tech manufacturing, and the optimal talent allocation of the high-tech manufacturing shifts to the left. This paper may have the following contributions: First, it provides empirical analysis to study the relationship between talent allocation and TFP from a new perspective. Some literature has recognized that talent allocation may be biased towards virtual economy sectors, but all of their analysis is descriptive. Second, it finds that China’s limited human resources are over-biased to the virtual economy sector which is represented by the financial industry. The policy implication is that, to promote real economy, policy-makers should attach great importance to this issue and adjust the manufacturing talent development plan properly.