Concentration gradients has been presented in many biological applications with the performance of chemotaxis analysis, cell growth, DNA detection and toxicity evaluation. Traditional methods for concentration gradient generation are usually with the characteristic of inefficient, inexact and unstable. To overcome the boundedness, we demonstrated a self-similar fractal microfluidic concentration gradient chip with rapid response, high precision and stable concentration gradient, based on the classical Christmas tree model, in this paper. We tested its performance using both numerical simulation models and experimental research. First, a multi-physical field coupling model was established and simulated by a software named Comsol Multiphysics under different sample introduction conditions. Then, a variety of concentration gradient distributions are obtained by coupled setting the normalized injection flow matrix with the concentration matrix. Finally, the experimental results from deionized water and red dyes are in good agreement with the simulation results, which proves the rationality of the design of two-stage self-similar fractal microfluidic concentration gradient chip. The advantage of the chip is that it can generate flexible concentration gradients by simply adjusting the injection flow ratio without redesigning the configuration of the microfluidic device.