As an important part of the modern economic system, the urban industry has its unique spatial layout and location selection. Based on the micro-level enterprise data, using location entropy, hotspot analysis and conditional Logit model, the paper investigated the spatial distribution characteristics and location selection of urban industrial enterprises in Lanzhou. The results show that: (1) The overall distribution of urban industrial enterprises present a spatial agglomeration feature of "one center area and two sub-central areas". There is not much difference in the number and density of enterprises in the central city, suburbs and outer suburbs. However, as the spatial scale shrinks, the differences are more obvious on the district and county scale, and more significant on the block scale. (2) From the perspective of various industries, food processing and interior decoration manufacturing industries have hot spots in each circle of the city. The garment manufacturing, processed tourism supplies industries and electronic information manufacturing industries are mainly concentrated in the central city. The packaging printing industries and cosmetics washing product manufacturing industries gather in the suburban and outer suburb. (3) Urban industrial enterprises are mainly affected by local economy, city economy, infrastructure, policy planning and suburbanization, while technological innovation have less impact on their location selection. (4) In terms of industries, the location choices of the labour-intensive businesses represented by garment manufacturing, food processing, packaging printing industries, interior decoration manufacturing industries, and process tourism supplies industries are mainly affected by spatial distance, historical location, infrastructure, etc. However, capital and technology intensive enterprises represent by the cosmetics and washing products manufacturing industry and electronic information manufacturing industry preferred locations with technological innovation and policy planning factors.