A cognitive industrial internet of things (CIIoT) resource allocation strategy was proposed to address the delay guarantee challenge of CIIoT services in dynamic and uncertain scenarios. A delay model based on CIIoT delay-sensitive services was constructed, and a rate-analytic solution for guaranteeing the delay demand of the services was derived. A CIIoT resource optimization model with the objective of maximizing network throughput was established based on the delay model, which considered the transmit power constraints of the base stations, the interference constraints between devices, and service transmission delay guarantee constraints under dynamic and uncertain environment. It is difficult to solve the model due to the dynamic uncertainty of the model. The robust optimization theory was adopted to transform uncertain parameter constraints into deterministic constraint problems, and a resource allocation strategy for CIIoT in dynamic uncertainty scenarios was proposed. The simulation results show that the proposed algorithm effectively guarantees the delay requirements of services and improves network throughput in dynamic and uncertain environments.