A startup is developing a generative AI application. They decide to build and manage their own physical servers and GPU clusters in their office instead of using a cloud provider's AI infrastructure. What is a primary business implication of this on-premise infrastructure choice compared to using the cloud?