Hybrid cloud is a widely adopted framework where on-premise storage and/or compute resources are combined with public cloud system. This paper explores the storage aspect of this framework, which requires designing coding schemes that are aware of both local and global components of the available storage space. The coding schemes should provide efficient repair mechanisms for the data stored on the public cloud (global storage space) and utilize the local storage space to facilitate seamless access to the overall information stored on the hybrid cloud storage. This paper presents a mathematical model for hybrid cloud storage which takes all these requirements into account. The paper then extends the information flow graph approach to characterize the fundamental limits on access bandwidth of the system, i.e., the amount of data downloaded from the public cloud during the data reconstruction process. This paper also presents several explicit coding schemes that utilize the available local storage space to attain the fundamental limit on the access bandwidth. The setup where multiple clients with varying local storage spaces are supported by a single global storage space is also addressed.