The definition of the CubeSat is already nearly two decades old. This results in the trend that a large fraction of CubeSat missions are leaving the domain of (educational) technology demonstration and entering that of commercial applications. Therefore a larger focus is put on the reliability and compatibility of commercial of the shelf systems and the spacecraft containing them. One subsystem that is more or less fixed since the first CubeSat missions is the serial data bus. This internal network is essential for interconnecting subsystems. However, recent investigations have shown that the industry standard I2C must be re-evaluated due to performance and reliability restrictions. The research contained in this thesis sets off to evaluate these characteristics of other data bus standards to propose a possible future-proof data bus architecture. By splitting up the analysis cases into two parts, the choice and design of both parts can be optimised. The Telemetry and Command (TC) bus carries essential commands and housekeeping data between systems, while the Payload (PL) bus is dedicated to high speed bulk data transfers. An initial requirement-based selection of bus standards reduced the selection of standards to several options for both bus cases; for the TC bus, I2C, CAN and RS485 were selected. For the PL bus, CAN, RS485, SPI and USB were selected. The selected standards were all implemented in a data bus testing suite, comprising of up to nine simulated subsystems providing pseudo data to be communicated over the bus. Measurements were conducted of the buses' power consumption and data rates in several realistic test cases. Furthermore, the complexity and ability to withstand noise and voltage transients were evaluated. Ultimately, this resulted in a recommendation of using RS485 in future TC bus configurations and SPI in the PL bus configurations. However, this conclusion must be regarded a recommended direction of future research for several reasons. Firstly, more investigations are needed in the ability of these buses to work when subjected to large amounts of noise and other extreme environments. Secondly, the performed trade-off does not apply weighting to its criteria, as these can and will vary wildly for different missions. Finally, the test setup was limited in terms of processor ability for the PL bus case. These specific tests should therefore be redeveloped with more powerful equipment, allowing an even more realistic simulation of future CubeSat subsystems.