Having established what Large Language Models are and how their size is measured in parameters, we now consider the physical hardware required to run them. This chapter focuses on the essential computer components involved.

We will review the roles of the Central Processing Unit (CPU), system memory (RAM), the Graphics Processing Unit (GPU), and its dedicated video memory (VRAM). A brief introduction to Tensor Processing Units (TPUs) is also included. Grasping the purpose of each component is necessary to understand how LLM size impacts hardware selection for tasks like inference, which we will cover subsequently.