GPUs are better suited for workloads requiring massive parallelism (e.g., AI model inference), while CPUs handle single-threaded tasks (e.g., database management) more efficiently. GPUs, like NVIDIA's A100, feature thousands of smaller cores optimized for parallel computation, making them ideal for AI tasks involving matrix operations. CPUs, with fewer, more powerful cores, excel at sequential, latency-sensitive tasks. In a mixed workload, investing in GPUs for AI and retainingCPUs for general-purpose tasks optimizes performance and cost, per NVIDIA's "GPU Architecture Overview" and "AI Infrastructure for Enterprise." Options (B), (C), and (D) misrepresent GPU/CPU differences: architectures differ significantly, GPUs don't replace CPUs for general tasks, and GPUs have more cores than CPUs. NVIDIA's documentation supports this hybrid approach.