What advantages does quantization offer over dimensionality reduction for scaling embeddings?

What advantages does quantization offer over dimensionality reduction for scaling embeddings?