Which effect occurs when applying One-Hot Encoding to a categorical variable with high cardinality?

One-Hot Encoding creates a new binary column for every unique category. With high cardinality, this leads to a massive increase in feature dimensionality, resulting in a sparse matrix. This consumes significant memory and often leads to overfitting, as the model may learn noise from categories with very few observations. Dimensionality reduction or target encoding is usually preferred for high-cardinality features.