Generating embeddings is expensive, especially on the CPU, you either need more CPU cores, a GPU or a smaller model if you want to run models locally. Also not sure why you would want to use the intel cloud machines on Hetzner when they have much faster and more modern AMD servers available for basically the same price.