Every prompt, log and cached token your team sends to a model is data. Runars exists so Canadian teams can run inference and GPUs on Canadian soil, billed in dollars, without giving up the APIs they already use.