Most AI-focused startups shouldn’t focus on training, fine-tuning, or otherwise making significant hardware investments (e.g., GPUs) before finding product market fit. (GPUs for inference is, of course, OK). In many cases, this is the wrong sequence for startups