Depends on the industry. At seed in computational biotech, we had a lean team. After validating our technology as much as we could, we played a lottery on AWS to get compounds. These workloads handled tens if not hundreds of billions of rows of data and part of the selling point was the parallelizibility, so we got badass instances to show we could turnaround results in minutes whereas other companies took months.

We got a 100k and 150k AWS grant from AWS, and that was a big chunk of the computational budget. We were small, but the nature of HPC and our confidence in our algorithm meant we burnt the war chest on a YOLO and raised Series A.