Ability to experiment without upfront cost

In order to best explain this use case, consider the scientists trying to cure deadly diseases, for which they conduct research on the human genome. 

A genome, simply defined, is the blueprint of making a full organism. In humans, these genes are encoded as DNA in the 23 pairs of chromosomes. 

A genome of a single individual, will take several gigabytes of storage with several million DNA. With the massive compute power that we have today, it takes up to 200 hours (8.33 days) in order to sequence a single human genome (without parallelization and excluding analysis).

Consider that a doctor needs to run a sample size of 5,000 people for lung cancer research. Without parallelization, it would take about 114 years. However, the same can now be done at a fraction of the cost in less than 7 days time using the public cloud. That's less than 30 hours per human.

The cost of genome sequencing has also dropped, from being over $100 million at the turn of the century to a little over $1,400 today. 

So, experimentation and testing with different permutations and combinations has now become a reality, without spending a fortune upfront, all thanks to hyperscale public clouds.