Opportunity cost is everything. Staffing for a small startup, after funding, is a huge bottleneck. Every second you spend managing an install of Rabbitmq, be it on a VM or on-prem, is time spent not working on the app, or anything else that's drastically more important.
And every dollar you bleed on expensive AWS (and expensive devops engineers to wrangle it), is a dollar less you can pay for a new employee to deliver value quickly and get you profitable (or aquired).
Startups spend way more on staffing than cloud. generally the cloud Premium is less than the cost of a single employee for an early stage startup.
As the startup scales it’s about velocity of product. Spending 10% your time to save 20% on cost is a bad strategy. Instead startups should take that 10% time and invest it into their core product. This would lead to an accelerated timeframe for raising their next funding, which will be much larger than any cost savings. Eventually in the future once growth slows down you can focus on costs to improve profit.
*you still want to ensure your costs grow slower than top line rev.
Personally I think run on-prem till you outgrow a server in the office closet, Digital Ocean/Linode till you run out of features, then AWS/GCP when you start needing to scale to handle hockey-stick user growth. Then again, you'll need an expensive devops engineer to manage those redeployments seamlessly, so have you really gained anything there?
How much funding do you have? A couple hundred thousand is different from a couple million.
