Storage, compute and networking mostly. I'm confused by your question - are you suggesting that early stage startups couldn't possible generate a workload to warrant that cost? Maybe if your idea of a startup business is limited to running a basic CRUD app. But other data-intensive projects have engineering needs that can easily reach that scale, especially in the early days when you're trying to bootstrap your data catalog or ML models or whatever.
Those cases are rare as hell, $150k should be enough for nearly anyone. You are making bad engineering decisions if you go through that in under a year.
Ha, well tell that to the hard drive manufacturers and cloud providers.
The storage bill alone exceeded that at four startups I've worked with. When your job is to manage X PB of data with Y TB arriving daily, you're fairly constrained on the cost floor of your operations.
> You are making bad engineering decisions if you go through that in under a year.
Fine. You lack the imagination to conceive of a use case, or you lack the knowledge and skills to get the job doing it. Either way, this comment is rude and ignorant of the diversity of software challenges that exists in the real world.
I've had "it needs the RAM" while staring at a 15% memory usage plot on a very expensive CI VM. Choosing a 20x cheaper machine had no effect on build time and stability, but this was the PR comment. This overprovisioned machine was possibly $20k.
The truth is that developers sometimes require expensive things to get their job done efficiently. The other truth is that we often vastly overestimate the SKU. And we also leave things running.
As someone who worked at an early stage video intelligence startup, surprisingly easy. Redshift + Elastic Transcode + Cloudfront makes it extremely easy to spend thousands and thousands of $ per month.
