Also, paying for cloud compute is an operating expense vs the capital expense of buying hardware yourself.

I'm not going to pretend to know all of the tax implications, but it can actually make a significant difference to the company because the operating expense can be fully written off. The cost is therefore offset slightly vs depreciating the value of physical hardware over a number of years.

I did this math, I've got an MBA. Physical hardware wins in nearly all cases.

For startups you cant beat used hardware from ebay and some IT experience in one of the cofounders. This is especially true for development machines (CI runners, compile caching, GPU servers, etc).

Yeah, if you're a startup you'd better have someone tech oriented and you'll be able to do a lot of things for cheap.

If you're a tech startup, then I don't even know how you'll be able to do anything without someone technical :-)

> If you're a tech startup, then I don't even know how you'll be able to do anything without someone technical :-)

I've spoken to some YC founders. You can definitely raise funds without a (real) technical founder.

It's also extremely comparatively cheap to get powerful dedi servers from places like Hetzner. Price/performance absolutely dwarfs the cloud services.

(Serious question)

What do you do with that hardware? Do you go to a colocation and stand the servers up there? Run them from home? What are your options for something like this?

Edit: I realized afterwards you may not be talking about public-facing servers

I've done a few different things with these types of machines. I also know a few people who run some big companies off infra like this.

The key for a small team is finding a colo provider with really good remote hands. Essentially you can setup your rack/cage, leave good instructions and room for expansion. Then you buy a server from dell with financing, ship it to your DC, someone installs an OS and racks it (or you PXE boot from your rack ;) ), then you can provision anything you want. It's essentially like cloud providers but scaling is slower.

Colo the servers. I had 17 servers + the hardware firewall and a massive data store in a single cabinet for $600 a month. The colo I selected was a former Enron data center, so it had a massive Internet pipe.

How did you deal with data redundancy / backups?

If your use case is 1/10th the monthly cost of equivalent cloud resources you can have 5x the regions and still have budget for plane tickets and remote hands.

This is probably true, thanks

I had 3 duplicate environments - dev, staging and production. Backups were an automated process between data stores, and integrated into the environments normal operation. The entire setup was automated to the degree it only required casual glances at resource limits about once a week, with minor devOps maintenance occasionally. We're all smart engineers, producing wonderful tech - automating a server stack is intern work.

Buy new hardware and just do the IT yourself - it is seriously not hard. That's the other side of cloud propaganda: running an maintaining servers is easy peasy..