Aren't these numbers also so high because they took VC funding? The VC wants to throw the company against the wall as fast and hard as possible to see if it sticks. That means no time to roll out custom infrastructure, and hiring rockstar consultants just to build the product as fast as possible.
The idea is not terrible, it's definitely better than doing everything the right way (tm) for 5 years, just to find out there is no way to fit the market.
>That means no time to roll out custom infrastructure
It is sad that the industry is so anti-intellectual that we're actually afraid of computers now. Giving dell a CC# and giving them your hosting providers shipping address is not a complex or difficult task. There is no need for "custom infrastructure" to run a trivial web app. You buy servers, you plug them in.
Depending on your service provider (hosting, msp, mttos, whatever), the up-front msa can be truly painful if you want "real" service. I.e., redundant power and multi-provider network connectivity with multiple connectivity points that have a massive capital cost but will be as reliable as your commercial or home utilities.
Couple the cost of establishing the relationship (especially with startup credit) with the NRC and MRC of hosting services, smart hands rates, reserved cabinet or cage space for those surprise boxes on the dock, and incremental pricing models, and it's very difficult to build out and manage poorly predicted growth in a financially constrained environment.
Saving money on this means you're allocating additional engineer hours to dealing with this yourself.
The 'middle road' is probably an operator in a decent NAP or superNAP that manages the big lease and then carves out a margin by offering smaller space, possibly even colocating your hardware in the same rack as someone else's (and maybe not in the same rack with your other gear), and charging more than a macminicolo but offering significantly improved and variable services.
There are a number of walls you can hit that will present themselves as an almost insurmountable obstacle when you own your own gear. There are mandatory professional services when setting up some gear (SAN vendors have been notorious for this) and it's just a bunch of attention you have to devote to something for which there is generally a good enough solution for a company that may not exist long enough to depreciate the servers that ate a huge amount of their startup capital.
It's not that we're 'afraid' of computers. It's like saying we were afraid of electricity or telephone lines 40 years ago. We can manage it, but it's almost irresponsible and risky to do so.
You are inventing a problem that doesn't exist. I can literally have a cabinet in a first class data center in 2 days and it takes 15 minutes of my time. You save assloads of money this way vs AWS which is incredibly expensive. You can buy an equally performing server for the cost of a month of renting the VM from amazon (for the big ones obviously, you can't buy servers as slow as a ec2 large instance.)
Alright, who in your company knows what hardware to buy at which datacenter for how much? What software is going to run on it? who picks that software? How do you get data onto it? How do you get data off of it? What happens when it goes down? How do you monitor it? How do you scale it to two servers? Or more? Who is going to do that? Who is going to know all that? Who else is going to know that so your company doesn't fail when the first person isn't around?
>Alright, who in your company knows what hardware to buy at which datacenter for how much?
Everyone? I don't understand how you can imagine a problem like this. "Who in your company knows how to tie their shoes?"
>What software is going to run on it? who picks that software? How do you get data onto it? How do you get data off of it? What happens when it goes down? How do you monitor it? How do you scale it to two servers? Or more? Who is going to do that? Who is going to know all that? Who else is going to know that so your company doesn't fail when the first person isn't around?
AWS solves precisely zero of those problems. It isn't magic pixie dust, it is renting Xen VMs.
In addition to being fast to deploy, cloud infrastructure lets you easily scale up and down to try things out. This is useful if you're still finding traction or distribution channels that work best.
>> Giving dell a CC# and giving them your hosting providers shipping address is not a complex or difficult task
Financially, cloud servers have properties that might make them attractive relative to this. Also -- buying and managing colo is a fairly specialized discipline. There are a ton of business reasons to not take on a risk in this part of your business when you can outsource it at nominal cost. A COGS of 14% is quite nominal.
>In addition to being fast to deploy, cloud infrastructure lets you easily scale up and down to try things out
And you can buy a single server for $5k that will outperform $5k/month of EC2 VMs.
>Also -- buying and managing colo is a fairly specialized discipline
What? That is like saying "buying and managing phones is a fairly specialized discipline". No, it isn't. You give money, you get service. Not complicated. People have been able to handle this just fine for longer than you've been alive.
I wonder how much of the resentment to owning hardware comes from the "fail fast" approach - if you're trying to vet a business model quickly or move onto the next thing, having actual physical assets to take care of if the company ends up failing could be an off-putting extra complexity for some people vs just clicking "delete instance" with some virtualized provider.
I think it has more to do with initial fixed cost.
Personally I used my cofounder's existing array, but that's not an option for most companies, particularly starting out, and cutting out server purchasing costs pays in the short term. However, if you have the initial funding, as in this instance, it pays to buy the servers you need.
Plus, it's an asset that, should your company fail in 6 months, can be sold to at least salvage some of the initial investment.
[edit]In the last boom, though, people bought millions of dollars worth of servers and equipment that are now less than worthless; in many cases you have to pay to get rid of them. So, I guess there can be a significant downside.
The idea is not terrible, it's definitely better than doing everything the right way (tm) for 5 years, just to find out there is no way to fit the market.