To reiterate, seems like the business you're describing fits a pretty narrow profile: free bootstrapped site making money from ad revenue and growing by word of mouth and organic SEO (to minimize marketing spend). That is a phenomenal way to run a company, and if you can pull it off, you have my sincerest congratulations – however, you're likely not the target audience for the post above.
For software companies which don't follow your model (i.e. most of them) OR businesses which are aiming to trade growth for profit (a perfectly valid business strategy, see $AMZN), cost of hardware is almost _never_ the limiting factor compared to cost of labor. And to pay for labor at a loss, one may need to look for equity funding since banks won't extend debt to high-risk seed-stage businesses.
Your assertion as to the desirability of trading growth for profitability is entirely orthogonal, and the arithmetic about cost of server maintenance vs revenue - while informational - seems to miss the point entirely.
There will always be companies so great that VCs need to beg them to take money, but one only has to spend a day in Palo Alto coffeeshops to see that generally demand goes the other way.
> growing by word of mouth and organic SEO (to minimize marketing spend)
How else? I never saw an ad for Google, Facebook, Snap Chat, Drudge Report, Instagram, Twitter, or Plenty of Fish. More generally, I'm unsure just what significant bucks should be spent on marketing a startup such as I described. I can see trying to have some stories in some parts of the media and various other forms of publicity, but actual marketing, and as you also mentioned, sales, what's to do?
E.g., I was a B-school prof and across the hall from me was the guy who did, IIRC, the Pillsbury Dough Boy ads -- it was something that famous and novel he did. I've done high end applied math optimization for several cases of marketing. So, from that marketing, I see no connection with the startup I described.
For SEO, I doubt that Google did that! Nor Facebook, Snap Chat, Instagram, Twitter, Drudge Report, Plenty of Fish, etc.
Sure, if I type into the Google one line text box
KVR1333D3E9SK2/16G
I'll get a lot of hits of sellers and can believe that Google got SEO $ for that keyword. So, there may be hits from Amazon, Newegg, eBay, Tiger Direct, Memory4Less, and Kingston, and I can believe that Google got $ for those.
But sellers of
KVR1333D3E9SK2/16G
are not "good information technology" startups or candidates for Series A VC funding.
And for VC funding, there's another huge point: What the usual, even what is common on some Sand Hill Road coffee shop, etc. is essentially irrelevant. Instead, necessarily, especially if take the Paul Graham definition, the only things that are relevant are the really rare, literally once in a decade, exceptions. Here, too, the VCs are stuck-o because they have nearly none of the background to judge what will be so exceptional. The NSF, DARPA, and many departments at MIT, Princeton, and Stanford do at least on the technical parts. But IMHO, the technical parts should be so powerful, solve so well such an important problem, that they are about all that matters and isn't just trivial. The VCs don't agree yet. Tough for the VCs.
You raised the issue of labor: Well, I'd strongly recommend a founder to be a solo founder. This way he is forced to understand all of his business, and a good founder to be CEO needs to do that. Also he saves big bucks. He never has a "co-founder dispute" wastes essentially no time in meetings.
You raised the issue of hiring: Well, for a solo founder before any equity funding with $100+K a month in revenue, he can hire. Let's see: $120K a year is $10K a month plus some overhead, and can cover that a few times for the $100+K a month in revenue.
For hiring a rock star, full stack programmer, for me, no way. In my startup, I'm the rock star. And for the work of my startup, essentially no one in startup land, in a Silicon Valley coffee shop, computer science major, experienced developer, etc. has any hope of competing with me. Why? What's crucial for my startup, although the users will never know this, is some original applied math I derived, complete with theorems and proofs, from some advanced pure/applied math prerequisites. That's not programming, full stack programming, rock star programming, ugrad computer science, AI/ML, or chaired full prof of computer science work.
For the programming, sure, for my startup I did that, all of it. It was no big deal and doesn't make me a rock star or full stack programmer. Instead I just used the Windows .NET Framework, Visual Basic .NET (I don't like the C-like syntax of C#), ADO.NET for database, ASP.NET for the Web pages, TCP/IP sockets and object instance de/serialization for program to program communications, etc. Simple stuff. Visual Basic .NET? Fine with me. I'm thrilled with it. While I've programmed in a lot of languages, including C, I've never touched C++, C#, Python, Java, or JavaScript and won't unless I have a need to. I've never touched Linux or Unix and don't plan to. So, I'm no rock star, full stack programmer and see no need to be. Then I see no need to hire any rock star, full stack programmers.
I believe that in a few weeks I can teach any talented, interested experienced computer user how to use a good text editor, Visual Basic .NET, ASP.NET, ADO.NET, simple cases of SQL, TCP/IP sockets, and de/serialization quickly. For joins and foreign keys, so far I don't have any! Each index I have is clustered and, thus, as fast as possible. For the crucial, high end, original technical stuff, that's my job and easy for me, and I couldn't hire for that work anyway.
So, for my startup, hiring, no problem. There will be a delay of a few months getting new people up to speed and nicely productive. I'll hire anyone, fresh out of high school or retired with 40 years in the computer industry or 40 years in anything.
If I need some high end technical info, I'll call Microsoft and pay them. So far I haven't had to.
Currently my code is in alpha test. It's 24,000 programming language statements in 100,000 lines of code for a Web server, a Web session state server, and two back end, highly technical compute servers. They communicate with TCP/IP, etc. They are all single threaded, and the queuing is from just the TCP/IP FIFO queuing. IIRC, the default maximum queue length is 10, but likely I could increase that to 100 and have plenty of slack. In the end, in general, the way to keep queues and stacks from overflowing is just to slow the responses to the users; the software and server farm architecture should exploit this fact.
I will soon add a log server to replace my current usage of what Microsoft has for a Web site log server that I don't like. The many places in my code that write to the log need not be changed; instead I'll just change the one function they all call. Writing the log server will take me an easy afternoon and will be just from removing some cute usage of a collection class from my Web session state server and adding a print statement.
Basically, then, the production code is written -- I thought of it, designed it, and wrote it, all in one stroke, no prototypes, no refactoring. Everything worked just as planned.
For being "narrow", essentially every promising startup is necessarily "narrow".
For software companies which don't follow your model (i.e. most of them) OR businesses which are aiming to trade growth for profit (a perfectly valid business strategy, see $AMZN), cost of hardware is almost _never_ the limiting factor compared to cost of labor. And to pay for labor at a loss, one may need to look for equity funding since banks won't extend debt to high-risk seed-stage businesses.
Your assertion as to the desirability of trading growth for profitability is entirely orthogonal, and the arithmetic about cost of server maintenance vs revenue - while informational - seems to miss the point entirely.
There will always be companies so great that VCs need to beg them to take money, but one only has to spend a day in Palo Alto coffeeshops to see that generally demand goes the other way.