Startup Reality Distortion #5: Web Hosting Is Not Free

About This Blog

This site is for  entrepreneurs.  A full RSS feed to the articles is available.  Please subscribe so we know you're out there.  If you need more convincing, learn more about the site.



And, you can find me on Google+

Connect on Twitter

Get Articles By Email

Your email:


Blog Navigator

Navigate By : 
[Article Index]

Questions about startups?

If you have questions about startups, you can find me and a bunch of other startup fanatics on the free Q&A website:

Subscribe to Updates


30,000+ subscribers can't all be wrong.  Subscribe to the RSS feed.

Follow me on LinkedIn


Current Articles | RSS Feed RSS Feed

Startup Reality Distortion #5: Web Hosting Is Not Free


Many of the startups I’m involved with today (including my own) are providing software as a service (SaaS) in a web hosted model.
There has been lots written about how costs for infrastructure have gone down dramatically.  Servers, storage, bandwidth and systems software are all cheaper today than ever before.  This is actually true.  Things are cheaper than when were during the previous Internet “boom” years.  
One of the great things about web software companies is that it is so cheap to get started.  If you’re a programmer, you can likely write most (if not all) of the code yourself host the end solution (cheaply) somewhere and start getting users/customers pretty quickly.  This is a wonderful thing and I’m a strong advocate of doing exactly this (i.e. having programmers start software companies and get to market as quickly as possible).
However, a common “reality distorting” phenomenon is believing that web hosting is free.  Before I get into my “talking points” on the topic, I’d like to make one thing very clear:   I’m not suggesting to “pre-optimize” the infrastructure and spend lots of money building a “scalable” environment.  As an entrepreneur, you should be more worried about making no sales (or minimal sales) than worrying about the ability of your hosting environment to deal with a big spike in volume.  Don’t lose sleep over how you’re going to deal with user #100,000 – lose sleep over getting user #10 on the system in the first place.
In any case, here are some of my thoughts on the whole web hosting for software startups topic:
The Myth Of Almost Free Web Hosting
  1. Cheaper, But Still Not Free:  Though most infrastructure components (hardware, storage, bandwidth, etc.) are indeed getting cheaper, they are still not free – and likely never will be.  These are physical things.  They cost money to whoever is providing them and though it’s a “scale” business, this scale is usually already baked into the low costs we see today.  The infrastructure/hosting industry built out a lot of capacity in the 1990s and there was lots of consolidation too.  This drove prices way down.  That means there are really big players with lots of scale already.  The other cost contributor is that of the components – which are also getting cheaper.  But, they are still not “free”.

  1. Customers Want More Storage:  At the same time that prices have been dropping, customer expectations for storage are rising.  It’s no longer enough to give 500M of disk storage for common web applications now.  Many companies are providing 1G, 2G or even more to their users.  Though I don’t have “hard data” on whether customer expectations are rising at the same rate as costs are declining, I’m going to make the easily defensible argument that the trend in terms of what you have to give customers is certainly going up.  

  1. Apps Need More Bandwidth:  In addition to storage, the same holds true for bandwidth as well.  Back in the good old “Web 1.0” days, it was enough to have a nice, efficient web application with a focus primarily on capabilities.  Now, it’s all about the “experience” which often includes much more richness (which generally equates to bandwidth).  More images, more audio/video, and more Javascript/AJAX/Flash/whatever mean more bandwidth consumption.  If you think bandwidth is free, think again (and check with your hosting provider).

  1. Horizontal Scaling Adds Complexity and Cost:  One thing we’ve learned is that it’s better to scale “horizontally” (i.e. have a higher number of small machines instead of one big, bad-ass machine).  This allows us to scale “linearly” simply by adding more pizza-box servers as traffic and usage grows.  This is a great thing.  However, from a startup’s perspective, what people don’t realize is that the cost of adding server #2 is not the same as that of server #1.  More on this in my example below.

  1. The Disaster Lottery Ticket:  Most web startup companies that are saving money on hosting costs are basically living on borrowed time.  If you want to make startup founders lose sleep, ask them one simple question:

If all of a sudden, your server died taking all data that was on it with it, how long before you could start serving customers again?  (hint:  Many bootstrap startups are not worried all that much about reliable backups).  For added fear, add the fact that your hosting provider does not respond to your email or phone calls within the first 24 hours.
Personal Example:  I’m working on a web software startup now (beta launch coming up in June) and am dealing with all of the above issues.  When I first started the business (April, 2005), I did what all good entrepreneurs should do.  I bootstrapped and spent as little money as possible.  I found a nice, cheap but reliable hosting service to host my web application and started bringing on users.  Since then, here are the primary “discontinuities in cost” that I’ve experienced as the company evolved.  I don’t think I’m atypical here.  Though your timing will vary, you’ll go through more or less the same things.
Discontinuity 1:  Moving from shared to dedicated hosting.  If you’re doing anything remotely serious on the web, you’re going to need a dedicated server at some point.  A shared hosting environment presents too much risk and too little control.  It’s going to happen to you someday.  You’ll need to move to dedicated hosting.
Discontinuity 2:  Discount hosting to “managed” hosting.  As you start bringing on customers, chances are, you’re going to need more support than just “automated rebooting”.  Though you may not move to a fully managed service (where there are carbon-based life forms in addition to an IP address at the other end of the hosting relationship), you’ll probably outgrow the discount hosting service.  This generally happens the first time things go very badly and you need help getting your server and/or application working again.  Trust me, that day will come.
Discontinuity 3:  More than one application server.  You would think that if one server costs you $200/month (let’s say), than two should cost you $400/month, right?  Not so fast.  Once you move to a server farm environment, things change.  Now you need to worry about hardware/software load balancing.  You have to worry about shared storage across servers (maybe even a NAS device).  You have to worry about state management across servers.  At this point, you’ve likely also outgrown your “discount hosting” service as many of them are not positioned to meet these types of “advanced” needs.  Pop quiz:  How much will your existing hosting provider charge you to increase your current disk space available by 10X (i.e. from 80G to 800G?).  
Discontinuity 4:  Disaster Planning.  Once you have real customers and need to provide a “service level agreement” (implicit or explicit), you have to start planning for disasters.  Components need to be redundant so there is no single point of failure.  You need a way to “recover” should that very, very bad day come.  And, if your startup is successful enough to survive a couple of years, you’ll see that day.
So, across the various discontinuities, I’ve now gone from spending about $25/month on hosting back when the company started, to spending over $2,000/month.  Realize also that this is my own money (I haven’t raised outside capital), so I’m not looking for ways to spend VC money. 
Summary Of My Point:  Web hosting is not free.  Much of your costs also depends on your proclivity for ulcers and how much money your customers are paying you (and what they can expect for that).  If you’ve got 100,000 “free” users, maybe you can get away with $100-$200/month.  But, even then, not for long.  So, the big message here is this:  Are you being realistic when planning out your business model?  Can you really afford to grow to a million “free” users in the hopes of making money some day?  Maybe you can.  But then again, maybe you can’t.
Note:  I’ve tried to stay away from quoting “hard” numbers here.  So, please don’t spam me with why Hosting Company X can do it so much cheaper than what I am paying now.  Or that I’m foolish for spending that much money on web hosting.  I may be foolish, but I am not incented to overspend on this stuff  (being self-funded, my interests are aligned with myself).

Posted by on Fri, May 26, 2006


Hi Dharmesh, as ever an insightful article, thanks.
We have recently rushed headfirst into this area and i've been unpleasantly reminded just how complex 'commodity' web platforms can be.
My late '90's experience at a dotbomb project involved 600k worth of Sun and orders of over 1million to Oracle, Cisco, Checkpoint etc. Our entire setup was run by three very clever sys admin and networking guys who often performed 24-48 hour marathon magic stints at holding the whole ball of string together.
Today, all of these vendors are unnecessary evils to early stage startups but those sysadmin and networking skills are still required at short notice (and sometimes for those extended marathon periods). Fortunately, these skills are plentiful, just make sure you factor this cost into all resource spreadsheets. Also, create and cultivate relationships with these resources early, inevitably you need them at short notice when things crash as you sail close to the wind trying to satisfy customers 10 thru 10,000.

posted on Friday, May 26, 2006 at 1:22 PM by Dan Walsh

Hi Dharmesh

Great article. I work for a reputable web hosting provider and always recommend our clients what you mention here. We never hear back from them until something bad happens. They just don't take their business that seriously and respect their clients. And when problems start, we get blamed even though we told them to get a load balanced solution, or get a backup drive. Very true.

posted on Friday, May 26, 2006 at 5:31 PM by N.D.

I work for a managed hoster, too, and I've had the same experience as N.D. had. Nobody ever says that we don't know what we're talking about, but yet they think that the extremely cheap service is "almost" as good as the managed, load balanced, backed up, etc service. Sometimes we've even had people come to us because their old provider didn't fill there needs, but then leave after a while to another cheap place, then come back again when that doesn't work out, then leave again for another cheap place, then ask to come back again, etc. (Really! That many times!) Sometimes they'll even try and come back after breaking their contract (ie by trying to get a few free months by non-payment until we shut them off). I can't wait for the wonderful (mythical) future when more people read this article and start to understand.

posted on Saturday, May 27, 2006 at 12:38 AM by SLB

Insightful article!

P.S.: What's up with these question marks next to the labels in the "Post comment" section? When I click on them, I just get a popup that says "undefined".

posted on Saturday, May 27, 2006 at 1:12 PM by Hermann Klinke

I've been thinking about starting my own company for quite some time, but I always assumed that I'd be doing my own hosting, and rolling my own load-balanced cluster.

Why do this? Yes the learning curve is steep, and it's an additional job the founder must take on, but who is going to be more concerned about the stability of your web app servers, an ISP or you?

Google could not have been created if they had outsourced their hosting.

posted on Sunday, May 28, 2006 at 5:12 AM by J Nelson

Very good article. Thank you. We've been in the business for over six years and seen the bubble, the burst and the recovery.

It's been an amazing change. We used to basically sell space. Now space is irrelevant but dedicated and managed is the norm. Bandwidth? Bandwidth is not just quantity. Most hosting companies that provide "All you can eat" mean "all you can drink through a straw. If you don't have the ability to increase width and depth and duration, you will have problems. It's not a simple thing to guess.

Doing it yourself is fine, absolutely. But nobody will pay you for your service if it goes down. At least using a hosting provider co-lo setup gives you some safety. We've spent a lot of money on backup generators, power conditioners, battery backup units, etc etc etc. So have all the other companies. Now that 1U servers are the norm, there's more rackspace avaialbale for co-lo. Manage the boxes yourself, go nuts. But I don't care how nice your living room is, our server room is designed for it. A startup gets one chance to take over the world, if you blow it from being cheap but still think in-office massages are ok, then you are on the wrong track.

Awesome article.


posted on Wednesday, May 31, 2006 at 2:55 AM by Robert Gagnon

Hey. I'm thinking of starting a web app soon. I was wondering if you think it is worth going straight to dedicated or waiting on shared until you realise that it isn't enough? I'm a student so budget is everything!

Nicer overall tips though. Disaster planning is an important one!

posted on Thursday, July 20, 2006 at 3:29 AM by James

Cannot agree when this article any more. In facts whats why I started - to educate the general mass about how to choose a webhost and what to expect from this industry.

posted on Sunday, March 04, 2007 at 11:38 AM by Hostvise

Hey dharmesh , great post. I'm getting ready to launch the beta of my online travel startup in a few weeks, and have been having a rough time making decisions on hosting. I'm pretty settled on a basic package from RackSpace which is fully managed and comes in about $300 per month. It was reassuring reading your thought process on this.

posted on Friday, March 28, 2008 at 5:48 PM by Adam

Comments have been closed for this article.