We have been traveling a lot over the past two weeks visiting investors for our day job. In one of our last meetings, about  five minutes into the conversation we realized they had no interest in semiconductors, our supposed topic of discussion. This was fine with us.  We have been having the same conversation for a while now, and did not mind when they started chatting about other topics. So somehow we ended up on the topic of Amazon, which we thought merited a post.

First, some full disclosure. We own 20 shares of Amazon in our personal account. It has been our best performing stock pick in the year that we have owned it. During that conversation we realized we have not been terribly diligent in monitoring our portfolio. We know that Amazon is not that profitable and trades at some wild, Internet-like multiple. Every quarter we check to see how the stock is doing relative to consensus. But in our conversation today we realized we had a more concrete investment thesis.

Several years ago, we noticed that every start-up we met with in the Valley was built on Amazon Web Services (AWS). At the time, AWS got little attention from investors. Amazon says very little about its size or revenue. But in a moment of Peter Lynch inspiration we thought there must be something to this AWS stuff. With so many people we know using it to do interesting things, we assumed Amazon must be on to something.

We do not follow the e-commerce side of Amazon. We use them occasionally to buy books and will probably do a lot of our Holiday shopping there, but that’s about all we know there. And in regards to AWS, our original investment thesis has gotten a bit frayed. Many of the companies who we knew used AWS have since told us that it was too expensive and they have gone on to build their own infrastructure. Still, AWS seems to be growing strongly and doing well.

However, there are two things about AWS that keep us in the stock.

First is profitability. If you look at the math behind AWS there is something interesting to find, something which says interesting things about Amazon itself.  Let’s assume every AWS server costs $20,000, which is probably wildly more than they really pay Synnex to build it for them. Then let’s assume that there is an equal amount of ‘overhead’ including the networking equipment, facilities costs, engineering time and bandwidth cost. That number is just a swag but is no more outlandish than the server cost. So $40,000 to run that server for the two years of its serviceable life.

Assume that through the magic of virtualization AWS can run 5 simultaneous users on that server. The figure could be much higher or much lower, but this is a reasonable average. And let’s assume the server runs non-stop for two years, but is down for 20% of that time. This accounts for slack time vacancies or what have you, but is just a degree of conservatism for the model. Add all that up, and those $40,000 in costs are far outweighed by the $140,000 in revenue they generate.

Real networking and IT professionals can pick these numbers apart, but our guess is that there are more ways to make the revenues go up than there are ways to make them go down. Our back of the envelope math seems like a reasonable swag.

Put that all together and it should be clear that AWS is almost ridiculously profitable. This does not appear in Amazon’s number because: 1) they have never disclosed AWS financials; and 2) AWS is growing so rapidly that deprecation costs probably swamp the revenue figures. Regardless, when Amazon decides that it needs to show profits, there are ample ways for AWS to deliver those. Which makes us wonder if the rest of Amazon’s business are run in a similar fashion. we do not know this, but our guess is that assumption is probably the bull case for Amazon, that they are just a mouse click away from profitability but would rather keep investing in the business. We am not recommending the stock, and will freely admit our ignorance about how real investors think of Amazon. We just like the AWS math and will hold onto our shares for a while longer.

However, there is a second interesting takeaway here. A couple weeks back Amazon had their re:Invent user conference during which they rolled out their desktop virtualization service. Their pitch is that IT departments no longer have to spend $100,000 per user to run virtual desktops. They can just pay Amazon $35 a month and save tens of thousands of dollars. You can see their math here.

The stock market thinks this is bad for Citrix who does a healthy desktop virtualization business. The investor we spoke with felt that this grossly understates the complexity of IT departments in big corporations. He maintained that the more expensive Citrix solution delivered considerable value in bridging that complexity, something which it would be impossible for AWS to truly emulate.  This is where we started to disagree. we think AWS has a value proposition that may be too good to pass up. While IT departments may like that complexity of the Citrix solution, CFOs and corporations at large may care less about it. In a nutshell, this is the argument for public cloud providers, and why they can beat private clouds. Sure, many companies and government agencies will have to build their own data centers (or get Amazon to build it for them), but many more will be happy to shed their massive IT infrastructure costs. This may not start with desktop virtualization, but it also seems likely that Amazon will not stop there.