Last week we saw Amazon break out their AWS numbers as a separate segment for the first time, so of course, the entire tech press lost their collective minds and wrote reams about it. It’s bigger than we thought! It’s smaller than we thought! They have margin! OMG.
I’ve previously written about the economics of cloud, both here and elsewhere, and after listening to some of the commentary on the In Tech We Trust podcast (well worth a listen, by the way) I’ve decided to write some of this up to hopefully shed some light on the situation.
Let’s take a deep breath and walk a little more calmly through the numbers. You can do this for yourself thanks to the excellent SEC EDGAR service, and grab the documentation from here: https://www.sec.gov/cgi-bin/browse-edgar?action=getcompany&CIK=0001018724&owner=exclude&count=40&hidefilings=0
I’m going to give you my analysis on this, and I encourage you to read a variety of opinions on this. But looks carefully at the chain of reasoning behind what someone says, and also look carefully at their track record. Really good analysts admit it when they get things wrong, while snake-oil selling pundits and futurists don’t. You want to learn the why, not just the what, because then you are better armed to make your own decisions.
The Headline Numbers
In the three months to 31 March 2015, the AWS segment had sales of $1,566 million, and operating profit of $265 million, which gives AWS a margin of about 16.9%. Compare this to the North America business (which is all the Amazon.com stuff, the retail stuff they’re known for) with sales of $13,406 million and operating profit of $517 million, which is margin of 3.9%. This ignores the seasonality of retail, which has a big bump in sales during the Primary Gifting Period in the last two to three months of the calendar year, but to get last quarter’s results I’d need to pull the first 9 months out of the 10-K report for 2014, since Amazon haven’t broken out the quarterly results for their last FY quarter.
Lots was written about this stuff, because it was easiest to find and talk about. 17% margin is impressive for something most people assumed is an “undifferentiated heavy lifting” business. But this is a single data point, and it doesn’t provide anywhere near as much information as people assume it does.
For example, for all of financial year ending 31 December 2014, Amazon had operating profit of $178 million. So AWS, in the last three months, apparently made more money than the entirety of Amazon for all of last year. And Amazon made $700m a mere three months, and in only three months, after making only $178m for all of 2014.
Does that make a lot of sense to you? Did the costs Amazon incurred really change that much in a mere three months?
Read the Notes
I remember reading something in business school about novices and MBA students doing better than a lot of analysts with financial statement analysis because they actually read the notes, while professionals just cut-and-paste numbers into financial models and ignore most of an annual report.
That’s one reason I like to read the notes.
And that’s where I found the first thing that surprised me. In note 8 of the 10-Q, there is this sentence: “There are no internal revenue transactions between our reportable segments.” which I initially took to mean that each segment is operated more-or-less as a separate business, and there’s no cross-charging, which is not what I expected to see. Cross-charging is common within large organisations so that costs associated with a shared service (like email or networks, say) are allocated to other business units that make use of the service.
Ah, but if we read on, on page 27 when we’re looking at the more detailed breakdown of the operating expenses, we see this: “Technology infrastructure costs consist of servers, networking equipment, and data center related depreciation, rent, utilities, and payroll expenses. These costs are allocated to segments based on usage.”
I take this to mean that Amazon does indeed do internal cross-charging. They don’t do arms-length trading between divisions, meaning that Amazon.com isn’t charged to use AWS like any commercial customer. Instead, Amazon will charge costs to the segments internally, based on ‘usage’, whatever that means. Depending on how Amazon decides the costs should be allocated, the segment figures could move around quite a bit, particularly when we’re talking about yearly revenues of $90bn. Moving a few hundred million around isn’t a large percentage of total revenues, and for a segment that’s less than 7% of total revenue, but apparently 38% of operating profit.
Here’s another curious thing to note: in note 8 we see this: “The majority of technology infrastructure costs are allocated to the AWS segment based on usage.” which could be interpreted in two ways. Most of the technology infrastructure costs are due to AWS, or that most of the cost allocation is based on usage, but not all. I choose to interpret this as “most of the technology infrastructure costs” are because of AWS, because retail should mostly be in cost of goods sold and distribution overhead. There’ll be a small amount of technology overhead, but not a lot, because retail runs on thin margins. AWS would need to invest a lot in technology infrastructure because that’s the nature of the business.
So let’s assume that most, if not all, of the costs for Technology and Content are due to AWS. This is supported by the note on page 26: “Costs to operate our AWS segment are primarily classified as “Technology and content” as we leverage a shared infrastructure that supports both our internal technology requirements and external sales to AWS customers.”
But here’s a curious thing: Technology and Content is 12.9% of Net Sales (i.e. revenue). But AWS as a segment is only 6.9% of total revenue, and 5.9% of operating costs. What’s all the gear for? Does Amazon really need 50% of all of its Technology and Content costs just to run its core retail business? That’s another 50% of the cost of fulfilment, and 1.2x the entire Amazon marketing budget.
Another curious thing to note is the growth in spending on Technology and Content. According to the 10-K from last financial year, Amazon has increased spending on Technology and content by just over 40% for the past 2 years. Assuming that investment continued (and why wouldn’t it?) this financial year’s spend would be about $13.2bn, an average of $3,306m a quarter. Now, the first quarter number will be lower than the average for the year, but last year it was only $200m under the average. Why is the first quarter spend only $2,754, a full $600m less than what we’d expect the average for FY2015 to be?
If we go and check the trend in growth of Technology and Content, we see that for all of last year, spending increased at between 8 and 10%, but this first quarter of 2015 it grew by only 4.5%, about half. Why?
Either growth for Amazon (either AWS or overall) slowed suddenly, or Amazon had invested a lot already and didn’t need to invest much money in this first quarter. Which of these options does Amazon have more control over, and which provides the better marketing message? Or any one of dozens of other, perfectly legal, options that Amazon has for spinning the story to their own advantage.
There’s More To It
There’s a lot more to this than the simple headlines from the tech press might lead you to think, and I honestly have a lot more questions than answers from the two-ish hours I’ve spent looking at these financial statements.But I don’t believe AWS has 17% margin any more than I believe it’s the cheapest option for infrastructure. Consider that in Brad Stone’s book about Bezos and Amazon The Everything Store Bezos is quoted saying Steve Jobs made a ‘mistake’ with iPhone pricing:
Bezos believed that high margins justified rivals’ investments in research and development and attracted more competition, while low margins attracted customers and were more defensible.
Maybe you should take anything Bezos, or Amazon, says with a grain of salt because why would he tell you the unvarnished truth? He’s a businessman, not a saint.As an aside, I tried to copy&paste that quote from Stone’s book from Kindle reader in Chrome, but it won’t let me.