Amazon: revenue mix shift; fulfillment losses; sum of parts valuation (w segment commentary)
First, take a moment to enjoy the most beautiful chart in business history:
What’s even wilder than how gorgeous this chart is, is the fact that the mix-shift of revenue is changing in a highly positive way. Every single year (nay, quarter) a growing portion of Amazon’s revenue is derived from its more profitable segments. Lots more coming on that below.
Amazon now has seven distinct reporting segments. In order of revenue:
Online Stores (OS)
Third Party Seller Services (3PL+)
Subscription Services (Subs)
Physical Stores (PS)
Currently, [OS / Subs / PS / Other] are all losing money (orange tints). Collectively they accounted for 53.9% of Revenue last quarter (Q2 2022):
The unprofitable revenue segments accounted for 63.83% of revenue in 2019. Crucially, the decline is being driven primarily by AWS. AWS is their segment with the largest total addressable market (TAM) by far, meaning the decline should remain in place for years to come. Here is the forecasted revenue mix shift through 2024 based on consensus estimates (with some minor adjustments by me):
Note that AWS share of revenue is expected to increase by 41% (18.48% / 13.1% = 1.41) in just a few years.
Amazon’s stock was punished more than the other tech giants (though as of July 31st it has had a sharp rebound) because its operating profits took a dive and free cash flow went highly negative. The good news is that more than half of the negative FCF is attributable to overbuilding the fulfillment network. These investments will end up having a positive ROI even if the R takes longer to realize.
The bad news is the rest of the negative FCF and most of the negative operating profits are due to surging fulfillment costs (mostly fuel) which will never be recuperated. Just to give you an idea of how much money Amazon is currently losing on fulfillment let’s do some napkin math.
This analysis comes from Morgan Stanley and was done earlier this year (diesel is now at $5.30 and had been even higher). If we multiply fulfilled units (21,095M) for 2022 by total cost per fulfilled unit as of the time of this analysis ($4.18) we get: $88,177 (88B). This is a pretty wild figure already. I did the calculation adjusting for the current diesel price and the figure is even more terrifying: $4.66 per fulfilled unit. $4.66 X 21,095 = $98,302.7 (98B).
Higher fulfillment costs wouldn’t be such a big problem except that Prime members don’t pay for shipping so there’s nothing that can be passed through. So - what portion of those shipping costs is Amazon currently covering? I.e. what portion of those costs are attributable to Prime customers? We can make an educated guess using what has been reported about total gross merchandise volume, # of Prime subscribers, and average spending per Prime subscriber.
GMV = $600B
Prime subscribers spend an average of $1400 per year
There are 200M Prime subscribers
200M X $1400 = $280B
$280B / $600B = 46.67%
There are other variables, for example product mix could be different between Prime and non-Prime purchasers, not all purchases by Prime members have free shipping, and so on. But we’re probably over-estimating in the right ballpark.
Now we can take total annualized fulfillment costs based on the current price of diesel: $98,302.7 and multiply them by 46.67% to get: $45,874.6B that Amazon would be spending if diesel cost $5.3 on average throughout the year. Diesel prices for the year are below, they’ve come down from a peak of $5.81 to $5.268 today:
Now the stock tanking makes sense doesn’t it? Amazon was looking into a very deep hole with no way to climb out of it - at least not quickly.
Amazon’s entire Subscriptions revenue in 2021 (which is comprised mostly of Prime member fees) was only $31.8B. Annualizing Q2 subscription revenue brings us to: $34.9B. Even after accounting for the monstruous price hike in Europe (which is far less than it seems after you account for the Euro plummeting against dollar) and the 17% or so bump in the US they will still be losing billions on fulfillment alone if diesel doesn’t come down fast. Worse still, remember that the Subs business has other costs. Amazon spent $13B on content last year, this will show up mostly on their balance sheet but still some of the spend gets amortized each year. Beyond content, there’s likely $2-3B in other operating costs like SG&A to run the segment. This explains how Amazon manages to hemorrhage money even with AWS, Ads, and 3PL+ all printing.
Amazon Sum of the Parts Valuation
Jeffries thought a reasonable valuation for Amazon by 2021 would be around $153 based on the post-split price. You can see the multiples and growth assumptions they used above. Covid-19 accelerated both e-commerce and the move to the cloud. Jeffries probably would have been damn near spot on without the pandemic.
The operating margins Jeffries was assuming for each segment at the time were:
With Jeffries prior analysis as a backdrop the first thing I’m going to do is recreate their own SOTP using today’s figures. Results below:
For reference the current price is $134, and I put in consensus growth rates for 2022-2024.
Now I’m going to re-do a SOTP but make adjustments based on my own view of the business, and taking into consideration what we’ve learned over the past 3.5 years. I’m going to go segment by segment starting with AWS.
Amazon Web Services. I believe AWS is the strongest component of what I call the “base-layer” of the cloud. You can read my previous post to get the details about how large I think the opportunity is and what makes the base-layer unique from an investment perspective. AWS grew at 33% YoY in its most recent quarter, has a massive back-log growing faster than its revenue, and is essentially the top dog in an oligopoly servicing the largest TAM in history. It’s growing more than twice as fast as Microsoft, yet is similar to Microsoft in being an essential “technology utility”. The one place it lags Microsoft is in operating margins, which are around 30% vs. Microsoft’s 40%. All-in I think the higher growth rate and being top dog of the cloud probably more than washes out the lower profitability, meaning AWS should trade at a premium to Microsoft. That said, we’ll do a low and high scenario.
AWS Low: Microsoft’s next twelve month operating margin multiple = 21.9
AWS High: 30% premium to Microsoft’s next twelve month multiple = 28.47
AWS Low: 21.9 X $31,284 = $685.12B
AWS High: 28.47.3 X $31,284 = $890.66B
Online Stores. Online Stores is a money losing business and will probably never do much better than break even. Amazon doesn’t think about this business as a place to make money. Rather, they see this segment as fulfilling two main purposes:
It supports most of Amazon’s moats (customer relationship, logistics, economies of scale, network effects, etc) - which in turn support profitable segments
It is an R&D laboratory to develop what Jeff Bezos calls 1,000 run scores (more below)
Amazon’s online business was the ONLY business for the first 6 years of Amazon’s life. It’s insane rate of growth (and the associated pains Amazon experienced managing through that growth) led both to Amazon’s 3P Seller Services business and to AWS. The highly profitable Ads segment that now taxes seemingly every transaction on Amazon is also enabled (in part) by the unprofitable retail and subscription businesses.
Amazon wasn’t the first company to have a hard time building out a fulfillment network and massive customer service operation. Amazon wasn’t the first company to struggle in scaling on-premise architecture. But - they were the first company to realize what a massive TAM these pain points represented and to actually build a business to service them. Amazon has a track record of turning pain into opportunity (this explains why they bought One Medical - which they see as addressing one of their current biggest pains: labor costs - of which turnover and healthcare are the two largest components after salary).
There are two Jeff Bezos quotes that illustrate how Amazon thinks about “placing bets” (investing for the future).
As a company grows, everything needs to scale, including the size of your failed experiments. If the size of your failures isn’t growing, you’re not going to be inventing at a size that can actually move the needle. Amazon will be experimenting at the right scale for a company of our size if we occasionally have multibillion-dollar failures.
Put differently - Amazon believes that unless they are occasionally losing “multiple billions on single failures” - then they aren’t experimenting enough.
We all know that if you swing for the fences, you’re going to strike out a lot, but you’re also going to hit some home runs. The difference between baseball and business, however, is that baseball has a truncated outcome distribution. When you swing, no matter how well you connect with the ball, the most runs you can get is four. In business, every once in a while, when you step up to the plate, you can score 1,000 runs. This long-tailed distribution of returns is why it’s important to be bold.
AWS, Ads, and the 3PL+ business all qualify as 1,000 runs. Alexa/Echo probably will. Their Just Walk Out technology and Buy with Prime are less certain (still early) but are also very promising. Ten years from now they’ll probably have some 1,000 run scores that haven’t even been swung at yet.
So what is Online Retail worth? Given its historical role in creating 1,000 run scores I’d be tempted to put some sort of premium on this segment vs. other retailers like Walmart who have created no such opportunities. However, as of 2015 third party gross merchandise volume (GMV) surpassed first party (what goes under Online Retail). At this point, any benefit (contribution to Moat, experimental lab, etc) Amazon is getting from Online Retail is also being delivered by their 3PL+ business. Hence, I think it’s fair to view this part of Amazon’s business like a faster growing online version of Walmart. I’m going to put the same revenue multiple as Walmart for my low case: .63X and a 1X (Jeffries’ low case) multiple as my high case.
Online Retail Low: $142B
Online Retail High: $225B
Third Party Seller Services. This business is a combination of four different business models:
eBay style commissions (very profitable, probably 25% operating margins
UPS style fulfillment (8-9% operating margins usually, lower currently)
Pick-pack-ship/Warehousing (3PL) - don’t know what these margins would be but I’m guessing similar to UPS
Outsourced customer service - going to assume same margins as #2/3 above
Jeffries put the operating margins of the entire segment at 5-6% as of 2018. My guess is that’s pretty close due to Covid-19, inflation & overbuilding. But, I think they will increase fairly rapidly. The split of revenue for this segment between commissions and the rest isn’t broken out but we can make an educated guess. Amazon takes about 14% of the average sale, so we can multiply this by their estimated third-party GMV of $390B to get $54.6B in commission revenue. Our estimated 3PL+ revenue for 2022 is $115.4B so that leaves $60.8B in revenue to the UPS / 3PL side of things. Hence, long term (assuming revenue mix stays the same) we can assume that this segment probably has something like:
$54.6B X .25 = 13.65B operating profit; PLUS
$60.8 X .08 = 4.9B operating profit
$4.9 + $13.65 = $18.55 / $115.4 = 16% operating margins.
Again, they’re definitely not there today, but this is probably in the ballpark of where they’re headed.
My view on the 3PL+ business is that it’s heading toward being as impenetrable as a railroad. Amazon spent so much building out their fulfillment network (and now has extra capacity to spare) that I don’t see anyone successfully competing here - and they will likely keep taking share from UPS and Fedex. For reference, they added as much square feet to their warehouse/fulfillment network between 2019 and 2021 as 1/3 of ALL of Walmart’s shoppable square feet in the United States. That’s insane.
What’s more, they’ve recently launched Buy With Prime which will enhance their logistics/3PL+ moat yet further.
Note that Amazon’s parcel volume likely overtook UPS in the second quarter of this year, and if it didn’t it certainly will in the third quarter.
Railroads are regulatorily protected oligopolies. As such even though they grow top line around 0-GDP they still maintain nice multiples (and historically have made great investments). Amazon’s 3PL+ business is not regulatorily protected but it’s moat is probably as strong as they come after actual regulations. It’s TAM is still growing at more than 2X GDP and it will likely take market share to boot. Consensus growth rates are mid-high teens through 2024 after a lull this year.
I’m going to value Amazon’s 3PL+ on the low side at a 20% premium to UPS’ next twelve month operating multiple, and on the high side at a 20% premium to Union Pacific’s net twelve month operating multiple - in both cases using Amazon’s normalized operating income.
Normalized operating income: 16% X $134.5 (NTM consensus rev) = $21.52B
UPS multiple: 13 X 1.2 X $21.52 = $335B
UNP multiple: 16 X 1.2 = 19.2 X $21.52 = $413
Advertising. Amazon is rapidly gaining online Ad share. The tricky part with this segment is figuring out what the TAM is. Most of the advertising today is what you see every time you search for a product:
But, Ads are also in other places like Amazon devices and Prime Video - which has been expanding the Ad TAM with the recent acquisition of MGM and Amazon’s purchase of exclusive rights to Thursday Night Football.
As saturated as Ads are on Amazon, there is still an opportunity for them to become more effective and hence to command higher fees. Further, Amazon ads are served to customers who have just literally told them exactly what they’re looking for (e.g. “Tennis Shoes”). This makes them the most effective direct response ad platform on the internet, which means they’re going to be far more resilient during a down turn than a company like Meta (Facebook) and probably more resilient than Google. Also, Amazon isn’t impacted by any privacy changes from Apple. It’s operating within its own ecosystem and isn’t reliant on outside data, so the pain of Meta/Snap/etc is (at least in part) Amazon’s gain.
One final point - the faster Amazon can ship you something the larger its advertising TAM. Ads being served to a customer who is about to buy something that they need to receive in the next few hours have a different dynamic than ads being served to a customer who can wait a few days. Neat right? Amazon will be able to serve more ads, and charge more for them, because it can ship faster than anyone else.
As with the 3PL+ business I’m going to assume that their long-term operating margin is going to be equivalent to Google’s, rather than guestimate what their margin currently is. For the low range I’m going to put a value of 10% less than Google’s operating multiple on next twelve months and for the high range I’m going to use a premium of 10%.
Operating margin = 30% X $45,069B = $13.52B
Low: 16.8 X .9 X $13.52 = $204B
High: 16.8 X 1.1 X $13.52 = $250B
Subscriptions. As we saw above this segment (right now) is essentially a cash incineration machine. It’s primary function - like with Online Retail - is to serve as part of the moat that supports other profitable segments. I think best case this segment commands a Netflix level revenue multiple (2.94X NTM), and worst case this segment is worth less than 1X, though I’ll use 1X for my low case.
Low Case: $42.43B
High Case: 2.94X $42.43 = $124.7B
Physical Stores. I’m going to leave this valuation the same as Jeffries - these are a rounding error so it won’t make much of a difference anyway.
Other. This segment includes PillPack, the newly acquired One Medical, Kuiper project and many other ventures. I’m going to toss in a $8B here on the low side (PillPack, One Medical and Rivian investment = $6.2B at cost alone), and a $13B on the high side - arbitrarily adding $5B.
Here’s the final result, and I’m reprinting Jeffries’ for easy comparison:
Thanks for reading 01Core! Subscribe for free to receive new posts and support my work.
Now, let me tell you how I really feel.
I wanted to walk through some SOTP valuations because lots of analysts use them and they’re definitely helpful in getting to know the business. That said, a SOTP valuation method works best when the parts are capable of functioning independently and could theoretically be split off, AND when you have high quality comparable businesses (publicly traded) to match the segment multiples against.
For example, we can use UNP’s multiple to estimate the value of Berkshire Hathaway’s BNSF railroad - perfect.
But, none of Amazon’s businesses are really similar to anything else, and while AWS could be split off from everything else, the remaining segments are all reliant on each other. Here’s my preferred framework for thinking about Amazon.
Pretend you’re in 2024 and all segments are operating at normalized margins…
The retail side of the business (3PL+ / Ads / Online Retail / etc) gets a 10% premium multiple to the S&P500 for the low case and a 40% premium multiple to the S&P500 for the high case because:
It’s got a collection of diversified and deep moated revenue streams
It is growing faster than the average S&P500 constituent
Amazon - unlike other businesses - is likely to create more 1,000 score runs and this optionality deserves a premium
Using 18X for the S&P 500, this gives us a 19.8 X normalized earnings for Retail on low side and 23.4 X normalized earnings for Retail on high side. I’m going to assume (to be conservative) a 1% operating margin for Online Retail, 2% operating margin for Physical, 0% margins for Subscriptions and Other, and 85% conversion of operating margin to Net Income. We now have the following:
For AWS, we’ll use the same multiple as we did above. AWS will still be growing at 20%+ (maybe 25%+?). My guess is that as time goes on the market will continually re-rate AWS’ terminal value higher as it realizes how big the TAM is and how long high growth will be sustained. Further, it will permit AWS to command a similar multiple even with slowing growth because it will become increasingly obvious how sticky and utility like the business is.
AWS Low: 21.9 X Forward Operating Margin
AWS High: 28.5 X Forward Operating Margin
Now we have the following:
Note, the RoR is NOT annualized, it’s the total return achieved by share price moving from where it is today to the implied share price.
Thanks for this! Great write-up on Amazon. I found particularly interesting how you pointed out that the money-loosing businesses of Amazon actually play their role in prodding the profit-making businesses.