What the early Amazon investors saw
Peeling back the curtains behind one of the greatest investments of all time
Potential investors in early Amazon would have been easily dissuaded by their sky-high valuations
Yet some prescient investors managed to look past the noise and discovered one of the best performing investments in contemporary stock market history
What did they see? We’ll explore it here.
There are few investment success stories which capture the collective imagination as much as Amazon. Once billed as a nondescript online bookstore, it quickly stole the spotlight in the business community by performing the corporate equivalent of blitzkreig and penetrating new verticals with blinding speed - general e-commerce, Prime, cloud, tech hardware, logistics, groceries, pharmaceuticals, video… you name it, there’s an Amazon subsidiary for that. Now at the peak of society, Bezos’ mantras of ‘customer obsession’ and ‘your margin is my opportunity’ have become buzzwords in the business zeitgeist.
Just as Amazon’s meteoric share price performance made Bezos excessively wealthy, it has also made countless other early investors globules of money as well. Big names in the investment space (e.g. Nick Sleep and Bill Miller) have been hailed as investing legends for identifying the future corporate juggernaut in its nascent days - leapfrogging even the Oracle of Omaha himself. More lyrics have been waxed on the what-if’s and could-have-been’s of missing out on an early Amazon investment - than perhaps the phenomenon of first love itself.
Hence the question begs - what did these prescient early Amazon investors see in what was at the time a fledging Internet bookstore? What did they behold in the company’s public records in which they saw beauty, which others found plain? What can we learn from studying its financial history, so that we don’t repeat the mistakes of omission? And more importantly, what lessons lie in the past that can teach us how to wield the capitalistic sickle with more dexterity and grace - so as to reap the rewards from other future Amazons?
Enough Shakespeare. It’s time to turn wisdom into wealth.
The dilemma of early Amazon investors (2004 - 2010)
Before Amazon became the bona fide American conglomerate spanning multiple continents and countless industries, in 2005 it was… still basically just a startup. While the online e-commerce site had several things going for it - including the rare negative cash conversion cycle - it was still nothing compared to the legions of offline brands that dominated America’s retail landscape at the time. True, the company had graduated from being just a bookstore to at least something resembling eBay - but it was already picking fights with household names the likes of Barnes & Nobles and Toys’ R Us, and its natural trajectory meant that it would eventually fall under Walmart’s crosshairs - whereupon the wider investment community was expecting the latter to wipe Amazon from the face of the earth.
Making the zit on its face even larger was the fact that Amazon’s stock was trading at sky-high valuations of 30x - 90x PE between 2004 and 2010 (ignoring the earlier outlier periods), for an average PE ratio of 60x. While that wasn’t unnatural for tech stocks at the time, in 2005 the world had just recovered from being seared by the dotcom bubble which had burst in late-2000 and bottomed-out in early-2003, with faith in overvalued technology stocks that were going to change the world significantly waning.
While revenue was clearly growing like a weed in 2005, earnings was still relatively stagnant - leading to questions about the company’s sustainability. During this business era, investors had still not gotten used to the concept of funding growth for extended periods of time in winner-takes-all industries; and it was difficult to ask the mainstream investor to part with their hard-earned cash in the absence of profit visibility.
Some astute investors drew comparisons between early Amazon to early Starbucks, which had in the late 90’s operated on an accelerated depreciation accounting model - where company-owned stores were depreciated faster than the industry standard (likely in line with accelerated tax depreciation policies) and therefore depressed accounting earnings - but which clearly did not affect actual look-through earnings. By 2005, Starbucks had taken the world by storm - with some stores literally facing each other on different sides of the same street - and gossip at the water cooler wondered whether Amazon was about to do the same.
However, that wasn’t enough to sway the mainstream investors. While it was possible that Amazon’s innovative accounting policy might be hiding potential growth, the fact remained that there was an 8-billion pound gorilla in the room - Walmart - which could still crush the insignificant e-commerce website should it decide to pivot in its direction. On top of that, the e-book wars between Amazon and Barnes & Nobles were heating up, and it was also at the same time burning cash making inroads into Toys’ R Us’ and eBay’s turfs (the latter of which had 3x its market cap). So while the upside potential of early Amazon was visible, to many - including Buffett himself - the risk was simply not worth the reward.
What the pioneers saw
Hence given the challenges above, what was it that the early Amazon investors grasped behind their monocles that everyone else seemingly missed? With the benefit of over a decade of hindsight, I’ll try to piece the puzzle together here.
As we can see above, Amazon’s revenue had been growing like gangbusters since its inception, alongside America’s then nascent e-commerce industry’s growth. Gross Profit also kept pace, which was (and still is) used by many early investors as a KPI to justify their high earnings multiples. The idea being that when a company is operating in the growth stage of its industry’s life cycle, gaining market share is by far the most rational use of investment dollars. Hence, it should be busy reinvesting to gain as much market share as possible - which leads to depressed earnings, but shows up as performance in Gross Profit growth.
As for what growth investments they were making exactly, that’s well known too. We can see above that Fulfillment expenses consistently made up nearly 10% of revenue, or about 40% of GP margin. In stark contrast, Marketing expenses (or what some refer to as S&A expenses) was only 2% of revenue - which was remarkably low for a company which was supposed to be justifying its low earnings by gaining market share through advertising & promotions (the bread and butter of retail businesses).
Clues to the reason behind this departure from the status quo can be found in Bezos’ biography, The Everything Store; together with his business mantra of customer obsession. Around this time, Amazon’s upper management team got together to determine whether they wanted to spend their scarce investment dollars on advertising or maintaining lower prices. Ultimately, Bezos decided that Amazon would keep offering lower prices, which would also serve as a natural form of advertising.
So that explains the relatively low Marketing expenses. But what about the relatively high Fulfillment expenses?
In order to answer this question, we first need to understand the composition of Amazon’s Fulfillment expenses. The line item Fulfillment largely represents the immensely complex logistics side of Amazon’s e-commerce business, involving:
the middle-mile and last-mile delivery operations,
the warehouses where they store and process the millions of pallets of inventory that flows through their locations, and
HQ operations which determine the strategic aspects of e-commerce logistics - e.g. route planning, inventory optimization and delivery fleet utilization.
Arguably, the most important part of the e-commerce business.
Giving some context to Fulfillment
Early Amazon’s Fulfillment expenses can basically be broken down into 3 parts:
Fulfillment by Amazon (FBA)
First, Warehouses. Warehouses are the bread and butter of any e-commerce business, as they are where the processing and storage of the literal mountains of inventory happens.
I won’t go into the details behind the operations of this, but imagine trying to sort through millions of little individual boxes that can’t be addressed in bulk each day to ensure that they get to the right doorstep as quickly as possible. On top of that, you need to ensure that all the little boxes delivered to your regional fulfillment centers are delivered on time to the respective local warehouses, with basically all the complexities of running a full-fledged Fedex or UPS back-end.
Now, if we take a look at the composition of early Amazon’s Fulfillment expenses, we can see that it’s roughly keeping pace with Revenue and Gross Profit growth. We can also see that Rental expenses under operating lease agreements (‘r’) makes up roughly 10% of total Fulfillment expenses (r / f), which seems par for the course for your average business. However, this conceals the immense amount of Operating leases (‘OL’) that Amazon was picking up, to the tune of 60% of PPE+OL [i.e. OL/(PPE+OL)]!
To appreciate why this is significant, we first need to understand the business model of supermarket logistics.
The supermarket - specifically Walmart - was the industry that the fledging e-commerce industry was closest to resembling; and hence closest to disrupting. The reason for this was simple - in a commoditized industry such as theirs, customers only care about getting the lowest price, which means that competitive strategy revolves around obtaining the lowest cost. And in an industry which involves ensuring the availability of tens of thousands of SKUs at all times to stores spread out across the entire country, business strategy simply boils down to acquiring maximum economies of scale.
Think about what is required to run the nearly century-old supermarket business model. You need to ensure the timely delivery of thousands of SKUs daily, so that your customer doesn’t decide to visit a competitor for that particular item that you’re missing - and risk being lost to them forever. That means you need to mobilize an army of delivery trucks and raindrops worth of storerooms all across the country to ensure full availability of every item at all times.
The problem with this is that you can’t always coordinate demand and supply perfectly. For instance, let’s say that a drop off in average inventory demand in California coincides with a spike of average inventory demand in Massachusetts, for whatever reason. Obviously your delivery vehicles and storage capacity on the East Coast will be significantly overburdened, while those on the West Coast will be underutilized. This involves an inefficient usage of the fixed costs involved in acquiring those vehicles and storage space, which represents wasted depreciation vis a vis if those demand and supply imbalances could be fully matched.
The solution that Walmart came up with to address this problem was the hub-and-spoke (‘HAS’) logistics model. Rather than a distribution channel that responds to demand and supply on a point-to-point basis, a HAS logistics model fills an entire 20 TEU trailer with goods that goes to one particular regional distribution facility (i.e. hub), which then breaks down the regional delivery into smaller trucks that service their respective local warehouses (i.e. spokes) around the hub. This is called the middle-mile in logistics, and you can see an illustration of how it works here: https://youtu.be/wVGdkBXeJ7s&t=64s
The beauty of the HAS logistics model is that it solves the intermittent mismatch between demand and supply of deliveries. If a truck on a particular spoke doesn’t have enough deliveries on a particular day while the one on a nearby spoke has too many, they could combine it into one truck serving both routes on the same day. It also allows planners to plan ahead for other logistical disruptions, such as peak traffic times or higher delivery volume during office hours. This dramatically saves on fuel costs compared to a point-to-point model, where trucks just randomly drive to wherever they need to be without any process. It also significantly optimizes efficiency, by allowing goods to be loaded and unloaded in a centralized area and optimal stock levels to be maintained.
In other words, it enables economies of scale - which as mentioned above, is what the business models of all commoditized industries ultimately revolve around.
Okay, now let’s take a look back at the same table as the previous one. A few things stand out. Firstly, even way back in 2005, Amazon was making huge operating lease commitments stretching out multiple years. Accounting standards define an operating lease as basically a rental contract that extends beyond 12 months, without an option to purchase the leased asset at the end of the lease period (in contrast to a finance lease, which does have such an option). This means that Amazon was basically committing to long-term rental contracts for the leased asset, which we can only presume to be warehouses given the size of the leases involved. In the table below, we can also see that they regularly leased warehouses up to 5 years out:
Considering that early Amazon was regularly signing operating leases making up >50% of the annual FCF (OL / FCF) - and presumably re-signing expiring contracts - this begins to look a lot more like an outright purchase of warehouses rather than leasing them - despite there being no purchase option at the end of the lease period. In other words, a conservative analyst should assume that these cash flow obligations represented CAPEX, rather than OPEX. Furthermore, we can see that they started this practice way before 2005, which means that Bezos likely had this vision of maximizing Amazon’s economies of scale from Day 1.
So what can we derive from all this? It seems likely that Bezos had begun outfitting Amazon’s distribution channel with a HAS logistics model even way before it was necessary, when they were still small enough to rely on 3PL operators like Fedex or UPS for delivery.
We can test this theory out by observing the ratios of rent-to-depreciation (r / depn) and OL / (OL + PPE). Operating Leases regularly made up >50% of early Amazon’s effective fixed costs - both on the income statement and the balance sheet - which infers that the Operating Leases were by design actually CAPEX in nature. This means that Amazon was actually building way more capacity than it needed for the present in anticipation of future growth - which lends credence to the theory that Bezos was actually constructing a nationwide HAS logistics channel.
Prime and FBA
Now, of course, it doesn’t stop at the warehouse. As mentioned above, Fulfillment expenses can be broadly categorized into 3 components: warehouses, Prime, and FBA. While there can be no doubt that establishing a nationwide coverage of warehouses was the central thesis behind early Amazon’s logistics strategy, it also enabled another one of the key success factors behind Amazon’s e-commerce triumph: Prime.
According to The Everything Store, the idea for Prime came from a rank-and-file employee named Charlie Ward who thought about giving away deliveries like a sort of club membership benefit, a la Sam’s Club. While the idea obviously worked in hindsight, it’s important to recognize that Prime could only work with an in-house distribution channel already in place. It would have been nigh impossible for Amazon to offer unlimited free 2-day shipping across the country for only $79/year in 2007 (when Prime launched) if Amazon hadn’t already spent the last half-decade outfitting its distribution channel with a HAS logistics model; and still had to depend on 3PL’s like Fedex or UPS for delivery. Hence, it’s highly accurate to say that the investments represented by Fulfillment expenses contributed to Amazon’s success with Prime today.
The same can be said for Fulfillment by Amazon (FBA). FBA was the 3rd-party outsourcing of inventory and delivery processing by merchants to Amazon - where they would store their inventory in an Amazon warehouse and have it automatically processed and delivered by Amazon warehouse employees whenever a customer made an order. It dramatically simplified things for merchants, and made it significantly easier for dropshippers and foreign sellers (e.g. from China) to make timely deliveries on sales. Obviously, this wouldn’t be possible if Amazon didn’t already possess a huge network of warehouses to enable such colossal levels of stock-keeping.
Hence, we can conclusively attribute early Amazon’s success to their outsized Fulfillment expenses, which we now know to be CAPEX for the purpose of acquiring warehouses and developing an in-house HAS distribution channel - in order to secure those indispensable economies of scale.
Now, that’s all fine and dandy, but how does that really help us in terms of making an investment decision?
I think most of you already know the answer to this by now…
As we’ve seen above, the primary pain point for those who wanted to invest in early Amazon was the extended PE multiples of 30x - 90x between 2004-2010 - or an average PE ratio of 60x. The reasons for their extended earnings multiples in the half-decade of 2005-2010 was due to their general lack of profits despite growing revenues. This led to the perception that maybe Amazon was just burning cash in futile conquests against much larger, more well-resourced foes.
However, if we assume that Amazon’s Fulfillment expenses actually represented CAPEX rather than OPEX, then it logically follows that we should back those expenses out from Net Profit to arrive at their look-through earnings.
How do we do that? Firstly, it’s important to recognize that not all Fulfillment expenses represented CAPEX. Sure, the Operating Leases for warehouses could be attributed to CAPEX, but the salaries for the employees working in those warehouses certainly couldn’t. Neither could the fuel expenses for the Prime delivery fleet, or insurance expenses involved in FBA. All those were certainly very much OPEX in nature.
Hence to be conservative, I didn’t back out the entirety of Fulfillment expenses, sticking merely to adding back the amounts for Rental expense under operating lease agreements (‘r’) to Net Profit. Further, since I couldn’t properly identify how much of Depreciation expense could be attributed to growth CAPEX vs. maintenance CAPEX (given that they were overbuilding in anticipation of growth), I simply added back the entire amount of Depreciation to arrive at a sort of “ex-growth” Operating Income figure. The idea was to give a standardized figure to others who wanted to build on this model, which they could then remove their own estimate of growth CAPEX from (i.e. some % of depn + rent) in order to arrive at look-through earnings.
The result is the line item in the dark green font, NP + depn + r = EBDA. I understand that some will nitpick over my usage of the term EBDA (since it usually includes rent/lease expenses), but again I wanted to use a familiar term that everyone could get behind and suited the context. Just understand that this EBDA figure is supposed to represent Amazon’s “ex-growth” Operating Income - in much the same way that EBITDA represents standard Operating Income.
Now, rather than using the P/E ratio as an earnings multiple to value early Amazon, let’s use P/EBDA. What might the results suggest?
As we can see, the P/EBDA figure reveals a much more palatable earnings multiple of 25x - 40x between 2004 - 2010, with an average P/EBDA ratio of 35x. While it’s true that “true” Depreciation OPEX hasn’t been deducted yet, the average look-through multiple would be closer to 40x - compared to the average P/E ratio of 60x. And while 40x doesn’t sound cheap either, it is certainly much more easily justifiable when considering the context described above and early Amazon’s gangbuster average revenue growth of 30%; plus the fact that they were consistently FCF-positive throughout and possessed a negative cash conversion cycle.
Certainly, had I been presented with the above figures of early Amazon in 2005, I would have wanted to spend a little more time understanding the company rather than immediately writing it off as too expensive. By looking beyond just the headline figures and making some bespoke adjustments, it is definitely possible to see things that others may not - driving another nail in the coffin against the existence of EMH. Considering that the reward would potentially have been this:
… it would have been worth the extra effort, wouldn’t it?