Tag: Amazon

What is AWS worth? Thoughts on the business and whether it should remain part of Amazon $AMZN

As cities across the U.S. embarked on huge marketing campaigns to attract “HQ2”, I couldn’t help but ask myself why Amazon was planning a second headquarters. Usually companies like coherence and for the business to act as one under one roof. It is more difficult to do this when you have two “headquarters”. To me, it reminds me of having two CEOs. Two voices. Two cultures.

I came to the conclusion that Amazon was preparing for a split up of the company. The traditional retail business + Whole Foods would go under one umbrella and continue to operate out of Seattle, while the second business would be Amazon Web Services (AWS) and would operate out of its own hub and have its own resources.

For those that do not know, AWS is much different than the retail business. It is designed to provide developers quick access to on-demand computing power, storage, applications and development tools at low prices. Thanks to its offering, developers, startups and other companies do not need to focus on infrastructure build out when they are starting a new venture, they can instead focus on their core product and utilize AWS’ resources. This allows faster speed to market in some cases. It also allows this spend to be a variable cost, rather than a high, upfront capex spend.

This led me to ask myself:

  • What do I think AWS is actually worth? Does Amazon’s valuation today reflect its strength?
  • Is AWS actually a good business?
  • Would splitting the company make sense?

I think I am going to go through the answers in reverse order. I do think it makes sense to split the company. AWS operates in a competitive environment with the Microsoft Azure, Google Cloud, IBM among others clamoring for a larger piece of the pie. AWS currently has top share of wallet, but it’s a much different business model than the retail business. Here are my three main reasons for splitting the company:

  • Allow focus in areas of strength.
    • Being a company focused on one goal helps keep all employees realize what the goal at the end of the day is: support customers.
    • Today, while a smaller part of the market, there is a conflict of interest with AWS and CPG sectors (e.g. Target announcement to scale back AWS, Kroger announcing post-Whole Foods it would be moving to Azure and Google)
    • Typically in conglomerates, new business in certain sectors can go unnoticed or underappreciated as “approvers” up the chain of command have other things on their priority list. Splitting the company reduces this kink in the chain. This appears to be limited risk today as Amazon is growing AWS and its services, but think about the other segments. Perhaps they are moving slower now in the retail segment or internationally than they otherwise would.
  • Optimize capital allocation
    • AWS is highly profitable from a EBITDA perspective (though is also highly capital intensive today as it grows). On the other hand, Amazon’s retail business is not only low margin, but also is capital intensive and its return metrics are much lower than AWS (especially internationally where the company still has significant operating losses).
    • Even if I think returns for AWS will come down over time (discussed more below), you could view the company as funneling good money after bad (i.e. taking resources from a high return business and pushing it into low return business). If I invest $1 dollar into AWS, I get 28 cents back in year 1 compared to 9 cents for North American retail or losing 11 cents Internationally.
    • You may be optimistic about the LT trends in the retail business and its improving profitability over time, but think about if AWS could funnel more of its cash into higher return projects. It is likely to perform better over time on its own. Freeing up these resources can help AWS invest in its core capabilities and build sustainable competitive advantage.
  • Better align incentives
    • Distinct and clear business with their own performance goals matter when you are paying your employees in company stock
    • Imagine you are a leading employee in the field and asked to work for Amazon. A significant portion of your comp will be in restricted stock that pays out if the entire company does well. But oops, something happens to the grocery business at amazon and the stock falls 20-30%, so now you are well out of the money.
    • Wouldn’t it be better if your performance and the company’s performance were linked?

Is AWS a good business?

“I believe AWS is one of those dreamy business offerings that can be serving customers and earning financial returns for many years into the future. Why am I optimistic? For one thing, the size of the opportunity is big, ultimately encompassing global spend on servers, networking, datacenters, infrastructure software, databases, data warehouses, and more. Similar to the way I think about Amazon retail, for all practical purposes, I believe AWS is market-size unconstrained.”

-Jeff Bezos, 2014 Letter to Shareholders

Clearly, AWS has grown like a weed. Amazon has disclosed the results of this segment going back to 2013 and we can see sales and EBITDA have grown at a ~50% CAGR since then and 2018’s pace actually picked up over 2017.

The business started when Amazon brought infrastructure-as-a-service (IaaS) with global scale to startups, corporations, and the public sector. By this, I mean that they provide the infrastructure that allow these companies to outsource computing power, storage, and databases to Amazon. It essentially turned this services into a utility for consumers, with pay-for-what-you-use pricing models.

As stated earlier, this allows customers to add new services quickly without upfront capex. Businesses now do not have to buy servers and other IT infrastructure months in advance, and instead can spin up hundreds of servers in minutes. This also means you benefit from the economies of scale that Amazon has built up. In other words, because Amazon has such massive scale, it can pass some of those saving on to you. If you bought the servers and other utilized ~70% of the capacity, your per unit costs would be much higher.


AWS has attracted new entrants as well. Microsoft’s Azure and Google’s Cloud have also grown and attracted share. Microsoft provides some color on the business, which I have estimated in the table above, but Google provides basically no disclosure. Other players with smaller market share include Oracle, IBM and Alibaba in Asia.

Despite growth in Azure, Amazon owns more market share than the next 4 competitors combined. And the rest of the market appears to be losing share.


The logical outcome from this, especially when considering how capital intensive the business is, is that the business forms into an oligopoly. Oligopolies traditionally have high barriers to entry, strong consumer loyalty and economies of scale. Given limited competitors, this typically results in higher prices for consumers.

However, this is not always the case and it leads me to some concern on the industry. In some cases, oligopolies become unstable because one competitor tries to gain market share over the others and reduces price. Because each firm is so large, it affects market conditions. In order to not cede share, the competitors also reduce price. This usually results in a race to zero until the firms decide that its in the best interest of all players to collude firm-up pricing.

How could this happen with Amazon Web Services? Well, we all now Amazon’s time horizon is very long and it attempts to dominate every space it is in by capturing share with low prices. As said in its 2016 letter to shareholders:

We will continue to make investment decisions in light of long-term market leadership considerations rather than short-term profitability considerations or short-term Wall Street reactions.

Jeff Bezos, 2016 Letter to Shareholders

In addition, other competitors have benefited from movement into the space and have been rewarded in their stock price (see MSFT). Ceding share to AWS may hurt their results (and stock price) so they may attempt to take more and more share at the expense of future returns. In a way, what could happen here is more of a slippery slope. They give a bit on pricing because the returns are so good now that a small price give here and there want matter. But then they add up…

Lastly, its no secret that building data centers to support growth is capital intensive (see AWS financial table above). They also serve a lot of start-ups and other businesses. This creates a large fixed cost base for the cloud players. If there was a “washing out” of start-ups, such as the aftermath of the tech bubble, AWS and others could see excess capacity, hurting returns. In order to spread the costs, they may attempt a “volume over price” strategy. This will result in retained share, but hurt pricing for the industry as competition responds.


Unfortunately, I DO think returns are going to come down materially over time. The industry reminds me too much of commodity industries. The industry is no doubt growing today, but the high margins we see today I believe are from high operating leverage. This tends to be more transitory in nature and means that returns on incremental capacity invested go down over time.

Take for example commodity chemical companies. AWS reminds me of some of their business models (unfortunately). Some operate in very consolidated companies, but operate at very high fixed costs and have exit barriers. The loss of one customer results in very painful results due to operating leverage. Although higher prices are better in the long run, the company will lower price to keep the volume and keep the fixed cost leverage. Think about commodity businesses that report utilization (note, we have limited disclosure on AWS, but that might be a helpful statistic in the future). Higher utilization usually results in good returns on the plant (or in the case of AWS, server). However, when you lower price on the commodity, the competitors, in turn, will also lower price so they don’t lose their existing customers. What results in lower returns for everyone.

As a another case study, P&G used to run its HR, IT and finance functions within each of its business units. It decided to form one, centralized, internal business unit to gain economies of scale. It actually formed a leading provider of these services. But then, P&G looked at its core business and said, “do I need to continue doing these functions or can I outsource it? Or maybe I can sell this whole business unit to another provider, which would give them more scale and lower the cost of the function”. The latter is what ended up happening. Although it wasn’t viewed as a commodity at the time, these functions began to become more commoditized and business process outsource companies (“BPOs”) now have very high competition and lower returns.

Another analogy would be to at the semi-conductor or memory businesses, though perhaps not as commodity as these two.

I personally think that over time, we will see this form with AWS. The fact that pricing has come down so much for the service (one unnamed engineer I interviewed told me that AWS will even let you know if they think pricing is coming down soon to keep your business). According to the Q3’18 earnings call, Amazon had lowered AWS prices 67x since it launched and these are a “normal part of business.” They’ve certainly been able to lower costs as well so far, but how will that change as competitors also gain scale?

How could I be wrong? I could be wrong in many ways, but the one case is that I am underestimating how “entrenched” the business is into everyday use. That could result in very low switching by customers. It could also mean that developers get trained using AWS so default to using its services. The counter to these arguments is that there is a price for everything and clearly Azure has taken share over the past year, so it’s not proving out thus far.


So is it a good business? Well, lets look at at it from a Porter’s Five Forces perspective. I think the view on these, in brief, plus the current returns we can see in the previous tables would point to the industry being solid. My one concern comes back to the competitive rivalry. We can have a high barriers to entry business and highly consolidated, but returns are not good.

What do you think AWS is worth?

Based on the market share numbers posted above, AWS is currently attacking a $73BN market. At the rates that the market has been growing, as well as the additional services, I think its reasonable the market expands to $100BN in the near term, especially as adoption increases. I assume Amazon maintains its market share due to its relentless focus on lowering price and maintaining share. I think the additional share gained by Google and Azure will come at the expense of the other players included in the market share chart above.

For the reasons stated above, I think this results in returns coming down over time. I still model good returns to be clear, but that the return on assets comes down meaningfully.

This may be hard to read, but I end up with a valuation of $164BN for AWS. Given the value of AMZN today is around $830BN even though AWS accounts for over half of EBITDA now, I cant help but think this opportunity is more than priced in, unfortunately. However, I maintain that the business would benefit in the long-term from being on its own for the reasons discussed above. In the short-run, it may get a higher multiple than the rest of the Amazon retail business since its growing quickly with such high margins.

What is Twitch worth? A look at what you need to believe…

A good friend of mine came to me full of excitement about the prospect of Amazon’s “hidden gems.” He had heard Amazon owned Twitch and was very excited for what that could mean for the company given the success of Fortnite and other games, which have driven a new era for e-Sports (that is, watching other people play video games). Twitch could also serve as a platform for other content. I decided to take a moment and explore what Twitch might be worth and put into context what that may mean for Amazon’s share price.

Warning: As with any investment analysis, I make a lot of assumptions in this post given I have very limited disclosure from Amazon.

With that warning out of the way, I hope that you can take these assumptions with you and think about “what do I need to believe?” when it comes to Twitch (and Amazon) and determine whether or not the market is pricing this in (as I wrote recently about here). Let’s get started:

Most internet platforms like Facebook, Twitter and YouTube are analyzed by a popular KPI called, Monthly Active Users (or MAU). MAU is important, and different than valuing a company via “eyeballs” as was common in the tech bubble, because MAU measures engagement with the platform and that helps advertisers determine whether placing an ad on the site has impact or not.

According to a very trustworthy source, Wikipedia, Twitch had 100MM MAUs in 2015. I assume that has likely grown significantly since then, given the dawn of Fortnite and internet celebrities like Ninja. I assume 130MM MAUs for 2017, which compares to Twitter’s 328MM.

As for Average Revenue Per User, or ARPU, I think Twitch is still in its infancy stage. I don’t think it is likely capturing as much money as it can right now because (i) it is owned by Amazon which has a long investment horizon and (ii) they likely want to keep engagement up and growing as much as possible in the near term to drive stickiness with the platform. However, I do have it growing substantially over time. In fact, this may be an aggressive assumption given Twitch has a monopoly on this niche for now, and competitors may move in. In addition, advertisers can move elsewhere, such as Instagram, Facebook, Twitter, Youtube, and so on. There is only 1 advertising pie and all these players are competing for the biggest slice.

On the flip side, Twitch should benefit from a “live TV” aspect of its content, much like ESPN which is able to charge a large premium in the broadcast world compared to its peers. As a result, I expect ARPU to ramp up quickly. Twitter’s total advertising revenue / MAU ramped from $2.20 in 2013 to $5.30 in 2017. I have Twitch scaling much quicker than that given what I’ve noted above.

For now, I don’t think Twitch is a major contributor to EBITDA given investment in R&D, sales staff and general expenses, but with price gains comes scale. For reference on how I derived my assumptions, I looked at Twitter. Twitter in 2015 spent ~36% on R&D, 39% on sales and marketing, and ~12% on G&A (though some of this includes a massive stock based comp expense, at ~30% of sales). This has stepped down to R&D being 22%, Sales being 29%, and G&A being 11.6% (again, these expenses include an 18% of sales expense for stock comp, which is non-cash but is a real expense at the end of the day).

Twitch est

Therefore, you may view my assumptions still as aggressive, but I’m trying to show what you need to believe. Net / net, I have Sales growing at a 127% CAGR from 2017 to 2020 and EBITDA growing at 450% (albeit off of a low base). I’ve also included some multiples so far as a proxy for where the value of Twitch may shake out.

I do not know where Amazon puts Twitch today in its reporting, but I’ll assume it is in the North American Retail business for now, given I don’t have a better idea of where to put it (I know that AWS is distinct from any ad revenues though so I didn’t put it there). here is a snapshot of Amazon today then, including my rough Twitch estimates.

AMZN Today - Segments

Now, for my next assumptions, I am going to go with some street estimates here, but a big assumption on my end is that I do NOT think they are baking in the value of Twitch. That is a big “if”, but given how small Twitch is today relative to the rest of the business, I don’t think it is out of the realm of possibility. Below are the 2020 estimates for AMZN, including my assumptions in value for Twitch.  All in all, it shows pretty strong growth for a company that has already grown massively. AMZN 2020

I took the liberty to assign multiples that you may disagree with, but I have a sensitivity table later that will let you be the judge.

The problem you may be seeing is that the total EV of Amazon of slightly less that $600BN is less than the current ~$890BN. As shown below, my “price target” is well below the current price.

AMZN PT

Uh ok… Ok OK I must have done something wrong. My multiples must be WAY off. That’s for you to judge and that is the “art vs. science” part of investing. As we see below, I think I am pretty comfortable given where other Tech titans trade on my multiples, especially when we look out 2.5-3 more years.

AMZN 2020

So what do you need to believe??

Below I show a sensitivity of AMZN’s stock price relative to (a) changes in my EBITDA estimate and (b) estimated changes in the multiple. If you think Amazon is worth 18x EBITDA and will produce $50BN of EBITDA, for example, then today’s price of $1,830 looks OK (if not a little rich given this PT is based on 2020 estimates).

AMZN Sensitivity

Does this seem reasonable to you? It may, or it may not. That’s part of doing the analysis. After all this you may say, “Yes, actually. Amazon is such a dominant force, with a loyal customer base, I think it is worth a high multiple and the street is underestimating it.” On the flip side, you may say to yourself, “Geez, I don’t know. Those are some lofty figures it will need to reach… Maybe I’ll stay on the sidelines for now.”

Both are equally fair.

That’s all for now. Full disclosure: I am long AMZN.