In my discussions with analysts covering the technology sector over the last several years, there are several points of view when it comes to how to value public clouds. I believe that most people covering this space have made the wrong assumption about the future growth of cloud with respect to contributions from enterprise clients and that this will play out in 2017 and 2018.
The value range usually associated by these analysts with the largest public cloud, Amazon Web Services (AWS), is a $125b-$200b market capitalization. This represents up to 50% of Amazon (AMZN) market capitalization based on yesterday’s trading data:
I personally don’t have a specific view of the exact value of AWS with respect to Amazon’s overall valuation but I feel comfortable saying it is a significant portion. Jeff Bezos has certainly built an amazing company that has the capacity to invest in almost any line of business, anywhere in the world.
I firmly believe that AWS has created a revolution with respect to its ability to broadly enable Digital Transformations. AWS has also changed the way that IT operations have been run since the 1980s and these broad transformations are still in the early phases with respect to being broadly adopted by every company around the world. Being the market leader in breadth of service and number of customers is a huge advantage.
However, AWS is certainly in the target sights of deep-pocketed companies like Microsoft and Google and will, in my opinion, now have to react to their moves to go after the same customer base. This includes a very interesting pricing update from Google Cloud that we covered last week that I don’t believe the market has fully grasped with respect to the broader implications of these pricing changes. I will be expanding on this competitive pricing area for the public clouds and related ISVs operate in it next week.
What you are about to read shouldn’t be misconstrued as a statement around public cloud not being the future. You should read the following note as one person’s point-of-view about how the future of cloud consumption economics and patterns may be different than how public cloud has been consumed to date. This may have impacts to modeling future revenue growth if today’s public cloud revenue forecasting assume similar growth contribution metrics from previous trends.
I’m definitely no Warren Buffet, but I will try to model this after the simplicity of his writing style and avoid deep technical concepts. I will insert links to articles that allow the reader to get deeper.
Humans & Bubble Thinking
For as long as there have been investment opportunities there have been bubbles. My view is that we are potentially witnessing a bubble built on incorrect future assumptions of public cloud consumption costs and their impact on margins and revenue growth. Most students of economics are familiar with Tulip Mania.
At the peak of the Tulip Mania, some of the most sought after tulip bulbs were trading for 10 years salary of an average laborer’s wage. Ten years! Today we still love tulips and Tulip Mania wasn’t an indictment against our love of tulips. The bursting of the tulip bubble was an opportunity to understand how the valuation of the tulip bulb had become so disconnected from the intrinsic value of a tulip bulb. We also must learn the lesson that when the true value was understood the correction is sharp and quick:
I want to be clear that I am not forecasting any contraction in AMZN stock value. I like starting with the Tulip example to show how easy it is for us to collectively not vet our investment thesis with rigorous debate and challenge the status quo of conventional wisdom. As I mentioned above, AMZN is more than AWS and Jeff Bezos is an amazingly sophisticated operator and investor. I would prefer to lay out a case about why the cost dynamics for greater cloud adoption may be modeled incorrectly by what is considered conventional wisdom in the cloud space of continuous smooth future growth. My view is that you must look at public cloud providers needing to increase their sales and support investments on a per client basis more than AWS had to for the first ten years for large clients.
The First Decade of Public Cloud (2006-2015) Led by Self Service Startups
People define public cloud to mean many things but for the purposes of this article I will refer to public cloud meaning Infrastructure as a Service (IaaS) offerings.
When I joined AWS in 2010, I heard a lot of common pushback from enterprises in the Fortune 2000 class of companies that were not interested in using public cloud. Most, but not all, had a negative view of public cloud as something too risky (“not as secure as my data center”), not mature enough (“we can run data centers for our business better than a bookseller can”) or too impactful to current IT workflows (“we don’t want to change the way our IT teams work because they are doing a great job already”).
Startups weren’t born because of public cloud, some just happened to realize the public cloud’s value in this time period better than enterprises and even other startups. The following reasons, in my opinion, led to explosive growth of startups that had the best ideas and execution at the time on public cloud:
- Startups in this time period didn’t have any technical debt coming from decades of past IT operations so they were free to choose technology that worked best for them.
- Startups most likely had limited resources compared to established enterprises and wanted to use Lean principles to iterate quickly and cheaply to see how their own clients responded to new products.
- Startups were given an opportunity to convert large, up-front CapEx IT operations purchases (servers, networking, data center leases, long term software licenses, etc.) before they ever made one into the ongoing, consumption-based OpEx investment cycle that public cloud offers. Startups could put their credit cards down and get started with AWS, no contracts required.
- Startups wanted to run faster than their competitors and could consume and pay for what they used in public cloud. Traditional IT departments were procuring IT assets via the traditional sales cycles that non-public cloud IT companies were also racing to preserve. Startups could get a server in AWS in seconds, pay less than $0.1/hr for that usage and turn it off when it wasn’t needed. I worked with some F500 companies that took no less than four months to obtain access through their IT teams for a server in the data center at costs that were 3x the underlying hardware and software to cover internal costs.
- Startups that maximized public cloud were able to convert the labor of traditional IT tasks into automated software code and eliminated labor costs, that didn’t have to be retrained later, in addition to CapEx to OpEx consumption conversions.
Some startups that I worked with at AWS were spending hundreds of thousands of dollars a month tied to a @gmail address and not even sharing details on their startup projects. As long as they weren’t doing anything illegal, AWS gave these startups stealth and cover from their competitors view. The early adopters of IaaS were therefore not established companies but those leading Digitial Transformations and disrupting industries or creating new ones. Early examples of this included Netflix streaming being built on AWS and Instagram’s twelve employees building everything on AWS before selling to Facebook for $1b.
Software developers at these startups could read the documentation of public cloud APIs and transform, what to date, had been people-focused IT operations into Infrastructure as Code. How many employees would a traditional company need to match the output of the twelve employees from Instagram? If one of these startups was growing fast enough and disrupting your company or industry you better believe that CEOs and Boards instructed their teams to looking into migration to public cloud. By the end of this decade of cloud we were starting to see this transition happen with interest outside of the self service startups. By 2015, almost every industry and company in the world was being transformed to serve their clients or by their competition whether they wanted to be or not.
Public Cloud Consumption Trends Today (2015-present)
One of my biggest problems with forecasts of AWS, or general public cloud growth, for the future is that I feel most analysts and people that cover the industry are not contrasting a startup’s ability to consume public cloud historically with how enterprise public cloud projects generally are fairing currently. Startups will still continue to be created and drive consumption of public cloud much like they did in the first decade of public cloud. However in order to meet the future growth assumptions of equity analysts on public cloud, enterprises need to contribute at similar rates and enterprises must sustain their projects over time (not migrate them off public cloud back to data center or to competitor).
Here is a list of my observations of enterprise IT that contrast greatly with startups that consumed public cloud in the first decade on their own:
- Enterprises are more likely to have political structures that they want to preserve in some way when migrating to public cloud instead of re-imagining a new organizational support structure built on top of public cloud.
- Enterprises generally fail to grasp that public cloud was originally intended for companies that were building their own software and that software developers skill sets need to be extended into traditional IT operations teams to maximize investment return, even if these enterprises are not developing their own software. Most enterprises don’t invest fully in the notion of Infrastructure as Code and can spend 50-100%+ more than they should leaving them to make comments like “Cloud is more expensive than we realized when we got started.”
- Enterprises are likely to view migrating or moving to public cloud as a task solely for IT operations. The reality is that it takes a broad level of support from all stakeholders at the executive level.
- In my experience, less than 30-35% of enterprise employees can transfer their traditional IT skills and learn the new skills that are required to successfully and efficiently use public cloud within an 18 month period. New hires are needed.
- Enterprises have made, and in some cases continue to make, investments in traditional infrastructure (firewalls, security software, etc.) that they want to use in public cloud environments and most of these products were never designed to go to public cloud.
- Successful transformations for enterprises moving to public cloud can take 24-36 months and almost always happen much slower than enterprises expect.
- Turnover for cloud engineers is higher in most enterprises, compared to startups, due to the more mature cultures and difficulties in transitioning to “cloud-native teams.”
- If you are an enterprise that is not developing your own software, you will not achieve the same benefits of public cloud to your business model that companies that were built on top of it like Netflix or Instagram. This increases the risk that the economic case that was made to move to cloud will not be fully realized.
- If you are an enterprise that wants to start this process tomorrow, your internal team is most likely not prepared to be successful on their own. You are now competing for cloud operations talent that is most likely more expensive than you are budgeting. Most great cloud services companies also don’t have cloud engineering talent not already working on other client projects.
- “It is not necessary to change. Survival is not mandatory.”- Dr Edwards Deming Some Enterprises have processes and cultures that are so entrenched that public cloud providers, and especially cloud services companies, must not waste business cycles attempting to fit a square peg in a round hole. It sounds simple but when you are public cloud sales exec and you only have a few accounts, it may not be possible to successfully play all of the hands you are dealt.
Why do these enterprise trends matter?
If we look at some of the first significant enterprise commitments that changed the market view of AWS, they happened in 2015; GE and Capital One. These public endorsements, along with Infor’s CEO in 2014 famously proclaiming that “friends don’t let friends build data centers,” signaled that most of the original concerns of public cloud consumption that had held enterprise consumption back were now yielding to its potential benefits and industry evolution(s). Very few Fortune 500 companies though have had significant public cloud workloads for longer than 24 months.
Public clouds will now need to make more significant investments on a per enterprise client basis vs a startup client in my view based on the different operating models of enterprises compared to startups. Public cloud players may already dedicate a sales and services team, in some cases, to a single enterprise client as well as several external cloud services companies (whose work may be subsidized by the public cloud) to assist large enterprises in their migrations. I see proactive programs and responses across all of the public clouds as they continue to scale to compete and capture enterprise cloud workloads, especially workloads that are best use cases for cloud success.
The enterprise trends of migrating legacy IT operations and infrastructure, as well as developing new software projects on public cloud, are very recent at a macro level. If we can broadly say that enterprise transformations take longer than 24-36 months on average and don’t yield the same return as startups, we should also be able to soon see if the projections for enterprise consumption of public clouds on a sustained basis are correct over 2017 and 2018. My bet is that the many analysts that have lumped startup consumption models and enterprise public cloud future consumption together as one and the same are about to get their bubbles popped.
I have more details to share on specific metrics and enterprise-related services that I will be watching across AWS, Azure and Google, please reach out to continue the discussion.