Technology watchers have been conditioned by the success of various hardware and software platforms to think of every technology market as being defined by network effects; namely that the more users of a platform, the more valuable it becomes to both users and providers. That economic principle certainly holds for things as diverse as Ethernet and Facebook.
The problem comes when people conflate network effects with monopoly control. Indeed, these two examples illustrate that success driven from network effects might (Facebook), but doesn’t necessarily (Ethernet) result in abusive market manipulation. Crossing that line requires one or a few dominant providers to create artificial barriers to entry while simultaneously erecting artificial obstacles and friction that increases the cost of switching to a competitor.
Perhaps the fact that the major players overwhelmingly dominate other markets in online shopping, search and PC operating systems arouses monopolistic comparisons and creates the assumption that cloud infrastructure is destined for a similar fate. However, recent news shows that the market for cloud services is a vibrant example of a market benefiting from network effects, but lacking the abusive tactics of true monopolies. Indeed, the cloud is full of vigorous competition that serves to expand the overall market, not merely reshuffle a limited supply of revenue between a few dominant providers. In sum, the cloud is not a zero-sum game.
Microsoft, Amazon cloud revenues continue to surge
I risked stating what should be obvious because recent earnings reports from Amazon and Microsoft showed some relative changes in each company’s cloud services business that some over-analyzed as two companies heading in opposite directions. If one assumes that economies of scale, network effects and lock-in barriers to service substitution destine the cloud for a winner-take-all future, it’s easy to interpret such changes as implying that one vendor’s success must necessarily cause another’s eventual failure.
The reality is that both AWS and Azure are growing at robust double-digit rates that are moderating due to sheer size. Such a slowdown is the natural result of what my former Bell Labs colleague and current Forrester Research Director Glenn O’Donnell terms “fiscal physics,” namely the eventual impossibility of maintaining geometric growth rates as a business reaches revenues of 8- or 9-figures.
The proximate cause for the cloud comparisons was Amazon’s recent Q2 2019 earnings report in which the company reported a 29 percent increase in AWS operating income on a 37 percent increase in revenue. While still stellar for a $30 billion business, the results precipitated several questions on the company’s earnings call probing the reasons for decelerating growth and declining operating margins. Indeed, the $8.38B in revenue was about 1.5 percent shy of analyst estimates. When combined with overall Amazon revenues and guidance that fell short of Wall Street’s hopes, the result was a 2.2 percent drop in Amazon’s share price the following day.
Source: Amazon Investor Relations
Source: Yahoo Finance
The unfavorable AWS comparisons were also a consequence of an outstanding Microsoft earnings report a week earlier, which saw its overall cloud revenue hit $11.4 billion, with Azure revenue popping 64 percent. Ironically, despite Azure’s growth rate declining 25 points since the same quarter a year ago, there was no analyst hang-wringing on the earnings call about decelerating growth. Indeed, the relative decrease in growth was larger for Azure than AWS, yet the headlines focused on Microsoft’s absolute cloud growth, not comparisons to prior quarters.
Objectively, both Amazon and Microsoft have enviable cloud businesses that are each vigorously growing. While it is impossible to directly compare them due to the way Microsoft mixes in other service revenue in its reporting business segment, it’s reasonable to conclude that:
- AWS is still larger than Azure, but the gap is closing
- Microsoft’s SaaS business (Office 365, Dynamics 365, Teams) is much larger than AWS’s budding applications business
- Microsoft’s PaaS, by exploiting tooling and APIs already familiar to Windows developers, is more popular, particularly among enterprises, than AWS’s platform services.
The latest earnings results suggested to some that Azure might eventually unseat AWS as the largest provider of cloud infrastructure, however that’s both unclear and irrelevant. Each will see their cloud business contribute a growing share of their respective company’s profits as enterprises continue the shift from DIY operations to managed services.
Google not going away, VMware playing Switzerland
Despite the recent focus on Amazon and Microsoft, the cloud market is not a duopoly as a reinvigorated Google Cloud is aggressively pursuing enterprise business. As I detailed last spring, the company is using its expertise in containers to smooth the migration of enterprise workloads to the cloud via its Anthos service. Accordingly, the company just announced the beta availability of Migrate for Anthos, which repackages VMs running on-premises or as cloud instances into containers running in Google Kubernetes Engine (GKE). Originally stated to only support Google Compute Engine, the migration service has been extended to AWS (EC2into containers running in Google Kubernetes Engine (GKE).
EC2 and Azure VMs.
Google has also enhanced Anthos with support for its Traffic Director service mesh controller, which manages container networking and load balancing for microservices. The company simultaneously announced the beta availability of Layer 7 Internal Loads Balancer (L7 ILB) for containerized legacy applications along with the general availability of two high-performance and -availability networking services, namely High Availability VPN, which provides a 4-nines SLA (versus 3-nines for standard VPNs) and 100Gbps Dedicated Interconnect.
Source: Google Cloud
Amidst the slew of product releases, Google also announced an expansion of its partnership with VMware in the form of a new service that enables running VMware workloads on Google Cloud. Designed and operated by CloudSimple, which offers a similar service for Azure, with front-line customer support provided by Google, the service includes the core VMware infrastructure components vSphere, vSAN and NSX.
The cloud infrastructure business has never been healthier and while growth rates might be decelerating for the two front runners, the absolute increase in revenue is not. Indeed, from Q1 to Q2 2018, AWS revenue increased $663 million (12.2% Q-to-Q), while over the same period this year revenue rose $685 million (8.9%). While we don’t have an exact comparison for Microsoft, if we assume that Azure makes up at least 30-40% of the Intelligent Cloud segment’s $40+ billion in revenue, that makes it roughly half the size of AWS or about a $15 billion Ebusiness that’s adding $9-10B annually. Thus, don’t be surprised when both Amazon and Microsoft announce the next major cloud milestone: passing $100 billion in annual revenue.
Think such numbers are hyperbolic? Consider what Intel CFO George Davis said on his company’s Q2 earnings call last week (emphasis added):
Cloud revenue was down 1% year-over-year as cloud service providers absorbed capacity after growing demand 40% in 2018. Enterprise and government revenue declined by 31% with particular weakness in China, while communications service providers revenue increased 3% year-over-year.
CEO Robert Swan added these cautionary words (emphasis added):
Enterprise and government has been brutal through the first six months of the year. Q1 was really soft, Q2 was even softer. And while we don't like it, it's been pretty much in line with how we expected the first half and even the second half of the year to kind of play out. A couple of dynamics, you'll remember, last year was really strong for enterprise and government, growth was much stronger than we expected...So our thoughts through the first half of the year and even going into the second half of the year is the EMG [enterprise market group] environment won't get dramatically better, cloud will get a little bit stronger and comms will get a little bit stronger in the context of our overall full year guide.
Source: Intel Q2 2019 Earnings Presentation
As Swan mentions, some of the enterprise softness is rightly attributed to increased spending discipline by companies facing an uncertain economic climate. However, the contrast with cloud infrastructure spending is stark with the implication being that more and more enterprise IT is going to cloud service providers, not direct investment.
In sum, the cloud infrastructure and applications market is a critical foundation of the digital economy that resembles big oil more than it does natural monopolies like water and power. Network effects help the providers by making their applications more useful, while the increased usage funds the significant investments required to build world-class data centers and develop innovative new AI and analytics services. There is plenty of business for three major vendors to fight over without existential risks to any of them, with enough leftover for smaller providers targeting niche markets and specialized needs.