A couple of weeks ago Amazon announced its quarterly numbers. As has been the case over the past year or so, the numbers looked good. Really good. Derided for years as a profitless company propped up by investor largesse, Amazon grew its revenues by 31 percent, from $23.9 billion to $30.4 billion, while profits leapt 832 percent, from $92 million to $857 million.
Most of the profit came from AWS: on $2.88 billion in revenues, AWS reported $718 million in operating income. In Q216, AWS grew 58 percent year over year (YoY), down slightly from Q1’s 64 percent, but still healthy. As I wrote earlier this year, AWS’s curious failure to align with Amazon’s overall low-margin approach to pricing indicates that it is deliberately keeping prices high to avoid further increasing customer demand. Said another way, AWS’s growth is governed by capacity, not customer demand — which means we can expect it to continue its 50 percent growth rate for the next several years.
Azure is a bit tougher to figure out. Microsoft claims its cloud business is over $10 billion, but that includes services like Office 365. The company does say that Azure is growing 120+ percent. But it’s hard to say how much of that $10 billion is Azure proper. The Register says Azure revenues were about $800 million in the quarter ending June 30. To be conservative, we could dial that back to around $550 million per quarter, or around $2.2 billion for the year.
If one accepts that AWS is going to continue growing at around 50 percent over the next several years, with Azure continuing its triple-digit pace during the same period, what does that end up looking like by 2020? Here is a chart:
If their growth continues as predicted, the two cloud giants will end 2020 at around $96 billion in total revenues — just shy of the magic $100 billion mark. Of course, you can quibble with the total — maybe it will only be $80 billion — but no one has clearly stated any rationale as to why the underlying growth figures for AWS and Azure will drop, and that’s the only reason their cloud revenues won’t be north of $75 billion by the end of 2020.
The other relevant number? Private cloud, which is not setting the house on fire.
Analyst firm Wikibon believes that no vendor is making more than $100 million via OpenStack. If that’s anywhere near true, the sum total of all vendors has to be less than $2 billion.
Private cloud is a concept propped up by legacy vendors desperate to provide enterprise IT groups a reason to keep buying their servers, storage, and network products, and that’s where the problem lies. It’s a solution in search of a problem. And not a very effective solution at that.
While private cloud proponents have spent the last five years focusing on getting their IaaS offerings working, the big three cloud providers have moved way beyond core computing services. They’re delivering the services IT groups will need in the future to keep their companies from being eaten by software.
Google, although its revenue is still small in comparison to AWS and Azure, offers an incredibly interesting machine learning set of services. I’ve worked with them, and they offer tremendous power at an affordable price, delivered in an easy-to-use framework. It’s clear we’re at the beginning of an AI-powered revolution, and Google is staking its claim to be the pioneer in the field, as demonstrated by its Deep Mind offering defeating the world’s champion Go player.
Microsoft is gearing up to make a big run at blockchain, with its Project Bletchley. I have written about Bletchley before, and it seems clear from what Microsoft has published that the company views blockchain as an enormous opportunity that will leverage Microsoft’s strength in middleware, cryptography and programming languages, not to mention its preeminent brand positioning, which will make companies getting started with blockchain initiatives comfortable working with a nascent technology. It’s early days for blockchain, but the technology holds the potential of restructuring a vast range of financial and governmental services.
Meanwhile, AWS has seemingly stumbled into being the hub of the smart home and workplace with its surprise hit, the Echo, powered by AI agent Alexa. Users can create and share Alexa skills by tying it into APIs front-ending external products and services, as in this video showing how one guy wired up Echo to his Tesla, automating the process of starting the car and backing it out of the garage.
What all of these offerings have in common is that they sit on top of cloud computing, but they are not IaaS services. The cloud providers are using IaaS to build the services that lie at the heart of the future of enterprise IT.
There are three elements regarding these services that are key to why they sound the death knell of the private cloud:
- Innovation. The cloud providers hire smart, really smart people and set them to solve interesting challenges. Supporting these people financially and placing them in organizations where they’re constantly exploring interesting problems means that the cloud providers create new functionality that legacy vendors with a private cloud could never discover the need for — and wouldn’t be able to create even if they understood the need.
- Scale. Machine learning is famous for needing massive data sets to train against. Having massive user bases solves the problem obtaining these data sets. Likewise, operating these higher-level services requires huge server fleets, well beyond the scale any single enterprise could hope to assemble.
- Network effects. By supporting lots of different data inputs and customer use cases, these cloud providers enrich their offerings. In essence, by providing your idiosyncratic needs, you help improve the quality of the services and define the need for new ones. These network effects are well beyond what one organization could possibly create within its own infrastructure.
The core problem for private cloud is that the providers are focused on delivering what solves their problem — pushing more hardware. This blinds them to the importance of top of stack services and causes them to spend endless energy on tweaking the infrastructure software to incorporate their latest switch. Or server. Or storage array. This is textbook Clayton Christensen material — and we all know how that story turns out.
Enterprise IT groups trying to solve the pressing business problems of their companies have to use these innovative “top of stack” cloud provider services — there’s no other way to access them. Certainly a few racks of a private IaaS cloud cluster can’t hope to deliver the same functionality. It’s impossible.
There’s no question we’ve reached peak data center. The only question is how steep the decline will be. Five or ten years from now private cloud will seem like a mirage that once fooled the industry into believing water lay just up ahead.