The End of Infrastructure

Cloud services providers have historically been built with an infrastructure component and a platform component.

Infrastructure-platform companies will continue to grow, but newer cloud providers build abstractions for developers that do not expose infrastructure.

The simplicity and focus of these newer cloud providers give them a competitive advantage against infrastructure-platform companies, while the infrastructure-platform companies have richer, more complex offerings.

The future is bright for every cloud provider who offers unique services rather than fighting at the margins for competitive advantage in an existing market.

The Infrastructure Era

Infrastructure-platform companies like Amazon, Google, and Microsoft represent an era where most developers need to interact heavily with individual machines.

Digital Ocean co-founder Moisey Uretsky portrays the future of cloud providers relative to the past:

Firstly, clouds are built backwards, not forwards. This is the main mistake that just about every company that tries to compete in the cloud makes. Amazon built AWS internally and then decided to resell it to developers. That gives them several advantages. First, they are able to capitalize on the scale of their operations from the primary core business which means they can start with an extremely large amount of capacity from day one. Because this service is needed internally to run their core business, even if the cloud fails publicly, they will still continue to invest in it internally so it minimizes their risks dramatically…

Clouds with a public infrastructure component have historically been “built backwards, not forwards”, if we are talking about Amazon, Google and Microsoft. These big three are the current icons of the Infrastructure Era: when cloud computing means offering access to both a platform as a service and the infrastructure beneath it.

Developers are suspicious of Amazon, Google, and Azure because the platform-as-a-service can lock them in.

Digital Ocean capitalizes on the anxiety of the Infrastructure Era by offering the simplest version of infrastructure with the lowest switching cost.

Digital Ocean decouples its brand from the platforms built on top of it by partnering with companies like GitLab and Mesosphere — Digital Ocean gets to evolve its PaaS offerings without brand risk or the risk of technical debt from building its own APIs.

Platform as an Infrastructure as a Software as a Service

For the past two years, Dropbox has been ridiculed by critics for its lack of new features, other than acquisitions.

Dropbox’s core product development looked stagnant because the core engineering team was focused on something that was not user-facing. Dropbox was undergoing a massive project — quietly migrating its cloud from AWS to Dropbox’s own data centers:

The whole process took two years. A project like this, needless to say, is a technical challenge. But it’s also a logistical challenge. Moving that much data across the Internet is one thing. Moving that many machines into data centers is another. And they had to do both, as Dropbox continued to serve hundreds of millions of people. “It’s like a moving car,” says Dan Williams, a former Facebook network engineer who oversaw much of the physical expansion, “and you want to be able to change a tire while still driving.”

Dropbox just announced its first big consumer feature since building new infrastructure. Project Infinite might have been cost prohibitive if Dropbox had not built its own data centers.

Now that Dropbox is done changing the tires on the moving car, there will likely be a long line of products that will have Dropbox critics going through blog archives and editing their bleak premonitions.

Dropbox exemplifies why the narrow lens of commodity is the wrong way to look at the cloud.

Zero-sum analysts have been framing Dropbox as a narrowly defined competitor to Box. Box encourages this, because Box imagines a world in which work and life are cleanly partitioned — Box is for your work stuff, and nothing else.

The lines between work and life are increasingly blurred for knowledge workers. By solving for the consumer, Dropbox future-proofs for the enterprise — and, inevitably, the developers.

In the coming years, Dropbox’s cloud will eat into market share of AWS, GCP, and Azure. The most obvious reason is user experience. Look at all the icons on the AWS dashboard:

Azure and GCP also have feature-heavy UIs.

Critics have mocked the spare interface of Dropbox, but it is much easier to add icons than it is to remove them. Developers will choose Dropbox for the same reason consumers use it: simplicity.

Why Azure is Underrated

Digital Ocean’s Moisey extends his history-focused perspective to Microsoft:

Looking at each [modern cloud provider] individually we see that IBM’s play to acquire softlayer was mainly to preserve their revenue base off of their existing customers. Microsoft, very much ended up in the same scenario as well. Given the large amount of revenue both companies have in the enterprise segment, these offerings will generate revenue, however, their growth will be slow, because they are simply canabalizing their own revenue and preventing it moving to AWS.

Microsoft is actually expanding the market of enterprise knowledge workers, not cannibalizing it.

As Amazon decays into Oracle and Google blossoms into AWS, Azure is doing something unique. Microsoft wants to turn business analysts into developers.

Solving this problem doesn’t require any fundamental technological breakthroughs, just tenacity and iteration — which money can buy. Since Microsoft has plenty of money, there is very low execution risk to Azure’s strategy.

Enterprises who are locked into Microsoft will have their cadence of migration to the cloud dictated to them by Microsoft.

A well-educated young investment banker recently said to me “the Microsoft Office suite of technologies is just unparalleled”. This statement is true to the extent that every company he has worked for uses Microsoft end-to-end.

An average software engineer is steeped in open-source, and the Azure stack seems confusing. But an average enterprise CIO is not steeped in open-source, and does not speak in engineering terms.

“Fast data”, “data lake”, “data-at-rest”, and “data-in-motion” seem like buzzwords to engineers, but they are the language which the enterprise CIO understands.

Similarly, Event Hubs and Stream Analytics may not excite a software engineer because these technologies aren’t built to empower software engineers — they are built for the business analyst.

Clearly a business analyst has no desire to write code — but even among developers, there are signs of growing infrastructure fatigue: containers, serverless architecture, and No-Ops suggest a desire to deal with smaller modules, less burdensome configuration, and simpler APIs.

Infrastructure-as-a-service is today’s assembly code. With Azure’s strategy of software development without code, Microsoft might end up building something uniquely beloved by developers.

Long Live Infrastructure

As AWS, GCP, and Azure compete at the infrastructure-platform level, newer companies are building simpler abstractions with better interoperability.

Stripe and Uber will eventually have developer platforms with the feature cardinality of today’s AWS. Today, Stripe and Uber are built on AWS, but they can always backfill their own infrastructure like Dropbox if they need to compete for commodity pricing.

Amazon, Google, and Azure look similar right now, but over time they will be as differentiated as Stripe, Dropbox, and Uber.

AWS will cater to legacy customers and developers who don’t want to learn a new cloud stack. Google’s cloud will be a better, cheaper AWS — the Gmail to Amazon’s Hotmail. Azure will be the cloud for people with a suit and tie.