Once upon a time, the main product of public clouds was infrastructure. That infrastructure was constructed using hardware that the clouds acquired mostly from third-party providers, and the infrastructure hosted whichever third-party software platforms customers chose to deploy on it.
Ah, how times have changed. Today, major cloud providers seem to be on a vertical integration bent. Not only are providers such as Amazon Web Services (AWS) and Google Cloud Platform investing heavily in software products that customers can run on top of their cloud infrastructure, they are now even sourcing their own hardware — which means that, at some point in the not-so-distant future, entire cloud computing environments may consist primarily of hardware and software components sourced from a single vendor.
Here's how vertical integration is playing out in the cloud industry, and what this cloud computing trend means for the future of public cloud.
The Days of Infrastructure-Only Public Clouds
To understand just how much vertical integration is changing cloud computing, you must first understand how the public cloud business model traditionally worked.
Originally, public clouds specialized mostly just in infrastructure as a service (IaaS). They sold virtual machine instances, storage, databases, and the like. To the extent that public clouds provided software, it mostly consisted of basic monitoring and administration tooling, like AWS CloudWatch and Azure Monitor, that worked only within each public cloud's own environments.
The only area where you could accuse public clouds of vertical integration 10 years ago was the platform-as-a-service (PaaS) front. Early on, the major cloud providers rolled out PaaS solutions (such as AWS Elastic Beanstalk, which debuted in 2011) that competed with third-party offerings like Heroku. By offering their own PaaSes, public clouds were able to couple their IaaS offerings with software tools.
But even here, they were targeting a narrow market. To the extent that PaaS vertically oriented public clouds, it did so for limited use cases and limited sets of users.
Public Clouds Become Software Vendors
That started to change around the mid-2010s, when public cloud providers began investing more extensively in software platforms that could run on top of their infrastructures.
Broadly speaking, most of these solutions fell into three main categories:
- AI and machine learning services: Services such as AWS Polly and Azure Cognitive Services have turned the public clouds into purveyors of machine learning software, in addition to infrastructure vendors.
- Data analytics: Through services like Amazon OpenSearch and Azure Data Lake Analytics, the public clouds now let you crunch big data without using third-party software.
- Business productivity: Platforms such as Google Workspace and Microsoft Office integrate productivity software into the offerings of public cloud vendors. Although not all of these services are directly tied to public clouds, some — Amazon Chime — are.
At the same time, the big public clouds doubled down on their investment in PaaS. Alongside basic tools such as Elastic Beanstalk, they built sophisticated managed containers-as-a-service and Kubernetes offerings that — when integrated with managed DevOps tools, like Azure Pipelines, and operating systems, like AWS Bottlerocket — provide developers with everything they need to build, deploy, and orchestrate applications without ever leaving a given cloud provider's ecosystem.
Latest Cloud Computing Trend: Hardware Sourcing
Not content to be just software vendors, the public clouds are now turning themselves into hardware vendors, too, by developing their own computing chips.
AWS, which already offers VM instances powered by chips sourced from the company, is currently the furthest along in this journey. But there are clear signs that Microsoft and Google are close behind.
Admittedly, there will probably always be room for third-party hardware inside the Big Three clouds' data centers. So far, the clouds are only building their own chips, not storage devices, NICs, and so on. It's also far too early to say whether cloud vendors intend to transition completely to chips that they source in-house, or to use those processors only for certain services or instance types.
Still, the move toward internal hardware sourcing is a very big deal. It means that the day will come (in fact, it's already here, at least for certain AWS use cases) when virtually the entire stack required to host a workload — from the CPU, to the VM instance, to the OS, to the software application — will originate from a single provider. Using the public cloud may feel a lot more like using an Apple product — where almost everything is sourced from one vendor — than running, say, a PC where the hardware comes from Dell, the OS from Microsoft, and the software from a variety of other vendors.
What Vertical Integration Means for Cloud's Future
Why are cloud providers investing in vertical integration, and how will it shape the future of the cloud?
It's hard to do more than speculate at this point, but I suspect that one outcome of this cloud computing trend will be stronger differentiations between public clouds. When all the clouds were basically just IaaS providers, there wasn't a lot of difference between them, apart from details like pricing and how VM instances were configured. But when public clouds become software vendors, and when they design their own chips, they have a greater ability to build unique products and services.
Vertical integration is also likely to have major consequences for "alternative" cloud providers, meaning smaller clouds that attempt to compete with AWS, Microsoft Azure, and Google Cloud. The ability to develop and source hardware and software is probably beyond the capability of most alternative clouds, which means they will have to chase customers in other ways — perhaps by undercutting the larger cloud providers' pricing for core IaaS services.
Conclusion
The vertical integration trend in the cloud computing market hasn't received a lot of attention, but it should. It's one of the most powerful forces in cloud computing today, and it promises to have profound implications for the way cloud providers operate and the way customers use their platforms.
