End of Cloud Computing? Didn't it just get started?

I know what you may be thinking. Who is this guy that is already predicting the decline of cloud? On the surface, it would seem that cloud is still riding a growth curve and, in many respects, has "crossed the business chasm", moving from early adopter phase to mainstream acceptance. A quick look at Amazon's quarterly revenues for the past several quarters illustrates the growth of cloud:

Amazon only started the breakout of the AWS cloud revenues in FY2014, which is significant, as it reflected a $1B/quarter business - a long time since its inception in 2006. As stated at this year's AWS re:Invent 2016 Conference, AWS, now 10 years old, is a $13B business with 55% annual growth and accounts for a whopping 100% of Amazon profits.

Similarly, Microsoft has enjoyed a growing business for both Azure and its SaaS assets, Office 365 and Dynamics Online. Azure posted a 102% growth in revenue and (starting last quarter), represents a $10B business for Microsoft, as they predict growth in Azure will account for a $20B business by 2018 - just 2 years from now. Not bad for a business that started 2 years later than its major competitor.

Office 365's growth was holding steady at 54%, but has shown slowing growth over the past few quarters, marking a plateau of growth for the subscription service. Since its inception in October of 2010, Office 365 now boasts over 23M subscribers.

With the appointment of Diane Greene (founder of Vmware, Nicira, Bebop) to head Alphabet's Cloud Platform, Google has become an emerging play in the space. What we learned from Google's earnings call is that cloud—along with Google Play, the company's online music and video business, Chromecast hardware, YouTube Red subscriptions and other things—drove 39% growth in Google's "other revenue" to $2.543 billion from $1.75 billion for the same period last year.

Given the results stated above, it would seem that cloud will continue to have major growth for the foreseeable future - so why predict a decline of cloud? I'm not suggesting that cloud will not continue to grow, just that it is more "yesterday's news" than the "next great technology trend." - and the technology demands are predictive indicator of this - as well as history illustrating this tale.

Just like the recent Ghostbusters movie, Cloud Computing is a remake - not a sequel. We have seen this before - in the form of mainframe. Centralized computing, like mainframe and cloud, provides resources in a cost-efficient manner, allowing businesses of all sizes to utilize those resources in a timeshare model - MIPS for mainframe, consumption for cloud. It also allows for aggregation of data for additional data insights through a common analytics engine. As the "story" develops, innovation sparks the need for these platforms to respond faster, have the ability to "customize" based on specific user needs, allow for development of platform enhancements that can provide a competitive advantage. As those demands increase, the see-saw tilts the the other side -distributed computing. Couple the increase in the demand for faster, customized, configurations with the decrease in cost to "build your own", you had the rise of Client/Server - the "sequel" to Mainframe. And we already see it now with Cloud Computing.

So what is Cloud Computing's "Sequel"? Simply put - distributed computing, again. Albeit in a different form, but a sequel nonetheless. It's not called "client/server" but instead referred to as Edge Computing, being driven by two canonical trends - IOT and Analytics. And while the cloud served as a great initial platform to incubate much of the innovation in IOT and real-time analytics, it doesn't offer the Speed/Agility, Customization, and platform enhancements required to power the next evolution of technology. If you're still unsure, let's consider some of these indicators.

The formation of the Fog Computing Consortium - an architecture that uses one or more collaborative multitude of end-user clients or near-user edge devices to carry out a substantial amount of storage (rather than stored primarily in cloud data centers), communication (rather than routed over the internet backbone), control, configuration, measurement and management (rather than controlled primarily by network gateways such as those in the LTE core network).

Amazon debuts AWS Snowball Edge - You can deploy AWS Lambda code on Snowball Edge to perform tasks such as analyzing data streams or processing data locally. Data is collected, stored, and processed using the Snowball Edge independent from any other storage or compute resource.

Microsoft announces Azure Stack - Transform on-premises datacenter resources into cloud services for maximum agility. Run Azure IaaS services—including Virtual Machines, Blob/Table storage, and Docker-integrated Linux containers—for applications like SQL Server or SharePoint. Empower developers to write cloud-first applications using on-premises deployments of Azure PaaS services such as App Service. Make your application counterparts productive by enabling the same self-service experience as Azure.

SAP acquires Plat.One - Plat.One has an IoT platform that communicates with over 40 machine protocols. A micro-services-based adaptive architecture can run logic on a device, at the edge (such as a router or gateway) or in the cloud. What happens “when you want your city systems to keep operating and the connectivity required is not there?”

The indicators above illustrate the the power of the demand on agility vs. the aggregate power of the cloud. While the edge will never have the processing capabilities of the cloud, for many important reasons, the edge will provide more agility than the cloud ever could. As the edge becomes more capable - size and cost of computing power and storage capacity continues to decrease - the ability for the edge to process decisions quicker than transmitting and receiving from the cloud will spawn the next set of technology innovations.

Peter Levine, from the VC firm Andreessen Horowitz, explains it beautifully using the analogy of a fighter pilot. Colonel John Boyd developed a feedback framework called OODA - Observe, Orient, Decide, Act. The concept was simple - if you could take a pilot and shrink the time between each part of OODA, the pilot who completes the loop the fastest will ALWAYS WIN. Agility and speed are the most important factors.

This same agility is required on the edge. In order to achieve the required agility, more of the responsibilities that exist in the cloud will move to the edge. These responsibilities include Sense (collect data), Infer (determine relevance), and Act (decision making).

Sense - Sensors are becoming more ubiquitous and the size and amount of data is requiring that data to be processed locally vs. pushing all the data to the cloud - video data as an example. Real-time data processing of video feeds would become cumbersome if you had to send the data back to the cloud for processing and analytics. While data points and insights can (and should) be sent to the cloud for aggregate analytics, these feed are going to require local processing.

Infer - the 3 V's of data (volume, variety, veracity) will require local processing in order to extract relevance of the data as well as driving machine learning to improve accuracy and automation of the data insights that are extracted.

Act - devices will become more sophisticated as the technologies become cheaper, allowing for local processing to provide more agility than waiting for the cloud. Think about Siri - a cloud service that processes in the cloud - vs. a locally processing version of Siri. While there will be some functions that the cloud will continue to provide, some of the functions that can be processed locally will provide for a better user experience.

Autonomous cars are probably one of the easiest examples of the need to have localized processing but still take advantage of cloud services, self-driving cars need to make driving decisions instantaneously. They can't wait to send telemetry data to the cloud and wait for those commands to transmit back. They also can't afford any connectivity issues when decisions are made centrally in the cloud. The only way to solve these problems are to allow for decisions to the processed at the edge. Cars are becoming moving datacenters.

So why do we care? The biggest reason is because you will need to determine where to invest your capital and people resources. Don't concentrate your datacenter transformation on specific technology buzzwords - Docker, Chef, Puppet, Git, etc - but consider the key business drivers of your organization and then work to tie technology functions to executing against the business requirements. Does your company need a hybrid cloud strategy or does it need to consider a modernization of the applications used to run the business? Are these apps conducive to a web-scale model? Are there new functions and capabilities that you should develop in a cloud-native format utilizing edge platforms to improve customer engagement/experience? The answers to these questions will paint a clearer picture of which technologies to invest in.

Final note - Application Development will also change.

"If-then-else" logic is not enough to process the copious amounts of data and provide relevant insights. Algorithms and mathematics will become the next evolution of development, especially as machine-learning becomes a common as a math co-processor did to the Pentium chip. We'll need more STEM and less "script kiddies".