There was a time, 10 years ago, when Amazon and Google were neck and neck in the cloud infrastructure race. But then Google dithered - dropping the rumored Google Disk - while Amazon moved ahead.
Amazon has built a multibillion-dollar business in AWS, while Google is far behind. But the cloud is a rapidly evolving beast, and Amazon's advantages are about to be turned against them.
The evolution of new technology
New technologies go through predictable phases. The hype cycle is phase one. Cloud is well beyond that.
Phase two: we build what we already have with the new technology. So, cloud-based file storage.
Amazon has moved far beyond storage. They enable customers to build entire data centers in the cloud. That is their key strategic advantage.
Phase three is where life gets interesting: we build what we could not build before. More on that in a moment.
That's the build side. What about the use side?
Today, customers are happy building data centers in the cloud. They are looking for AWS to add more capabilities so they can run their legacy apps and get rid of their internal data centers altogether i.e. cloud admin will be a fast growing occupation; sys admin won't.
The next step
The cloud has upended the enterprise storage market, but that isn't its competitive advantage. Local scale-out storage can be competitive with cloud because network bandwidth isn't cheap.
What the cloud has that no enterprise-scale datacenter will ever have is the ability to spin up 10s of thousands CPUs - a virtual supercomputer - to run analytics against the data. CPUs are expensive - and they'll remain so as long as Intel can keep them that way.
The ready access to massive CPU cycles means that cloud will always be better at deep analytics, especially ad-hoc queries, than enterprise scale datacenters. But more importantly, cloud-based machine learning, neural networks and artificial intelligence are the next major evolution in how we use data.
And that's where Google has a huge lead over Amazon. Amazon's focus on building cloud-based datacenters makes them irresistable now, but the future of the cloud is with applications that can use thousands of cores to create value.
Look at what Google - and Microsoft - has done with machine translation. Yes, you need many petabytes of storage for the corpus, but the real key is in the compute resources and algorithms that make it work.
Or Google's new tool that can determine the rough location of photos using a neural network. They started with 126 million geotagged images to train and test the neural net.
And while their PlaNet is much better at guessing location than well-traveled human beings today, it should continue to improve as it is trained on more images.
The Storage Bits take
Henry Ford dominated the auto industry for decades because of the Model T's low price and manufacturing consistency. Then General Motors came along with a different model - multiple price bands - that appealed to a more affluent nation, and GM dominated for decades too.
Amazon's data center in the cloud model is powerful because that's what people know and its cost displacement model appeals to CFOs. But that's only for this generation of applications.
As more users discover the unique advantages of cloud-based infrastructure, the nature of leading applications will change. And those changes will favor Google and Microsoft, with their armies of PhDs pushing the limits of computer science, over Amazon's more pragmatic approach today.
Comments welcome, as always. Parts of this post first appeared in the Druva blog.