As Amazon Web Services Inc. opens its ninth annual re:Invent conference this week online rather than at its longtime Las Vegas venue, it confronts a world that needs its services more than ever during the worsening COVID-19 pandemic.

Amazon.com Inc.’s cloud unit has helped a raft of companies during the pandemic, enabling Zoom Inc. to handle rocketing growth from millions of people working and studying from home, providing a cloud foundation for Moderna Inc. to develop its COVID-19 vaccine, and much more.

At the same time, it’s facing more credible rivals in Microsoft Corp. and Google LLC, as well as nagging followers such as Oracle Corp. and IBM Corp. That’s all the more the case as AWS Chief Executive Andy Jassy (pictured) increasingly targets companies that want to reinvent their data centers and network edge locations so they’re more cloudlike — that is, responsive to rapid change and easier and less expensive to run than managing their own servers and software.

To learn more about AWS’ strategy and figure out what’s coming next for the company and the industry, I joined Jassy recently via Amazon’s Chime videoconferencing service for an exclusive, in-depth conversation ahead of the virtual conference that runs through Dec. 18.

AWS grew into a $46 billion annual run-rate behemoth by staking out an early lead among cloud-native startups, some of them now-huge companies such as Lyft Inc. and Airbnb Inc. But the next big market is pretty much all of information technology residing in on-premises corporate data centers, which still account for 96% of nearly $4 trillion in annual IT spending.

As Jassy put it, AWS is looking way beyond the current cloud computing market for its next stage of growth. “The market segment that we really address is the global IT market,” he said. “Mainstream enterprises and governments are moving to the cloud. We have a lot more growth in front of us.”

In this first of a four-part series, Jassy hints at some of the themes and announcements he’ll cover in his Tuesday morning keynote, including special focuses on edge computing, databases, and artificial intelligence and machine learning. The interview is lightly edited for clarity.

Look for more strategic and competitive insights from Jassy in my summary and analysis of the interview, as well as in three more parts in coming days and weeks. And check out full re:Invent coverage by SiliconANGLE, its market research sister company Wikibon and its livestreaming studio theCUBE, now in its eighth year covering re:Invent.

Beyond the cloud

Q: How do you view the cloud computing market today, especially with the stronger competition coming on?

A: Sometimes people get confused by those year-over-year percentages because they only really matter relative to the base revenue number. If you have a much smaller base revenue number, you’re actually growing in absolute dollars by a lot less. So if you look at where we are to be growing 29% year-over-year to a $46 billion revenue run rate, that means you were at a $36 billion revenue run rate. And that 29% year over year growth is an incremental $10 billion of trailing-12-months revenue. That’s a lot more than you’ll find from anybody else in the space by a pretty large amount.

We keep reminding ourselves that it’s still the relatively early days of mainstream enterprises and governments moving to the cloud. And we have a very significant market segment share in infrastructure cloud computing, what they call [infrastructure as a service and platform as a service]. I think Gartner’s latest estimate is we have 45% market share, which is more than double the second-biggest player.

But the market segment that we really address is the global IT market segment. If you look at that, still 96% of the total global IT spend is on-premises. The lion’s share of growth is ahead of us, as we’re kind of in this moment where mainstream enterprise and governments are moving to the cloud. We have a very broad offering. We have lots of things coming, both in the next few weeks, as well as in the next few years.

Q: Years ago you said everyone’s going to move to the public cloud. But the market’s shifting. If cloud’s growing and it’s still massive, then you still got this massive other market, it’s not about you guys pivoting. It’s just more customers want you to do more, right?

A: We’ve thought a lot about it as a team over the last 10 years. And I’ve thought a little bit about it as I get ready to do a keynote at re:Invent. We made a very strong statement, which we believed then, and I would say I believe even more strongly now: In the fullness of time, the vast majority of companies are not going to have on-premises data centers. And those that do will have much smaller footprints than they have today, even for things that just have to live there because they have to be close to something, like a factory, or pieces of mainframes that they couldn’t tease away. But their footprints will be much smaller.

It means that the overwhelming majority of computing is moving to the cloud. We always said we didn’t know how long the fullness of time would be. And I think people extrapolated that to mean that we’ve said everything will be in the cloud, and that’s not what we said. In fact, if you look for 10 years, we’ve been building what we think are sensible bridges to on-premises footprints, because we always knew it would take a long amount of time for that transition to happen.

And it’s why we built Virtual Private Cloud. It’s why we built Direct Connect, why we built things like Storage Gateway. I think people assume that the definition of hybrid is something that’s set. Because it was set many years ago. But the definition of hybrid is moving pretty quickly, evolving and being reinvented as we speak. Because when the vendors and the people that put together the term hybrid, they really envisioned it as cloud, alongside on-premises data centers. And that’s because a lot of those people who were pushing hybrid were on-premises data centers infrastructure technology providers, and they were trying to sell. And also at that time, most of the computing was in the cloud and on-premises data centers. But if you look at what hybrid is today, I think you’ve got to challenge that definition.

Amazon on the premises

Q: What do you think hybrid computing should mean?

A: Hybrid is the cloud alongside any edge node that you can run computing on-premises. And what’s on-premises? Is on-premises data centers? Is it stores or restaurants or hospitals? Is it at the battlefield edge? Is it down in the agricultural fields? Is it at the 5G edge? Is it in metropolitan cities where nobody in their right mind will have giant data center footprints because it’s more expensive, but people need certain number of workloads to work there?

We think that hybrid is the combination of cloud alongside these edge on-premises nodes that could be on-premises data centers, but it could also be all those other things I mentioned. And the way people want hybrid delivered is they want us to distribute AWS to those various edge nodes where they can use it in a way where it’s the same APIs, the same control plane, the same tools, the same hardware, and use it consistently.

The notion of there being on-premises presences alongside AWS and the cloud has never changed. We always believed that was going to be true. But I think the number of edge on-premises nodes has changed a lot. And I think the way we’ve thought about it as having all of these kinds of edge presences, of which on-premises data centers is just one, has also evolved. And you see that in how in the products we have and some of the things you’ll continue to see us deliver.

Taking the cloud to the network edge

Q: Related to hybrid cloud, you’ve made Outposts, the managed service to extend AWS infrastructure and services to data centers, available in most areas of the world and I suspect we’ll hear more on it at re:Invent. What’s your vision of the challenges and opportunities at the edge, and how does it differ from those of competitors such as Microsoft, Oracle and Google offering similar services?

A: We announced Outposts at the 2018 re:Invent, and then we launched it at the 2019 re:Invent, and we’re very excited at how quickly Outpost is being adopted and how centrally it’s part of people’s plans to move to the cloud. And we had a really different approach with Outposts than most of the other players took with hybrid. I think most of the other players took these approaches like they’re going to ship this software, but the software has different APIs and different control planes, different tools and different hardware, and you can buy custom hardware to do it. And it was trying to find this bridge between two very different things, which are on-premises and the cloud, and they were really clunky bridges. It didn’t work, and so they got no traction, so we never liked that model.

It’s part of why it took us time to define a solution there, because we saw all the pitfalls. You can never keep the on-premises version as up-to-date as the cloud. If you look at just the features that we launch in a year, it’s a couple or a few thousand, but that’s a fraction of the total number of deployments we make that aren’t customer-facing features every year in AWS. So you’re never going to keep them consistent unless you actually start from the get-go thinking that we are going to be consistent, that what we’re going to offer, instead of having this clunky bridge, we’re going to take AWS and we’re going to distribute it on-premises.

And we may not have all the services we have in AWS, and we’ll build a seamless bridge to allow you to leverage those services when you need it, but what we ship on-premises that we distribute to allow you to compute in storage and database and analytics, machine learning, is going to have the same API, the same control plane, the same tool set and the same hardware so that you can use it consistently. That vision has really resonated with customers.

Q: But presumably people ultimately want all the AWS services they get in the cloud on-premises, if they can.

A: Not surprisingly, whenever you have something that’s growing and people like, they have lots of suggestions, which is awesome and very helpful, and we have customers who’d like to see us find ways to enable them in other environments than just their on-premises data centers. If you look at all the places people want to run all of those edge nodes I was talking about, they’re everything from some kind of asset on an oil field to some kind of 5G location, to something in a metropolitan area that they don’t really have capacity to an office building or a restaurant, to their on-premises data centers, to the battlefield.

They want offerings in all of those places, and for several of those, Outposts will be appropriate, and for others, you’ll need different solutions. Snowball Edge are appliances that allow you to store data and run some computer on them, but where you can do it, where you’re disconnected and it’s a ruggedized terrain, you don’t have to be connected, that requires a different solution. So you’ve got to be careful about not having one tool to rule the world and use one offering for all those places.

Multicloud myths

Q: You have been strongly against the notion of multicloud for awhile now. But lots of companies want to do just that, however they define it. What’s your view of multicloud today?

A: Most enterprises start off thinking, OK, I’m going to try to figure out how to be multicloud and split my workloads rather evenly across a couple providers. When you look at the deals that get done, and we see a lot of them obviously, the vast majority of companies don’t split those workloads rather evenly. They tend to predominantly choose a cloud provider, and then if they want to make sure that they can run on something else or they have certain teams that really want to run something else, or they want to make sure they have leverage if something goes sideways, they’ll run a certain amount of workloads on another cloud provider, but it’s rarely 50/50 — very rarely. It’s almost always like 70/30, 80/20, 90/10 when they’re not going all-in somewhere.

They do that for a few reasons. First, not all the cloud providers have the same capabilities and you’ve got to decide how much you want to optimize for a rainy day versus enabling your builders to build anything they can imagine.

The second thing is that it’s hard to manage all of these disparate environments. It’s a big move to go from on-premises to the cloud, but then ask your builders to make that move, plus being fluent in multiple platforms is tough, and it leads to lower productivity. It makes development teams unhappy and you get less buying leverage because you’re splitting your volume across multiple providers. Everybody will do it a little bit differently and I think over time, you’ll see it continue to evolve.

Photo: Robert Hof/SiliconANGLE