“What we’re really trying to do is to look at that end-to-end journey of data and to build really compelling, powerful capabilities and services at each stop in that data journey and then…knit all that together with strong concepts like governance,” Selipsky told Protocol in a recent interview in Boston.
AWS now has more than 200 services, and Selispky said it’s not done building.
“I don’t know when we ever will be,” he said. “We continue to both release new services because customers need them and they ask us for them and, at the same time, we’ve put tremendous effort into adding new capabilities inside of the existing services that we’ve already built. Both prongs of that are important.”
But cost-cutting is a reality for many customers given the worldwide economic turmoil, and AWS has seen an increase in customers looking to control their cloud spending.
“Some customers are doing some belt-tightening,” Selipsky said. “What we see a lot of is folks just being really focused on optimizing their resources, making sure that they’re shutting down resources which they’re not consuming. By the way, they should be doing that all the time. The motivation’s just a little bit higher in the current economic situation.”
This interview has been edited and condensed for clarity. Read Protocol’s other story based on our interview with Selipsky here.
Besides the sheer growth of AWS, what do you think has changed the most while you were at Tableau? Were you surprised by anything?
The number of customers who are now deeply deployed on AWS, deployed in the cloud, in a way that’s fundamental to their business and fundamental to their success surprised me. You can see it on paper and say, “Oh, the business has grown bigger, and that must mean there are more customers,” but the cloud and our relationship with these enterprises is now very much a C-suite agenda.
There was a time years ago where there were not that many enterprise CEOs who were well-versed in the cloud. Then you reached the stage where they knew they had to have a cloud strategy, and they were…asking their teams, their CIOs, “okay, do we have a cloud strategy?” Now, it’s actually something that they’re, in many cases, steeped in and involved in, and driving personally.
That’s just indicative of how much so many organizations are using the cloud now in every facet of their business — to run their core IT enterprise applications, of course, to do all sorts of new analytics, many of which involve machine learning now that there were never possible before, and then many, many end-customer applications as well.
It’s not just about deploying technology. The conversation that I most end up having with CEOs is about organizational transformation. It is about how they can put data at the center of their decision-making in a way that most organizations have never actually done in their history. And it’s about using the cloud to innovate more quickly and to drive speed into their organizations. Those are cultural characteristics, not technology characteristics, and those have organizational implications about how they organize and what teams they need to have. It turns out that while the technology is sophisticated, deploying the technology is arguably the lesser challenge compared with how do you mold and shape the organization to best take advantage of all the benefits that the cloud is providing.
How has your experience at Tableau affected AWS and how you think about putting your stamp on AWS?
I, personally, have just spent almost five years deeply immersed in the world of data and analytics and business intelligence, and hopefully I learned something during that time about those topics. I’m able to bring back a real insider’s view, if you will, about where that world is heading — data, analytics, databases, machine learning, and how all those things come together, and how you really need to view what’s happening with data as an end-to-end story. It’s not about having a point solution for a database or an analytic service, it’s really about understanding the flow of data from when it comes into your organization all the way through the other end, where people are collaborating and sharing and making decisions based on that data. AWS has tremendous resources devoted in all these areas.
Can you talk about the intersection of data and machine learning and how you see that playing out in the next couple of years?
What we’re seeing is three areas really coming together: You’ve got databases, analytics capabilities, and machine learning, and it’s sort of like a Venn diagram with a partial overlap of those three circles. There are areas of each which are arguably still independent from each other, but there’s a very large and a very powerful intersection of the three — to the point where we’ve actually organized inside of AWS around that and have a single leader for all of those areas to really help bring those together.
There’s so much data in the world, and the amount of it continues to explode. We were saying that five years ago, and it’s even more true today. The rate of growth is only accelerating. It’s a huge opportunity and a huge problem. A lot of people are drowning in their data and don’t know how to use it to make decisions. Other organizations have figured out how to use these very powerful technologies to really gain insights rapidly from their data.
What we’re really trying to do is to look at that end-to-end journey of data and to build really compelling, powerful capabilities and services at each stop in that data journey and then…knit all that together with strong concepts like governance. By putting good governance in place about who has access to what data and where you want to be careful within those guardrails that you set up, you can then set people free to be creative and to explore all the data that’s available to them.
AWS has more than 200 services now. Have you hit the peak for that or can you sustain that growth?
We’re not done building yet, and I don’t know when we ever will be. We continue to both release new services because customers need them and they ask us for them and, at the same time, we’ve put tremendous effort into adding new capabilities inside of the existing services that we’ve already built. Both prongs of that are important.
We don’t just build a service and move on. Inside of each of our services – you can pick any example – we’re just adding new capabilities all the time. One of our focuses now is to make sure that we’re really helping customers to connect and integrate between our different services. So those kinds of capabilities — both building new services, deepening our feature set within existing services, and integrating across our services – are all really important areas that we’ll continue to invest in.
Do customers still want those fundamental building blocks and to piece them together themselves, or do they just want AWS to take care of all that?
There’s no one-size-fits-all solution to what customers want. We absolutely have customers who very much want to have their hands “on the wheel,” if you will, and to be working with our services at the at the deepest layer, at the most primitive level — so EC2 for compute, S3 for storage, one or more of our database services — and they want to be interacting with those services directly.
It is interesting, and I will say somewhat surprising to me, how much basic capabilities, such as price performance of compute, are still absolutely vital to our customers. If you’d asked me 15 years ago, “hey in 2022, how much of the cutting edge of innovation do you think would be around raw performance or price performance of a unit of compute,” I wouldn’t have necessarily guessed that was still as important as it is. But it’s absolutely vital. Part of that is because of the size of datasets and because of the machine learning capabilities which are now being created. They require vast amounts of compute, but nobody will be able to do that compute unless we keep dramatically improving the price performance.
We (also) absolutely have more and more customers who want to interact with AWS at a higher level of abstraction…more at the application layer or broader solutions, and we’re putting a lot of energy, a lot of resources, into a number of higher-level solutions. One of the biggest of those … is Amazon Connect, which is our contact center solution. In minutes or hours or days, you can be up and running with a contact center in the cloud. At the beginning of the pandemic, Barclays … sent all their agents home. In something like 10 days, they got 6,000 agents up and running on Amazon Connect so they could continue servicing their end customers with customer service. We’ve built a lot of sophisticated capabilities that are machine learning-based inside of Connect. We can do call transcription, so that supervisors can help with training agents and services that extract meaning and themes out of those calls. We don’t talk about the primitive capabilities that power that, we just talk about the capabilities to transcribe calls and to extract meaning from the calls. It’s really important that we provide solutions for customers at all levels of the stack.
Given the economic challenges that customers are facing, how is AWS ensuring that enterprises are getting better returns on their cloud investments?
Now’s the time to lean into the cloud more than ever, precisely because of the uncertainty. We saw it during the pandemic in early 2020, and we’re seeing it again now, which is, the benefits of the cloud only magnify in times of uncertainty.
For example, the one thing which many companies do in challenging economic times is to cut capital expense. For most companies, the cloud represents operating expense, not capital expense. You’re not buying servers, you’re basically paying per unit of time or unit of storage. That provides tremendous flexibility for many companies who just don’t have the CapEx in their budgets to still be able to get important, innovation-driving projects done.
Another huge benefit of the cloud is the flexibility that it provides — the elasticity, the ability to dramatically raise or dramatically shrink the amount of resources that are consumed. In the first six months of the pandemic, Zoom’s demand went up about 300%, and they were able to seamlessly and gracefully fulfill that demand because they’re using AWS. You can only imagine if a company was in their own data centers, how hard that would have been to grow that quickly. The ability to dramatically grow or dramatically shrink your IT spend essentially is a unique feature of the cloud.
These kinds of challenging times are exactly when you want to prepare yourself to be the innovators … to reinvigorate and reinvest and drive growth forward again. We’ve seen so many customers who have prepared themselves, are using AWS, and then when a challenge hits, are actually able to accelerate because they’ve got competitors who are not as prepared, or there’s a new opportunity that they spot. We see a lot of customers actually leaning into their cloud journeys during these uncertain economic times.
During Amazon’s Oct. 27 earnings call, it was noted there was an uptick in AWS customers wanting to cut costs, and Amazon’s CFO said customers were looking to save versus their committed spend. Do you still push multi-year contracts, and when there’s times like this, do customers have the ability to renegotiate?
We’re an $82-billion-a-year company last quarter, growing 27% year over year, so we have, of course, every use case and customers in every situation that you could imagine. Many are rapidly accelerating their journey to the cloud. Some customers are doing some belt-tightening. What we see a lot of is folks just being really focused on optimizing their resources, making sure that they’re shutting down resources which they’re not consuming. By the way, they should be doing that all the time. The motivation’s just a little bit higher in the current economic situation. You do see some discretionary projects which are being not canceled, but pushed out.
But every customer is welcome to purely “pay by the drink” and to use our services completely on demand. Every customer is free to make that choice. But of course, many of our larger customers want to make longer-term commitments, want to have a deeper relationship with us, want the economics that come with that commitment. We’re signing more long-term commitments than ever these days.
AWS’ margins took a hit this past quarter, but do you think its margins in general are kind of fat?
We provide incredible value for our customers, which is what they care about. There have been analyst reports done showing that…for typical enterprise workloads that move over, customers save an average of 30% running those workloads in AWS compared to running them by themselves.
(Australian airline) Qantas, for example, is using AWS to do advanced analytics on flight paths — fuel-efficient flight paths, given wind conditions and what their flight paths should be — and they actually project they’re going to save $40 million a year, in addition to…lowering their carbon footprint through better fuel efficiency. That kind of analysis would not be feasible, you wouldn’t even be able to do that for most companies, on their own premises. So some of these workloads just become better, become very powerful cost-savings mechanisms, really only possible with advanced analytics that you can run in the cloud.
In other cases, just the fact that we have things like our Graviton processors and … run such large capabilities across multiple customers, our use of resources is so much more efficient than others. We are of significant enough scale that we, of course, have good purchasing economics of things like bandwidth and energy and so forth. So, in general, there’s significant cost savings by running on AWS, and that’s what our customers are focused on.
The margins of our business are going to … fluctuate up and down quarter to quarter. It will depend on what capital projects we’ve spent on that quarter. Obviously, energy prices are high at the moment, and so there are some quarters that are puts, other quarters there are takes.
The important thing for our customers is the value we provide them compared to what they’re used to. And those benefits have been dramatic for years, as evidenced by the customers’ adoption of AWS and the fact that we’re still growing at the rate we are given the size business that we are. That adoption speaks louder than any other voice.
Do you anticipate a higher percentage of customer workloads moving back on premises than you maybe would have three years ago?
Absolutely not. We’re a big enough business, if you asked me have you ever seen X, I could probably find one of anything, but the absolute dominant trend is customers dramatically accelerating their move to the cloud. Moving internal enterprise IT workloads like SAP to the cloud, that’s a big trend. Creating new analytics capabilities that many times didn’t even exist before and running those in the cloud. More startups than ever are building innovative new businesses in AWS. Our public-sector business continues to grow, serving both federal as well as state and local and educational institutions around the world. Only … in the vicinity of 10% of IT has moved to the cloud. It really is still day one. The opportunity is still very much in front of us, very much in front of our customers, and they continue to see that opportunity and to move rapidly to the cloud.
Do you ever see a cloud environment where customers could easily run say your machine learning services and Google’s data offerings and Microsoft’s X offerings as one big tech stack easily?
In general, when we look across our worldwide customer base, we see time after time that the most innovation and the most efficient cost structure happens when customers choose one provider, when they’re running predominantly on AWS. A lot of benefits of scale for our customers, including the expertise that they develop on learning one stack and really getting expert, rather than dividing up their expertise and having to go back to basics on the next parallel stack.
That being said, many customers are in a hybrid state, where they run IT in different environments. In some cases, that’s by choice; in other cases, it’s due to acquisitions, like buying companies and inherited technology. We understand and embrace the fact that it’s a messy world in IT, and that many of our customers for years are going to have some of their resources on premises, some on AWS. Some may have resources that run in other clouds. We want to make that entire hybrid environment as easy and as powerful for customers as possible, so we’ve actually invested and continue to invest very heavily in these hybrid capabilities.
For example, in the management capabilities, that’s the first thing that customers ask for: “We want to be able to see and have visibility into and, in some cases, manage resources on AWS, on my own premises and, in some cases, on other clouds.” So we’ve built capabilities, many of our management services, to see and, in some cases, control what’s going on across those environments.
A lot of customers are using containerized workloads now, and one of the big container technologies is Kubernetes. We have a managed Kubernetes service, Elastic Kubernetes Service, and we have a … distribution of Kubernetes (Amazon EKS Distro) that customers can take and run on their own premises and even use to boot up resources in another public cloud and have all that be done in a consistent fashion and be able to observe and manage across all those environments. So we’re very committed to providing hybrid capabilities, including running on premises, including running in other clouds, and making the world as easy and as cost-efficient as possible for customers.
Can you talk about why you bought Dilip Kumar, who was Amazon’s vice president of physical retail and tech, into AWS as vice president applications and how that will play out?
He’s a longtime, tenured Amazonian with many, many different roles – important roles – in the company over a many-year period. Dilip has come over to AWS to report directly to me, running an applications group. We do have more and more customers who want to interact with the cloud at a higher level – higher up the stack or more on the application layer.
We talked about Connect, our contact center solution, and we’ve also built services specifically for the healthcare industry like a data lake for healthcare records called (Amazon) HealthLake. We’ve built a lot of industrial services like IoT services for industrial settings, for example, to monitor industrial equipment to understand when it needs preventive maintenance. We have a lot of capabilities we’re building that are either for … horizontal use cases like (Amazon Connect) or industry verticals like automotive, healthcare, financial services. We see more and more demand for those, and Dilip has come in to really coalesce a lot of teams’ capabilities, who will be focusing on those (areas). You can expect to see us invest significantly in those areas and to come out with some really exciting innovations.
Would that include going into CRM or ERP or other higher-level, run-your-business applications?
I don’t think we have immediate plans in those particular areas, but as we’ve always said, we’re going to be completely guided by our customers, and we’ll go where our customers tell us it’s most important to go next. It’s always been our north star.
Correction: This story was updated Nov. 18, 2022, to correct the name of Amazon EKS Distro.