On the surface, the timing of Amazon Web Services Inc.’s annual re:Invent conference in Las Vegas next week doesn’t seem ideal — but AWS Chief Executive Adam Selipsky doesn’t seem too worried.
The event is always a relentlessly upbeat gathering, but this year Amazon.com Inc.’s cloud computing unit is facing, well, a lot of storm clouds. Although some companies might envy the 27% revenue growth AWS reported in the third quarter, it’s the slowest growth rate for the unit since it began reporting results separately in 2014.
Selipsky (pictured), however, says AWS is steaming full speed ahead, and it’s set to introduce its usual raft of new products and services at re:Invent, which runs Nov. 27-Dec. 2. In fact, he thinks the economic downturn will make moving to the cloud even more important for companies looking for more efficiency and agility.
In particular, he thinks the ability to store, manipulate and analyze data — something he steeped himself in during a stint as CEO of Tableau Software Inc., now owned by Salesforce Inc. — will become even more important. That’s likely to be a big focus at re:Invent, as well as AWS’ recent push into more industry-specific services beyond basic compute, storage, data analysis, and artificial intelligence and machine learning services.
I joined Selipsky recently at the 37-story re:Invent building in Seattle for a conversation ahead of re:Invent, where SiliconANGLE and its video studio theCUBE will be providing wall-to-cloud coverage of the event.
We talked about how much cloud keeps changing, its transformation of the cultures of companies that adopt it, the continuing march of innovation in compute and other infrastructure, the increasing importance of higher-level services atop AWS, and, not surprisingly, the central role of data for cloud computing. The interview has been lightly edited for clarity.
Q: It feels like we’re at a tipping point of another wave of cloud from customers such as Snowflake, Goldman Sachs, Capital One — more horsepower, more higher-level services, rendering itself into a new kind of cloud, a next-generation cloud. What are you seeing customers doing differently now?
A: How customers have used the cloud has dynamically and rapidly morphed since day one. When I came back a year and a half ago, it had changed a lot. And I think it still continues to evolve. We continue to see just deeper and broader adoption inside of individual customers. And I think it’s not so much a single tipping point that you see within customers, but just in an inexorable move. And a lot of customers go, “Hey, let’s accelerate this move.” Or they’re just moving hundreds and in many cases now thousands of applications onto the cloud and they’re getting more sophisticated about the way that they move those applications and setting up landing zones, which work really well technically.
And then some of our more technically sophisticated customers are saying, “Hey, we have capabilities which we could externalize.” And so you alluded to Goldman Sachs as a financial cloud for analytics. So Goldman on top of AWS is essentially trying to become an ISV, a SaaS offering on top of us. This has been one of the most noticeable things to me in the year and a half I’ve been back, is just how deeply AWS is now deployed at so many really big companies and how the conversation has really become a C-suite conversation.
Q: Is that your big takeaway that in talking to customers this past years is that there’s so much more deeper in with it? Or are they more replatforming and refactoring?
A: I would say the biggest thing I’ve seen in the past year and a half is that as a result of how deep and broad they’re going with the cloud, what’s really happening and the conversation that most often CEOs and CIOs want to have with me is around transformation. It’s not around technology. So a lot of CEOs and CIOs are now proactively bringing up, when I meet with them, “I can see the cloud is transforming how my organization functions, not how my technology works. So talk to me about that. What have you, AWS, seen? How can you help me think about and shape that organizational transformation?”
That’s a very big claim. So why is that? I think that when our customers really deeply adopt cloud, what happens is that they get more agile, they get faster. They do more things more quickly. So that encourages innovation.
The second thing is that there are very few penalties for failure. So you boot things up, they don’t work, you spin them down. And so you quickly get into this model of it’s OK to experiment. And you just start to think of failure as part of innovation, which Amazon has always done. But a lot of companies spending a lot of capex on a lot of big projects don’t tend to. And then also of course, with all the different capabilities we have, particularly around data and really taking an end-to-end view of helping you manage your data, we see so many more organizations now moving toward actually putting data at the center of their decision-making as companies. Those phenomena I just went through actually at the end of the day cause cultural change inside of organizations.
Q: You’re talking about digital transformation.
A: It’s the digital transformation by moving to the cloud that’s driving organizational transformation and people are saying, “Hey, how should I organize my company? And I can see my culture changing, how do I proactively shape that?” But the transformation potential of the cloud is to make organizations faster, more risk-taking, more innovative. And to drive far better and far quicker decision-making.
Q: AWS is now an economic force, and this is the first recession where Amazon is an economic influencer and could change how recessions look. The impact of what you guys do could accelerate out of the recession. So if I’m a business owner, I can get faster time to value. So this is kind of a new variable. It’s never been seen before in economics.
A: We’re absolutely seeing customers using AWS to both weather uncertain economic times much better and also to accelerate out of them much faster and in much better shape than they were before. A great example is Airbnb. Airbnb was already a significant AWS customer in 2019. And then of course the pandemic hit in early 2020. The bottom fell out of the hospitality industry. And Airbnb had to go into very significant cost-cutting mode. And they were able, with incredibly low lead time, to cut their cloud spend by 27% with almost no lead time. Huge. I mean, imagine you’re running data centers and there’s no metrics to cut your expense.
And that was a very significant part of their overall cost reduction efforts. At the same time, they were still able to be far more innovative than other companies and continue to experiment, continue to roll out new functionality for their customers, because you can do so much more with so much less in the cloud.
Coming out of the worst part of the pandemic, of course travel has since exploded. The hospitality industry is back to very high demand. And Airbnb was able to really accelerate out of that, turn on vast new amounts of cloud resources in a way that again, you couldn’t if you weren’t already deeply steeped, deeply expert in the cloud. They were able to accelerate past a lot of competitors who had not prepared themselves, had not become expert in how to use the cloud, how to innovate in the cloud, how to save costs in the cloud, how to rapidly expand resources when the situation called for it.
What a lot of our companies understand is that now is exactly the time to lean into the cloud, not despite economic uncertainty but literally because of it. It is the most compelling time to lean into the cloud. If you think just about pure cost savings, you’ve got all sorts of companies who have saved huge amounts of money by moving to the cloud and are able to do so very quickly.
Dealing with the data deluge
Q: If you take digital transformation to completion, companies have been completely transformed with technology, so IT isn’t a department but is the business. What does that look like? In particular, how do you see data emerging as the lifeblood of applications and application development?
A: I do believe that data, and IT in general, is moving from departments or topics into becoming just intrinsic parts of how businesses function. And so that they’re moving into lines of business, into applications, intrinsically into the things that companies deliver to their end users.
Now, that doesn’t mean the companies will not want to have a centralized data warehouse or experts who are data stewards and in charge of companies data strategy. But I don’t think it’s an either/or. It’s an and. Companies will continue to stand up very robust chief digital officer organizations, who will drive a mission-critical part of the company strategy around data.
In addition to that, data expertise and the use of data will permeate all parts of a company. And, in time, we’ll stop talking about people who have the word data analyst in their title. Rather, we’ll have hundreds of millions of people who analyze data as part of their day-to-day jobs, most of whom will not have the word analyst anywhere in their job title. We’re talking about graphic designers and pizza shop owners and product managers and data scientists as well.
Q: This whole generative AI category that the VCs are funding, it’s based on data.
Q: So this is the natural outcome you’re saying from a data perspective, it’s embedded in everyone’s job.
A: Yes. The data is just becoming intrinsic to so many business functions, so many jobs that didn’t used to believe that data was central to what they did for a living. Now, this is not just happening by itself. There are a lot of enablers. At a high level, there’s a couple things happening — one of which is, well, first there’s this external condition of data’s never been created faster in the world.
Q: Because the tsunami’s coming.
A: The amount of data, everybody keeps saying “Look how much data has been created.” And yet every single year, if you look forward, you make the same statement of, “We’re just getting started,” and the amount of data in the world is only accelerating. That is both a problem and an opportunity for companies. It’s a problem because there’s the expression, the best place to hide is in plain sight. And so for companies who don’t know how to deal with all of this data, they actually are unable to generate insight from their data because there’s just too much of it. But there are also huge opportunities to transform in entire businesses and industries with this data.
Q: What are the enablers AWS is focused on?
A: One is really having a broad end-to-end set of tools and technologies that work really well together to help you manage your data. You sometimes see folks talking about point solutions: “We have this solution for this part of what you do with your data or this other part of what you do with your data.” But that’s not really helpful to customers.
Customers really need to be able to take an end-to-end integrated view of the entire journey that their data goes on from ingestion to storage, to harmonizing the data, to being able to query it, to being able to do business intelligence and human based analysis on it, to being able to collaborate and share on that data. And we’ve been putting together a broad suite of tools from databases to analytics to business intelligence to help customers with that.
And then we’re more and more integrating across those different services so that it really works as part of a collective, as part of a whole.
Q: Like a data cloud kind of model for the customer.
A: It’s not a data cloud. It’s really, I think we’ve got one cloud which has got a set of integrated capabilities. So for example, you want to be able to move data between different services when you want to. If you’ve got data in one place and you want to analyze it using another capability, should you have to move it every time you want to analyze that data? It would be much better if you could leave that data in place, avoid all of the ETL, which is a very nasty three-letter word.
Q: And moving the data costs.
A: Right. Avoid the muck of ETL as well as the cost and the complexity and just query that data in place. And so more and more we’re trying to build capabilities where you can query data in place, whereas you do have to move it, you move it seamlessly with a push of a button essentially with zero ETL involved. That’s not a separate cloud, it’s really just a series of broad but integrated capabilities to help you manage the end-to-end life cycle of your data.
A second phenomenon is infusing machine learning into all of those capabilities. So we’ve invested very significantly in building machine learning capabilities. Things like our SageMaker platform, which is the most widely adopted machine learning platform.
Q: Swami [Sivasubramanian, AWS’ vice president of database, analytics and machine learning] is probably going to have some announcements on that.
A: Yes. I might have a couple as well. We have too many things to announce on one keynote, John. I love sharing the love, but it’s also a necessity.
Q: The conversation we’ve been having with customers in the marketplace and reporting on is it used to be how to store, you store it somewhere. So then it turned into sharing data and then it’s building apps with data. We call it data as code. So in other words, the progression is store, share and now more developer-focused. The problem is how do you govern it and how do you secure it? What are you guys doing with customers there?
A: We’ll probably have some news around some of these areas, around many different data topics.
I think data governance is a huge issue. Really what customers need is to find the right balance for their organization between access to data and control. And if you provide too much access, then you’re nervous your data’s going to end up in places that it shouldn’t be, viewed by people who shouldn’t be viewing it. And you feel like you lack security around that data. And by the way, what happens then is people overreact and lock it down so that almost nobody can see it.
Q: And then it’s not available to machine learning.
A: Or for the intrepid product manager who’s creative and actually has an idea to look at fields from two different data sets and draw some new conclusion that nobody told them to go do.
Q: And that’s the downside of not doing it right.
A: So what if you need is governance solutions, which actually provide really flexible, really powerful access controls that are configurable by the customer. You need to provide great visibility and discovery of that data. Things like metadata exposure, business catalog so that people can find and discover and share and collaborate on data. And if you have those capabilities, then you put the appropriate guardrails in place, which will be different for different organizations and different for different sets of data. And within those guardrails, you can empower people, you can set them free to explore their data and to derive new insights. So we’ve already got a good number of governance capabilities and we’re going to continue to release more really powerful governance capabilities, more and more that work across multiple AWS services to enable companies to find that balance between access and control that enable them to set their data free.
Q: We see a lot of conversation around clean rooms and sharing data’s been a big discussion.
A: Clean room is a really, really interesting area. And I think there’s a lot of different industries in which clean rooms are applicable, ranging from pharmaceuticals to advertising. So yeah, I think that clean rooms are an interesting way of enabling multiple parties to share and collaborate on the data while completely respecting each party’s rights and their privacy mandate.
Q: How about price-performance? Jeff Bezos used have that line, “No one wants to have their packages delivered slower.” I’ve never met a developer that said I want to run my app and workload on the slower infrastructure. So this has become a huge topic on workloads. Performance of the infrastructure is huge, I know [AWS General Manager of EC2] Peter DeSantis is going to have something Monday night.The performance, silicon advancements, chips, instances, what’s coming down the pipe there?
A: If you’d asked me in 2006 when we launched, “Hey, in 16 years, do you think that the first words out of our customer’s lips will still be price-performance of raw compute?” I would not have guessed the answer was yes. I would’ve thought that probably would’ve been on to other topics and I would’ve been wrong. Price-performance for compute is still absolutely a central enabler of all cloud workloads. And it’s because data sets are now so large and continuing to expand so quickly.
And there are workloads like machine learning-based workloads, which are incredibly compute-intensive. So for example, we’ve got this segment called LLM, Large Language Models. And these require truly massive amounts of processing on truly massive data sets. And the potential to then apply those to machine learning models, to all sorts of different business situations is enormous. But nobody will be able to do it if the price-performance isn’t there. And so we started investing in custom silicon back in about 2012. Yeah, well even slightly before that with our Nitro platform, which we started working on probably close to 10 years.
Q: Which everyone’s trying to copy, by the way.
A: Other folks are trying to move on custom silicon, but they’re just years behind us. So if you fast-forward, we’re now in a third generation of Graviton chips. We’ve also got, not one, but two specialized chips for machine learning, Trainium for training models and Inferentia for running inferences. We’ve got Nitro, which is important not only for performance but also for security by offloading a lot of work from the EC2 instances itself onto a separate chip. And so we’re multiple generations in. We’ve got absolutely leading price-performance. Importantly, our Graviton-based instances consume 60% less energy than comparable EC2 instances which are not Graviton-based.
Q: Power is huge, prices going up.
A: And sustainability. I mean sustainability is huge, and it’s very important. And other cloud providers are really just struggling to get internal alpha versions of custom silicon apps. So we’re years ahead and we continue to invest very heavily in this area because we don’t see any sign of customers wanting us to slow down.
Q: On the sustainability and supply chain side, we’re hearing that people if they’re trying to do anything on-premise with hardware, they’re having a hard time getting equipment too.
A: Oh absolutely. We have many customers coming to us saying either we can’t obtain chips or we can’t obtain energy, power commitments from utilities. By the way, it sounds that these things are easy for us, but given our scaling and that this is our business, we’re able to provide a lot more continuity in the entire supply chain than almost any of our individual customers can themselves.
From primitives to higher-level services
Q: If you look at networking advances and serverless for instance, with massive innovations coming from the field, we’re seeing signs of just people recognizing new ways to do things with, say, serverless and some of these network compute advantages. So latency becomes a big issue and I think it’s one of the reasons why I think multicloud is a complete nonstarter.
Now you got the higher-level services. A couple years ago we asked [former AWS and now Amazon.com CEO] Andy Jassy, “If you could redo Amazon today back in whatever, what would you do?” He said, “I’d build on serverless.” And we laughed, but his point was the advancements now are so good you can actually refactor. So what’s your view on how you see customers handling all this new higher-level services?
A: Well, first I’ll just say that we’re not done with serverless. I sometimes jokingly use the term “internally serverful.” I haven’t found a better opposite of serverless. We’ll continue where possible to take server-based services and continue to build serverless offerings for those services. So you should expect us to continue down that path.
In terms of customers, I think we continue to see rapid adoption of containers, which is not quite the same as your serverless question per se, but you tend to see a high correlation. And there’ll be with both our ECS-, Elastic Container Service-based offerings as well as our Kubernetes-based, EKS-based offerings. There’s very rapid adoption of those types of services. And so I would continue to see really broad adoption of containers.
John: That’s what [Amazon Chief Technology Officer] Werner Vogels was getting into his talk about microservices and event-based capabilities. I think he was teeing that up.
A: This wave of serverless offerings that we’ve enabled — Lambda is another one, which is rapidly growing. Customers have gotten so sophisticated now with the use of the cloud, particularly the ones that have invested deeply and have been working with us for years, that they’re now thinking about event-based architectures and thinking about how to use services like Lambda. In which case you are truly only triggering an action in the cloud at the moment when it’s needed.
The other trend you’ll see is just continued use of specialized resources, really powerful, specialized resources for specialized workloads and specialized purposes. So that’s why we’re now up to over 575 different EC2 instance types. And it’s because you really want to optimize the price performance. If you want something for high performance computing, for example, which is compute-optimized, that’s different than if you want something which is compute-plus-network-optimized. And compute-optimized versus compute-plus-network-optimized will give you very different price-performance and very different latency for your application than just choosing one or the other.
Q: So the specialized services enable the customer more choice. How do you make it easier for the customer? Because we’re seeing skills gaps and I got to move faster and make money and be profitable for their business using technology and they want to transform their business. How do you see that becoming actually turnkey? Are there out-of-the-box templates? And how do you roll that out to customers?
A: I just wanted to be clear, we are going to continue full throttle on compute, on storage, on database, on enabling all those fundamental capabilities for customers. I see no horizon on the appetite from customers for those core, sometimes we call them primitive, capabilities. And I think some people think, oh, developers don’t want to develop at that level. Many do. And many need to.
And — we have many customers who are asking us for solutions higher up in the stack, abstractions on top of those basic services. Sometimes they’ll call them applications. And so we’ve already started to build applications for customers, and you should expect to see us to build a lot more applications in both vertical industries, as well as horizontal use cases for specific business functions.
One of the better-known applications that we’ve already delivered is Amazon Connect, which is our horizontal solution for call center management. And you can get a simplified call center up and running in AWS in minutes. We’ve had customers deploy a thousand agents on Connect literally overnight.
We’ve also built for vertical applications for vertical industries. So, services we’ve released like FleetWise, which is a fleet telematics and fleet management solution for auto manufacturers and fleet operators. We’ve done a fair amount in the healthcare space with things like our health data lake called HealthLake.
The cloud is moving into lines of business, and your average business user is going to want to consume the cloud in different ways than your bleeding edge developer or a centralized IT team. So more and more, we are going to build and enable customers with line-of-business solutions, you can call them applications, in different areas, either for specific industry verticals or horizontal business use cases. And we’ve stood up very significant teams internally with senior leaders.
Q: On the other hand, you’re still going to enable the ecosystem to deliver some horizontal and vertical applications.
A: Absolutely. The partner ecosystem has been a fundamental plank of the AWS strategies, literally since day one. We now have over 100,000 partners around the world in every conceivable country addressing every conceivable use case, and we continue to put a ton of resources into helping to build and make vibrant the partner ecosystem.
The security challenge
Q: Security has become top of mind, so what’s AWS doing to improve that situation?
A: Security’s always been job zero at AWS, and I think it’s only become more important for our customers. We’re really thinking about security in different layers. First, customers get, without doing anything, the intrinsic security of AWS, the intrinsic security of the cloud. We build security in at every layer from the physical data centers through our fundamental services like EC2 and S3, and we talked about Nitro and security benefits of Nitro, and then all the different security capabilities which we put in place around that. And we have many thousands of security professionals whose only job is security. And customers get all of that without ever thinking about it or having to know about it or care about it.
Then on top of that, increasingly we built out a really powerful set of externalized services and capabilities that the customers choose to consume, which help them with security at the application layer, help them to ensure security of the workloads that customers are running on AWS. So, we built services over time, like Detective and Macie, and then GuardDuty, which is very popular, essentially a threat detection service. And we continue to build out tools like WAF, which the web application firewall, tools which help customers prevent and mitigate DDoS attacks, things of that nature.
More and more customers are saying, “Hey, there’s all sorts of bold things we want to do out there. We want to go and explore all the innovation that we can dream of, but we need to be confident when we do it. We need to know we’re going to be secure as we venture out.” And I think that’s our attitude in building this layer above the baseline security of AWS, of building these security capabilities that customers can layer in.
Q: What are the challenges in helping customers improve their security?
A: One big challenge is around data and security and being able to bring together all of your data around security. And so more and more, we’re trying to integrate these capabilities and make sure that our customers can view all the data which is relevant to security and to be able to glean insight from that and take action. Basically, data is incredibly important to security as well. We actually view it as not only a requirement for any organization, but also as an enabler of the innovation and exploration that our customers want to do.
Q: What kind of enablement are you guys looking at there for the customer?
A : One area is goes back to compute. We’ve released interesting capabilities around containers, but customers continue to ask us for more security capabilities around containers, and so there’s different facets of that that we’ll be working on. Another is we’ll do things to just continue to help bring data together and help customers shine a light, be able to take action on data which is relevant to security. Many aspects of security, which are fundamentally a data issues.
Q: How’s the partner ecosystem developing?
A: We continue to invest deeply in the partner ecosystem. It remains as important as ever, whether it’s ISVs, systems integrators, resellers, VARS, distributors. We invest in a lot of dollars and a lot of human talent in making sure we have the best partner capabilities in the world. And at the end of the day, it remains all about our customers. We and our partners both exist to serve our customers, make sure we’re adding value to our customers. And we always take the attitude with our partners of “Make sure you’re adding tremendous value to our joint end customers, and we will all be very successful.”
Q: If you had to put a bumper sticker on re:Invent, what would be the bumper sticker on the thematic big news coming out?
A: There’s too much going on to encapsulate it in bumper stickers. But a few of them are the end-to-end data story, security is an enabler of innovation, specialized purpose-built solutions for industry verticals and horizontal business areas. And one more: performance. Powerful, powerful capabilities around performance.