The next hills for Amazon Web Services: AI and the Internet of Things


When Amazon Web Services Chief Executive Andy Jassy showed off a raft of new products and services Wednesday, including a big semi truck that carries a device that can store 100 petabytes of data, it was hard not to think of it as a rather ungainly fix for the difficulty of moving large amounts of data.

Indeed, Jassy is clearly more excited about engineering the reverse course: moving Amazon’s cloud computing out to where all the data is.

Jassy explained his vision of AWS’s future during 90-minute conversation recently at his Seattle home. It’s a vision that resonates with some observers, who note that it may never make sense to move so much data, even in the age of the cloud. As Peter Burris, head of research at SiliconANGLE Media, put it, “you’re not going to move most data to the center. You’re going to move the cloud to the edge.”

In this final installment of a three-part series, Jassy talked about the new technologies that he thinks will drive new opportunities and challenges for AWS. The interview was edited for clarity.

Part 3:  The Internet of Things and beyond

Q: How do you see the opportunity in the Internet of Things, which inevitably involves cloud computing?

A: IoT is one of the fastest-growing and adopted types of new paradigm shifts that you’ve seen in a long time. And a lot of it is because the device manufacturers are highly incented to actually build in this fashion because it leads to additional capabilities in units. Most of the big [IoT] implementations are on top of us: Illumina’s genome-sequencing analysis, Philips’ smart-lighting project, John Deere’s 200,000 telematically enabled tractors that are sending information and sending analysis back on where farmers should plant. Or Major League Baseball’s Statcast.

Q: How is IoT different from other tech markets?

A: Some of it has to do with what’s happening with everything going digital. But I think that for many, many years people have had ideas or assets that were technically part of their on-premises footprint, but it didn’t sit in a data center. And if you think about these devices, the ability to put a sensor in them so they can actually capture information, send it somewhere and do something back, is very appealing.

You want those devices to be more intelligent. Being able to marry a small sensor in those devices which doesn’t meaningfully change the cost structure or the performance of those assets but have them be connected to the cloud to supplement on the storage and the compute and the analytics side, and have it send information back to those devices — that’s wildly attractive to companies.

Democratizing innovation

Q: Who’s driving enterprises’ move to the public cloud — lines of business or executives?

A: It’s a mix. In the early days of the cloud, it was largely line-of-business leaders, architects, and engineers themselves that were so frustrated by how hand-tied they were with their current IT teams. They basically went and did it themselves and said, “Look what I just did.” And then they would build credibility inside the organization and ultimately have both more influence as well as get the rest of the organization to use it.

A lot more over the last two to three years, we find that the CIO and the CTO are often driving the organization, pursuing the cloud in a much broader way. Almost always, whether it starts bottom-up or top-down, at some point the CIO is going to want to be aligned with whatever infrastructure provider they’re using and have that relationship and understand how they’re going to actually operate inside the company with the cloud. And then very often that enables the rest of the organization to safely go pursue using the cloud in all those organizations.

Q: Cloud computing has helped democratize business, allowing people to innovate both internally and allowing more new companies to compete in the market with more established leaders. Has the cloud in fact become instrumental to compete in this new way?

A: Most CIOs and most enterprises badly want their organizations to be able to operate much more quickly and in a much more agile way than they have in the past. To survive competitively you are going to have to move and innovate much more quickly than ever before. You’re going to need access to those same capabilities as all those other companies that are moving quickly. And if you want to do that, you’re going to have to work with different suppliers and infrastructure providers than you have for the last number of years.

They’re going to have to put some guard rails and some controls around how they’re going to use the cloud. But the cloud is going to allow those teams to move at a much faster rate, and they’re going to have to use the technology of the largest cloud providers.

Building blocks vs. stacks

Q: Your traditional IT competitors emphasize integrated stacks, and although you offer one as well, a lot of companies view AWS as offering a lot of building blocks they have to put together themselves. What’s the best model?

A: We believe in both models. For companies that have software developers who want to invent new products and new capabilities, they want the building blocks. They want to take the nine different compute families we have and figure out for their application [what works best]. They want those building blocks so they can kind of build whatever application they see fit.

jassyreinvent16-2How would MLB Statcast get built on an integrated stack? They basically have missile-radar technology in all 32 ball parks. They want to take all that data and they want to actually stream that data in. They want to actually process that data and all the analytics, and then they’ve got to send it back close to real time so they can kind of show on the diagram. There’s no integrated stack that does that.  

Q: So in a software-driven world, an integrated stack does not work?

A: It satisfies a slice of the use cases. But you can’t think of the integrated-stack solution for thousands of ideas. And remember, a core tenet of what we tried to build with AWS with the building blocks we have is we wanted to enable a thousand flowers to bloom.

Now, where we see patterns that are repeatable and happening over and over again, those we will build. We built [the data processing and analysis service] Elastic MapReduce because we had a lot of customers who were rolling their own MapReduce on top of AWS, and they said: “Well, we can keep doing this, but you know there’s a lot of people who want to run Hadoop on top of EC2. Could you just make this easier?” Or a data warehouse is another good example, with [AWS’s] Redshift.

Taking on the IT giants

Q: As you build out more and more services, you’re starting to compete more directly for dollars spent on traditional databases and other mainstream IT. How are you trying to get past the classic lock-in of Oracle databases, for example?

A: It’s already happening. You can kind of see that born out in the results of various companies. And you can just talk to customers. This new generation of the cloud, in addition to not hoodwinking them with all the hand-waving before they decide what they’re going to buy, is not going to have the database lock-in they had before.

In the old days, you had to basically choose between the commercial-grade, old-guard databases or the open engines that you could use but were hard to tune and get the performance you wanted. Customers didn’t want to have to make that binary choice, and it’s why we built Aurora [AWS’s cloud database]. It is the fastest-growing service in our history of AWS. Since the beginning of this year, we’ve had over 13,000 databases move via our database-migration service.

Q: Why are you going after that market?

A: We wouldn’t go after it if it didn’t matter to customers. In the first seven or eight years of the cloud the database was the slowest part of the stack to get with the cloud. Most of the database providers didn’t want to change their licensing model and they didn’t want to change their cost model, and customers were very frustrated.

And so we really got driven to build our own database engine by what customers wanted, and I think the horse is out of the barn there. I think people are not going to tolerate the cost structure and the lock-in and the punitive licensing terms that they’ve been subjected to over the last couple decades. We have a lot more planned in the database space.

The next hill

Q: What is the next hill you have to go climb?

A: The next 10 years will be even more innovative and game-changing. The first 10 years, as you’ve heard, there’s been so much debate and so much conversation and so much energy wasted inside of companies arguing about whether they should use the cloud or continue to use on-premises infrastructure. The next generation, that’s not going to be a debate. They’re not going to waste any time or any energy on the undifferentiated heavy lifting of infrastructure. They are going to use the cloud and AWS right from the start.

And so they will spend all that energy on trying to figure out how to reinvent every single space that exists today, including new ones that haven’t been invented. You’re going to see the way that people are able to analyze their data and then take action on that data will be wildly different. Virtually every application will have intelligence in it.

Q: What couple of things are you most excited about?

A: I’m excited about compute becoming much smaller than it ever has before. Instances [of computing] will be around forever. I still think most enterprises are figuring out how to use real-time analytics, which completely changes the decisions you can make and the intelligence of your applications and the actions they can take.

I think almost all applications over time are going to have artificial intelligence. We’re going to build a lot of services in that area and provide a democratization of those capabilities that are going to change what people can do in their applications.

I remain very, very bullish in the IoT space. The edge of what people consider their computing realm is wildly going to change. And I think cancer is going to be solved by the cloud, by computer scientists doing research in the cloud.

Read Part 1 from Tuesday: Charting the cloud landscape and the emergence of the enterprise cloud.

Read Part 2 from Wednesday: How Amazon Web Services CEO Andy Jassy manages at cloud scale.

Photos by Robert Hof