Amazon chief security officer on cyber outlook for second half of 2024, 'platformization' and chips

Welcome to cnbc.com. I'm Frank Holland, anchor of Worldwide Exchange. Today, Amazon is holding its AWS reinforced conference, where one of the main topics is creating a strong security culture. And I'm joined now by Steve Schmidt, the Chief Security Officer of Amazon, overseeing AWS, the retail business and the digital ad business. Steve, great to have you here. Great to be here, Frank. Thanks. So I want to pick up where your keynote left off. Hit two really big points in that keynote. First, the challenge and finding talent that understands security and also understands AI and also the fact that the cybersecurity landscape is constantly evolving. Can you give us a snapshot of the landscape right now for cybersecurity and also your outlook for the second-half of the year? Certainly, so AI is an area that's pretty new for most companies. So people are still learning how to secure it appropriately. And one of the things that we discovered in building AI applications across Amazon for quite a while is that we have to build ways for our developers to do things safely, give them good guardrails, help them ensure that they're doing things to our standard, and most importantly, build the tools that continuously test services and features. All right, So with your services, specifically AWS, you have obviously a lot of customers on there. You also have your bedrock service that gives access to basically AI through that. What are the challenges that you see on the horizon as we go throughout the rest of the year? Of course, hackers, and we'll talk more about hack attacks and things like that coming up, but hackers are always looking to exploit vulnerabilities or any opportunities they have. So what are the challenges that you're seeing? Where do you see the threats coming from? So the threats are the same threats we've dealt with before. What's new is their techniques. So in the traditional world, we had a pretty good understanding of what the vulnerabilities and software were. The AI world is different. We haven't had a lot of experience as an industry as a whole, and we had to develop new tools and techniques. So one of the things that we focused a lot on was taking the learnings that we have from our many years of building AI apps inside Amazon and giving them to our customers through our services like Amazon Bedrock and Amazon Sagemaker. It's all about finding ways to detect what the adversaries are doing before they have an opportunity to actually exploit things. So I mean, just a kind of a bottom line question here for you. Our networks are, is cloud infrastructure including your your own, is it more susceptible to a cyber attack now with AI or is it better protected with AI? It's actually in a better place. If you look at traditional infrastructure, it relied on things like firewalls and perimeters as a defence mechanism, which are just totally ineffective in a modern day. So what we have to do now is focus on how do we protect the data appropriately, the data that we use, that we consume, that we train models with, and that our applications give back to our customers. So it's building the new tools, deploying them and monitoring to make sure that we're focused on do we have the right controls around data itself. Alright, something I want to talk to you about. It's been a big thing for cybersecurity all year. I'm sure you've spoke quite a bit about it. I spoke to people about it at RSERSA. Excuse me, we're talking about platformisation. So I want to ask you very directly, how does Amazon view platformisation? Is it possible to have one company meet all your needs? I know currently use at least two vendors that implement platformisation, Palo Alto Networks and Crowd Strike. So when we look at things at Amazon, we want to give customers choice and that's always been something that we've done. So if you look at the way we give people access to a bunch of different generative AI foundation models as an example, it's give them an opportunity to choose the options that are best for their business. You mentioned Crowd Strike in Palo Alto. Both of those are excellent partners of AWS and they give our customers a choice when they want to do their own endpoint detection response or XDR or other things in the security space. But I mean, as, as I'm sure you know, each of these companies, they're trying to at least present themselves as a not A1 size fits all, but somebody that can meet all the needs of a particular company. Is it possible for one company to have a platform that can really cover everything that AWS needs done and your customers need done? I know you said it's about giving choice, but is that even possible? I think it depends a lot on what the business is and what the individual implementations are. That's why if if people have marketing language and say we do everything, I always look a little bit sceptically at it. Instead, what I focus on is my specific workloads, my specific use cases. Does a vendor have the tools, techniques, capabilities to address what I need as a customer? All right. So I mean, it's, it's becoming more and more just granular when it comes to cybersecurity, especially with AI. If you wouldn't mind, can we talk about some of the other vendors you're using and, and other particular use cases for those other vendors? Palo Alto and Crowdstrike, obviously two of the big giants in the cyberspace. There are quite a few others. We particularly like our partnership with Wiz. They're doing a great job in the vulnerability management space. They do a wonderful job in understanding how cloud infrastructure is actually configured and used. You see that in a bunch of other places as well. OK, Are there any other vendors that you wouldn't mind sharing in the use cases? I think that we should just stick with those for now. OK, fair enough. I want to get to your cloud business here. You're the global leader in in cloud infrastructure. Most estimates have you somewhere around 47% market share, give or take. It's a huge installed base. And now you have rising demand for AI and of course, you need AI chips. How important are AI chips, specifically from NVIDIA to protect AI workloads and then what percentage of chips are NVIDIA compared to the in house chips that you have for your servers? One of the things we recognise many, many years ago that we needed to build our own chips in order to make sure that we gave customers the options that they need in order to train and use and generate VI applications effectively. It takes a long time, as you know, to to build chips. You can't just do it and turn it around in a day. It's years of investment and we started that a bunch of years ago. The Graviton processors, each one of those chips has a A level of security that you have to have in order to secure the stack from top to bottom. And when you get up to the top for the the inference workloads and the ML training model workloads, you look at our training chips and our Inferentia chips. They're the ones that we use to deploy to customers who want very cost effective, high performance training and inference for their generative models. Now, like I said, we always love giving customers choice. That's why we also offer NVIDIA chips for our customers because some people have written things that are specific for Nvidia's workload sets, in which case they want to run them on AWS. No problem, we'll give them the opportunity to. OK, So it sounds like it's really a, a case by case basis, but are your chips in all servers or there's some servers depending on the customer where you have all the NVIDIA chips? It's depending on what the customer chooses. We have, I don't remember the number off the top of my head, but several hundred different SKUs for server types within the company that we offer to customers. And as a result, some of them have NVIDIA chips, some of them have AWS chips. The CPUs are the same way. We give customers the option to use AWS Graviton or an Intel CPU or an AM DCPU. Alright, Speaking of chips, there's a report out a few weeks ago that Amazon refuted, to be fair, but the report was that you would wait to buy Nvidia's wait from buying Nvidia's current chips to buy later models. A lot of people call that an Air Pocket. So just kind of important for me to ask you again, even though I know you refuted the company that previous report, will you pause buying at any point to possibly wait for the Rubin architecture that's expected in 2026? The fact that NVIDIA surprisingly announced some new models, does that change any of your strategy? Our strategy is set and it's the one that we talked about in the in the refutation. OK, fair enough. I also want to talk to you about municipal customers today, municipal customers, actually cyber security, municipal municipalities are in the news. The city of Cleveland is being impacted by a cyber attack. Wichita, KS also impacted. There have been reports all around the country when it comes to municipalities and cyber attacks. What customers do you have? What are you hearing? So our customers, whether they're a municipality or state government or a federal government, all have the same concerns, which is how do they protect themselves against things like nation state hackers on one side and ransomware actors on the other. And fundamentally, the security controls that are necessary to do that are the same regardless of who you are as a target. It focuses #1 on multi factor authentication. As you and I have talked about before, ransomware is a huge thing for a lot of companies and governments and the most effective way to protect oneself against ransomware actor is hardware based multi factor authentication for every single login. It's the thing that prevents a lot of those actors tools from working successfully and it's what we focus on with customers. The 2nd is reducing human access to data. If people don't have access to data, their identities cannot be compromised and used to leverage that data. I know AWS along with another, a number of other major tech companies signed the Secure by Design pledge from SISA. I want to talk to you about that pledge. How meaningful is that for your business? And also just to come full circle, finding talent that can handle both cybersecurity and AI. We talked about a cybersecurity worker shortage here in the US. Is there a shortage of those kind of of workers that you're talking about that understand both AI and cybersecurity? And where would you put that at? And so as I said in my keynote, finding security talent is hard. And you're right, we've talked about that before. Finding AI talent is also hard. The intersection of those two is even harder. And so it really made clear to us the fact that we've got to take the limited talent that everybody's got access to, have them focus on building guard rails for all of our developers. But most importantly, tools that can be used to enforce the behaviour that we expect across the developer population, whether it's ours or somebody else's. All right, Steve Schmidt, Amazon Security Chief of Security. Thank you so much. We really appreciate your time. Always great talking to you about cyber security and the entire landscape. And we want to thank the audience as well for watchingcnbc.com. Thank you. Thanks, Frank. Have a good one.