🦄 Making great presentations more accessible.
This project aims to enhances multilingual accessibility and discoverability while maintaining the integrity of original content. Detailed transcriptions and keyframes preserve the nuances and technical insights that make each session compelling.
Overview
📖 AWS re:Invent 2025 - Accelerate mission outcomes w/ recent AWS government regions innovations-WPS204
In this video, David Schatzman and Sean Osfeld discuss AWS's commitment to security, operational excellence, and AI innovation for regulated industries and government sectors. They highlight AWS's $50 billion, ten-year investment in federal, state, and local governments, delivering 1.3 gigawatts of GPU/CPU capacity across four US regions including GovCloud, Secret, and Top Secret partitions. The session covers AWS's compliance certifications (FedRAMP, DoD Impact Levels), managed AI services like SageMaker and Bedrock, and the importance of clean data strategy for successful AI implementation. They introduce the regulated industries AI innovation flywheel, emphasizing responsible AI frameworks, Bedrock guardrails, and incremental delivery approaches. Key recommendations include working backward from customer requirements, leading with compliance, and leveraging managed services for cost efficiency.
; This article is entirely auto-generated while preserving the original presentation content as much as possible. Please note that there may be typos or inaccuracies.
Main Part
Security, Compliance, and the Convergence of Cloud Computing, AI, and High Performance Computing
Without further ado, let me welcome David Schatzman and Sean Osfeld. Thank you, John. We appreciate it. Can everyone hear me okay? Awesome. So for how many of you is this your first re:Invent? Welcome. Are you having a good time so far? Learning a lot? Excellent. This is a conference for builders and learners, and we're so thankful that you came today to our session. We have a lot to cover in just a short amount of time. We have 15 minutes for this lightning talk, so we're going to move really fast through a bunch of exciting content. If you have questions, Sean and I will be down front afterward and can answer your questions.
Just three things today. We're going to be talking about security and operational excellence and regulated clouds. We're going to be talking about customer obsession, and we're going to talk about how these things come together to help you transform your mission and vision using AI. The first thing I always like to start off with is to make sure it's front and center of your mind that security is the most important thing that we do here at AWS. Protecting your data, protecting your customers' data, and protecting your reputation is the most important thing that we do. As we go through this brief set of content, keep in the back of your mind that AWS is here to secure your data and your customers' data.
Let's ask ourselves a question: Why do customers use AWS for AI and innovation? Sean, talk me through it. It's really our global scale and being able to deploy the resources you need where you need them. We have 38 regions and four government regions. We have a lot of AI infrastructure, cross-region inferencing availability, and security behind all of that. It really comes down to offering in the cloud the services, features, software, and technology where you need it and at the compliance and security levels that you need it at.
When we think about security and compliance, we're going to talk in the context of the United States, but just know we do have a global reach as far as security and compliance goes. AWS was founded in 2006 and interestingly enough, when we began to offer cloud on a global basis, we didn't see a lot of adoption within the public sector, specifically within the United States federal government. We started to ask ourselves why we couldn't help the United States federal government, state and local governments, and the broader public sector market transform their mission. At the time, Jeff Bezos, our founder and executive chair, sat down with us and we thought about what we needed to do to advance the mission of government. That led to the birth of our public sector business and also led to the introduction of a brand new set of regions and capabilities in 2011 that over time has grown into a portfolio of multiple partitions and multiple security enclaves. It's an investment that we're proud to say has been able to help transform government.
Now let's fast forward to today. What's going on in the market today? For me and my colleague Sean, we see a great convergence going on. You started to see a lot about that even yesterday morning when Matt began talking about AI factories. I would believe that within the public sector there are four things converging that will allow your agencies and industrial-based customers, whatever market you're in, to transform impact. The first is the convergence of cloud computing as we've noted since 2006. The second is an interesting shift: deployment options are becoming more and more interesting to our customers. When Matt yesterday morning announced AI factories, that further extends our agile deployment option. If you'd like to know more about AI factories, you can visit our website or we'll be down front later. Customers want not only region-based capabilities but also capabilities extended closer to them.
Of course, AI factories is one option, but within our edge computing portfolio, we have AWS Outposts in a rack format, a one-rack and a two-rack unit form factor. We also have our content delivery network. These agile deployment options are becoming more and more popular. As you look to deliver capabilities to your constituencies, think not only about cloud as an infrastructure and capacity option but also about how you deploy it in a customer-obsessed manner to deliver the most value to your customers.
When you have things like Outposts, you can still use that same cloud formation infrastructure as code to deploy closer to where you are. That is a part of the agility that you really get with cloud infrastructure at the edge. Now, while capacity is a wonderful thing and we all need it, capacity needs services and a value proposition on top of it. When we start thinking about what services are most relevant to our customers for developing solutions, well, obviously here at AWS we value generative AI. We think it has a transformative impact on just about every area of business, public sector or not.
I would suggest to you that generative AI is another cornerstone of your execution strategy. Managed services like SageMaker and Bedrock are two wonderful AI services I would strongly suggest you take advantage of. But even if we start looking at other services, code development and the ability to transform legacy code into modern innovative solutions are even more important now than they have ever been before. Now let me set you up for the next slide by suggesting to you that high performance computing as a capability is becoming more and more important.
I know a lot of folks might think about high performance computing in the context of drug discovery or the scientific community looking at the cosmos and stars and plotting distances. But high performance computing has a much broader impact. What I start to see in the public sector a lot over the last year is more and more interest in high performance computing solutions, from small commodity class workloads all the way up to exascale workloads that you see sensationalized, for example, in the Department of Energy. What I'm saying to you today and what I'd like you to take away from this is that these four factors on this slide right here are coming together, and if integrated properly by you, will be able to transform the value of your mission and deliver additional business impact that you might not realize by looking at these elements in isolation.
AWS's $50 Billion Investment: 1.3 Gigawatts of Capacity Across US Government Regions
It's really exciting because I led up to you with high performance computing and generative AI. How many of you saw the Amazon post a week ago Monday that talked about a major investment in the public sector? Well, let me give you a little bit of information about this. This time last year we started thinking about how we could really help the United States federal government transform and execute on citizen services and mission and to modernize and be relevant. We were also listening to the feedback we were getting directly from our customers. We were working backwards from the customer, not just leading in ourselves.
When the turn of the administration happened early this year, there was an executive order about AI leadership and global AI leadership and being able to integrate technologies into the federal government using AI to transform mission impact. So we made the decision and announced it in November last year, as the slide suggests, a huge investment in the federal, state, and local governments as well as the supporting industrial base. A ten-year investment of fifty billion dollars to help transform the federal, state, and local governments. Well, what does that look like? It's about a 1.3 gigawatt investment. What does 1.3 gigawatts mean? That's a pretty nebulous number. We measure GPU consumption and CPU consumption in the context of power, right, with Moore's Law, which we can talk about later. 1.3 gigawatts is a little less than the Hoover Dam, which produces about two gigawatts of power.
Just to say something kind of anecdotal and funny to make it real for you, does anyone remember Back to the Future and the DeLorean? It was 1.21 gigawatts. So we're investing about 1.3 gigawatts in the federal government over ten years. So what does that look like? How can you access it? What does it do? Well, a couple of things. One, we've heard loud and clear from our customers that they are investing heavily in generative AI. Going back to that previous slide, generative AI in the bottom left corner, if you remember, is a wonderful set of services and tools that have transformative impact. But what do they need to operate? They need capacity.
GPU, CPU, storage, memory. The 1.3 gigawatts of capacity in the form of CPUs and GPUs will enable the use of our services, and customers can also purchase raw GPU capacity for their own model training. The other interesting aspect is that we will be offering supercomputing capabilities as well, HPC at a broader scale. We believe that this investment will be transformative for the government and will help you, whether you're in scientific or manufacturing or whatever business line you're in, to deliver value to the government as well.
How will this be operationalized? It will be available in the console starting in 2026. We will be building this out, and whether you're a commercial customer or a GovCloud customer, it will be available in your console as additional capacity and services that you can take advantage of. Let's talk about where we're going to put this capacity and how it's relevant when you think about it through the lens of compliance. We are going to land this capacity across all four of our US regions. What you don't see on here that we offer outside of the US is the European sovereign cloud and other offerings, but this is primarily for the US. We have commercial, we have our GovCloud US, we have our Secret region, and we have our Top Secret region. The three boxes on the right are all US citizen operations.
For those requirements and compliance requirements—FedRAMP, Department of Defense Impact Level 4 and 5, IL 6, Top Secret—we have all of the regions and we will have the resources available for the workloads that the government needs. We also have our commercial offering for state and local governments and other types of agencies that don't require that level of security. We are making this available across all of our partitions, which represents a large investment across all of these. We will not be adding any additional regions or anything like that. These will go into the current regions and availability zones that we have, extending those and making the capabilities even greater.
When I think about security, I think about the services that we offer. We already maintain all the data center certifications and accreditations that we need, and we offer you the tools, services, and features that you need to meet your compliance requirements or the government's compliance requirements. When you think about the AWS Government Regions portfolio and how you may be able to use it, think about the different compliance programs that you might need in order to perform on your contract if you're within a government agency in order to consume it.
The Regulated Industries AI Innovation Flywheel: Customer-Focused Strategies for Government Success
Let me use the last six minutes of our time here. I want to give you some insights based upon actual conversations with customers at scale, so we can provide you with some value when you think about how you're going to go to market to deliver your solutions. This is our regulated industries AI innovation flywheel. It aligns very closely with what I believe are the policy priorities of the United States. While we show these, if you look to the right of your screen here, we show it as a circular wheel, but it's really an interconnected mosaic. If you take all of these different factors and bring them together into a market strategy or into a solution for a government customer, you will realize business value. We won't go through all of these in detail in the time that we have, but we'd like to highlight for you some of the most impactful statements that customers have made to us to help you as you move forward.
The first one I think we need to talk about ties back to the first slide that I raised with security. Talk about the components of security. Well, security is our top priority—it's at the top of the flywheel. With the security and the options that we offer right now, when you're able to earn trust with your customers, we're able to earn trust with you as our customers. Going the other way, you're able to have that rapid pace of innovation knowing that a lot of our services come with those security features built in, or we give you all of the options that you need to build those compliantly.
A lot of that starts with data. If you have to have good, clean, labeled data to really do AI innovation, that's why we have all of the S3s and the services and the data storage that we have with all of the compliance that you need in the appropriate regions. I want to make a comment on data because you said something that really caught my attention—data strategy, clean, structured, and organized data.
Data strategy is critical to success with generative AI. Your data must be clean, structured, and organized. If your data is not clean, not structured, and not organized, and you try to run inference against it, the expression "garbage in, garbage out" comes to mind. Your output is only as good as your input, so it is very important to implement this advice when implementing an AI solution.
Responsible AI encompasses many things working together. We have executive orders that guide what responsible AI is. We did a chalk talk on Bedrock guardrails, where we discussed allowing security to wrap around your AI, putting roadblocks and guardrails in place to ensure you don't have prompt attacks and bad actors misusing systems. You need to show compliantly that you can meet regulatory compliance requirements. A lot of compliance is not just technical controls, but demonstrating those controls and having the talent in place to meet them. Data residency, sovereignty, and capacity all come together to form responsible AI because it is not just a standalone thing. It is part of a flywheel that includes innovation, pace, and all of those elements wrapped together.
Let me give you a few closing thoughts based on customer feedback. If you want to be successful interacting with the federal, state, and local government, whether you are working for an agency or part of the industrial base, the most important thing you can do is understand your customer's end requirements and work backward from them. First, figure out what compliance programs are important to your customers, whether it be FedRAMP, ITAR, or the DoD Security Requirements Guide. Figure out what that is, how to achieve it, and how to deliver incremental value. Lead with compliance.
Second, make sure you have a secure and responsible AI framework that protects your reputation and your customers' information, such as Social Security numbers and other personally identifiable information. This is very important. If you want more information on responsible AI, we will provide that content after the presentation. Additionally, in the age of efficiency and cost efficiency, we can help you architect the most efficient solution. Use managed services where you can, but if you need to pay for instance hours, work with us to figure out how you can be most efficient.
Another important point relates to risk management. Do not start out wanting to deliver a major, complete solution that does everything. No plan survives first contact, so we work around that through incremental delivery. We do proofs of concept and proofs of value to show your employees, constituents, customers, and stakeholders that it works. Proof of value is very important because you can build the best AI tools, but if they do not have value and are not making up for their costs, you need to start over and make sure you are solving a valid problem.
The last thing I will leave you with is to ask us. There are thousands of AWS employees here today, and we are here to help you be successful. We will be available afterward if we can answer any questions for you. We have prepared a paper that summarizes much of the content we spoke about today and gives you additional information. If you cannot take a QR code because you are with the federal government, you can simply Google GovCloud, hit our home page, and it is right there.
I also want to mention that David Levy, our Vice President of Worldwide Public Sector, is doing an innovation talk today at 4 p.m. You can use your phone to get the room number and sign up. It is also being live casted, so you can pick it up through the app and on YouTube, just like this presentation will be in about two weeks. David will be talking about the announcement based on the Amazon post from last week, as well as a number of really exciting topics that would be impactful for anyone working with us in the public sector market.
Thank you very much for your time. Have a wonderful re:Invent, and we hope to see you at the re:Invent watch party rewatch on Thursday night.
; This article is entirely auto-generated using Amazon Bedrock.
Top comments (0)