Securing the Sovereign Cloud with AWS and SUSE

Hey everyone, welcome back here to Techstrong TV. I think this may be our last episode for Prague, SUSE CON. Last.

Saved the best for last. We really did save... Well, you were second to last and last.

I know, and I'm here with my bud. Let me introduce your bud, Manu Prabakar. Yes.

Hi. Very good. Perfect.

I had a little help. Someone told me to think of a rhyme, and it made all the difference in the world. I just think it's amazing he came up with a rhyme for Prabakar.

I know. Now I'm going to use that moving forward. Prabakar.

Absolutely. Yeah. But thank you.

Manu is with AWS, but there's a lot of people with AWS. Manu, tell us a little bit about what you do at AWS. Yeah, for sure.

So Alan and I lead all of our business application partnerships, as well as our Linux partnerships globally. So we've been associated with what we're doing together with SUSE for almost a decade now. That's fantastic.

And this, I feel necessary- Okay ... to introduce Margaret Dawson, who, if you've been watching along, was just on, and of course, Margaret is the CMO at SUSE. She hasn't been here for the last 10 years, but in the short time she's been here, she's made a huge difference, and I think this show- For sure ...

is testament to it. Incredible. Guys, let's jump into some things.

Okay. Before we get into our formal questions, I feel compelled. Look, AWS has made a significant, as in billions, significant investment in sovereignty, digital sovereignty, in sovereign data, sovereign infrastructure, a field that SUSE has really emerged as a leader in.

And, so you had a solid partnership for a decade. Mm-hmm. But this has taken this partnership, I feel, to another level.

Before we even get into the SUSE side of it, Manu, I'd like to hear from you about how does AWS feel about this sovereignty issue, and why SU- Well, I think I know why, but tell them why SUSE is a great partner for you with this. Absolutely. I think at AWS, we think of sovereignty not as a constraint, but as a design choice, and as a design choice to give our customers the comfort to scale responsibly.

Mm. And so in that context, there's a sovereignty of data- Mm-hmm ... there's a sovereignty of operations, and there's a sovereignty of the technology stack.

When we think of data, as your data resides in the context of European Union, it sits in our data centers, our regions in Germany. They sit in the region. From an operations standpoint, all the employees that are handling that data, the processes, the operations, they are EU residents.

Mm-hmm. And from a technology stack, we are working together with SUSE on infrastructure that is scalable, that is resilient, that is open source, so the customers have the flexibility to move their workloads. Mm-hmm.

SUSE was part of the launch of ESC that we just did in mid-January. We had 90 partners on launch. Sorry, 180 partners on launch.

180? 90 services. So it is incredible.

5 billion investment towards ESC, and we just are now launching local zones in different parts of different cities in Europe as well. Hmm. So I think it is a secular investment for us.

It is what we're hearing from our customers, and SUSE's been part of that. The operating system, the rancher offering, everything is now available on ESC. Yep.

Margaret, what's the SUSE view on that? Yeah. No.

Similar type of storyline in terms of digital sovereignty overall takes an ecosystem. None of us can achieve this on our own, right? And I think it goes back to what we don't want to see is for sovereignty to end up meaning a closed ecosystem or a closed environment or take away people's choice to be agile because you still have to innovate.

You can't just lock everything down and say, "Oh, now we're safe, we're sovereign," because you're going to kill your business growth. And so for us, having a world leader in cloud infrastructure like AWS give that commitment to sovereign cloud and be able to be part of that as well as how we are a partner in on the marketplace of AWS overall, is a perfect match, right? And then we're looking to other parts of that ecosystem, whether it be other software, other hardware, et cetera, right?

Because at the end of the day, it's about giving the customers choice in how they want to deploy and build and manage their sovereign environment, right? And it means something a little bit different, like you said. I love that for some people it's more on the operational side, right?

But there is a commitment as any technology vendor in what we are providing, and for us it's the same. If we say sovereign support, that means it is support that has the people in the jurisdiction- Mm-hmm ... of which you need to be kept sovereign.

Right? The same thing with data. But that doesn't mean you can't have the flexibility, the agility, the choice in how you are building your environment.

Yeah. Great. And I think I love Margaret's point about keeping it open and not having this as a tax.

So it's an enabler. Yeah. And I think the other point is there is no price premium for customers in the European Union who are using the European Sovereign CloudSo I think that's just also another data point.

I think that is important. But you mentioned the C word, customer, and that's really... We could talk in circles all day.

One of the nice things about being at a SUSEcon or a re:Invent for that matter- Mm-hmm ... is talking to real-life customers and hearing what do they want. Because sovereign shmoverein, if this is not what they want, it's often not.

What are you hearing, you're hearing as CMO here, this is your event, what are you hearing from your group of customers? Yeah. And I'll kind of tie it to other themes that we're talking to, right?

What I'm seeing is that this sovereign conversation is not separate from the AI conversation. That's right. Yeah.

Right? Is not separate from the modernization conversation, is not separate from all these different... The VMware conversation, whatever it is that we're talking about.

So it's one element or capabilities or criteria that is kind of an overlay across all these other things. So the fact we're now saying, "What is my sovereign AI strategy? " Yeah.

That's really critical because the agents and the data that they're referring to sometimes are going to need to be sovereign. Yeah. And I think when you asked what are we hearing from our customers, I think we are hearing AI agents- AI ...

and more AI. And maybe- Yeah ... I think we all are.

Let's face that. All. But you mentioned modernization, and I spoke about it before.

There's migration, modernization, transformation. Yeah. I think those are way stops along the way of moving to the cloud, moving to a new virtualization layer, adopting AI.

Right. Right? But when we talk about mission-critical industries, telcos- Defense ...

healthcare, defense- Yeah ... things that we can't afford to experiment with, right? We need to know these are not even five nines, six nines and above.

That's a different animal to a certain... You just throw that SLA out there like it's easy. I did that for years.

My Lord. It's a whole another story. Yeah.

But it's scary for a customer because it's their- Well, I think- ... for your customer, it's their business on the line. Yeah, but I think it's more than just the mission-critical nature of their business.

It's where their work is going. And what we're seeing, kind of going back to your earlier question, is a lot of this is moving further away from a traditional centralized data center, right? Yeah.

We're moving further to the edge. AI is moving to the edge. Sensors are moving to the edge.

Intelligence is moving to the edge. The data is moving to the edge, right? So how do you think about just the structure of that, let alone the intelligence that's coming from that, the residency of that data, the latency of that data?

There's just so much complexity. How do you make people comfortable? Right.

I was talking to a gentleman who works for a very large European grocery chain this morning, and they're thinking about everything from sensors on their inventory to their point of sale system, and they're trying to figure out, which development platform should I allow my developers to use? Yeah. And then where do I standardize?

Where do I... There's just so much complexity when you're getting out to things being on the edge. Yeah.

And I think talking about mission-critical workloads, SUSE and AWS have more than 5,000 customers that are running those mission-critical workloads for a decade. And you think about these are SAP workloads. We talked about telecom.

Great point. O2, Telefonica, British Telecom. These are all customers that are building.

And then we are also jointly supporting Ericsson and Nokia as part of the- That's right ... underlying substrate that's really powering this telecom ecosystem. So, bringing this back to AI, AI, we have to meet where customers are.

Their data right now sits on the SUSE stack together with AWS on our platform. And so we are really building the AI layers on top of that, whether that is innovation on the silicon layer, or we are innovating in terms of what we're doing with Bedrock or on top of that- That's right ... with building the agentic infrastructure or high-level services.

So it really comes back to meeting customers where they are. And there's a decade of earned trust that the two companies have done together, so it really helps there. Yeah.

Agreed. We've been skirting around AI here now. We mentioned it a half a dozen times.

Of course, my personal opinion is we've crossed another Rubicon in the last two to three months where it's not AI, it's not enough, it's agentics, it's agents. Even at Techstrong, we're a little nothing company. We are big into agentic experimentation, transformation.

Again, a matter of trust, though, a matter of how do we deal with this, right? First of all, it's so fast. We think about governance, right?

But unfortunately, we always think about governance after the horse ran out of the barn. How does an AWS SUSE partnership help us, again, give us confidence, make us feel comfortable with... Because the fact of the matter is, whether we like it or not, our people are using agents, right?

We call it shadow AI. Right. Call it whatever you want.

But they're being used. How do we get our arms around this? Yeah.

So first of all, SUSE and AWS have been working on a mission-critical workload. So I think there's already a long history of supporting our customers from a governance standpoint, from in a public sector, financial customers. So there's already we are working together.

I think how we think about agents is that there is definitely risks of hallucination from both the LLMs and then the agents that are built on top of that. So every time there's a hallucination or something bounces off, you can keep the agent, the blast radius of the agent, small and narrow, and you can build these incremental agents on top of that, so you can have a bigger impact on the workflow. So we have services like Agent Core, Bedrock, Guardrails- Yep ...

that really make sure that the agent stays in the state and narrow. And then on top of that, it really depends on what is the data that we are really training. Exactly.

What is the RAG and what are we training that data from a RAG perspective? And so the more cogent your data strategy is, the cleaner that data is, the more it is functional and it is tagged, the more likely the LLM it is going to be, and the agent is going to be to make the right sets of action. So I think it is both in terms of our data strategy and then tying that to what we are doing from a guardrails perspective.

I think some of the data that we have is we are able to catch about 88% of the hallucinations that the LLMs, and then of course, the agents are making on top of that. And that guardrail function from LLMs on Bedrock is now also being extended to Agent Core as well. In addition to that, we are also now partnering with a bunch of our ISVs, for example.

Mm-hmm. IBM is a big example, with SUSE, to really build on that trust layer to make sure that we are checking on the hallucination. But I think it's still a work in progress- Yes ...

as we scale. Yeah. It is, but you made me think of something.

When we were in early cloud, which I know feels like a lifetime ago, it wasn't that long ago- Yeah ... which is just so strange. I remember working with customers and saying, "Network security- Cloud security ...

is cloud security," right? Like sometimes- I was thinking the same thing ... we over-pivot to this, and what I would say is data security is AI data security.

And the things that we are working with at the infrastructure layer is role-based access controls are also relevant to agents, right? Just as you would look at your employees, and there are employees that don't get access to the keys of the kingdom. You probably don't want an agent to have access to your source code, for example, but they may want to have access to other parts of your development policies and language libraries and things like that.

So, at the fundamental level, set up your policies just like you would sometimes people or whatever, give it a persona, I don't care, to start with. So there's some foundational things. I don't mean to oversimplify it.

No. But there are some things that are foundational to how you operate that should continue and be applied to this new world. Now there's new things that go with it.

Yeah, I know. Yeah. But still.

And to Margaret's point about role-based access, and a lot of the data right now is residing in business applications. Right. Think about companies like Salesforce and Adobe- Absolutely ...

and ServiceNow and Workday. Yep. And a part of the other rich ecosystem that we have at AWS is really making sure that as these companies, which have the full context on the data, they have like, who's the user, what privilege of data access that this person has, we are able to power agentic experiences.

For example, Agentforce is being built- Mm-hmm ... on Bedrock. The Workday agentic solution is being built on Bedrock.

The stuff that we're doing with ServiceNow is being built on Bedrock. So we're really meeting customers and our ISVs where they are to really keep the guardrails on. I think that really is- And I think you're speaking to the ecosystem again, right?

Right. I think what customers want is for us to be working together, and this has been an age-old thing from enterprises, right? Enterprises have always said, "What I need is for my vendors or my technology providers- To work together ...

to collaborate so that I'm not the one having to worry about all this," right? And in the day it was like, do you integrate together? Do you play nice together?

But in the age of agentic AI, right, and all these different models that we're doing, we need to come together to have that governance, to provide those rules of the road, and to help them be successful. On the governance level, yes. But another important thing, and there was announcements this week, the MPC servers.

MCP. MCP. I'm sorry.

MCP. Model context protocol. Yes.

Think about the words I- Multiplayer. No, not multiplayer console. I was going the other way.

We could do gaming if you want. No, we could. But yeah.

That's not mission critical today. But yes. MCP.

Because yes, first they used to have integrations, then we had APIs. Yes. Right.

Right? And we had the API economy, if you will. Right.

I think that's now- It's still there ... being replaced. Right.

Now we have it's running on top. No, I think it's MCP gateways. No, I think absolutely, and I think how those are interoperable, and fortunately, we do have a standard, but as you and I discussed earlier, I do worry.

I'm feeling the universe, people wanting to take things back to the black boxes, right? Yeah. And I think it behooves all of us.

Oh, no. You know. Look, there's a company that's just changing their CEO who has made a really good living having a black box, a walled garden.

There's always going to be, I think, a place for that in the market. I think overwhelming, though, people vote for open. They vote with their wallets and their- Well, those black boxes, they still want to know how they can interoperate with other, either black boxes or open standards.

So there still has to be that interoperability. And I think, too, compliments to SUSE, the entire product portfolio now is being offered as an MCP server. Right.

So, for example, if a customer is sitting within Quick, that is our enterprise user interface. Mm-hmm. They can configure their SLES armies running on EC2, or the Rancher workload, which is now managing multiple EKS clusters.

Right. So we just love the capability that we are building. No, no, but that's becoming the connector.

Right. Where we used to have, do you integrate, do you work with... Now with this, anybody could work with anyone.

They could talk to each other, at least, right? And make that happen. I mean I mean, the difference is which is great is- You're out in front with this ...

yeah, I would actually say with APIs, it didn't have the context that we see with AI. So that the benefit we're having in this next level of connectivity, is you've got the inference, you've got the context. You can actually take actions from it, whereas APIs were static.

It was meant to do one thing. You had to teach it that rule. I think a lot of the ISVs also have to reimagine their API infrastructure in terms of, as you said, before API was primarily to get information.

Right. Give me the data pull on this section. Right.

Now, the APIs, because of the more context, they're also now being rebuilt- Yeah ... to also take action. Right.

" So I think that is a new capability. Yeah. And LLMs, for that matter, they can be decision-making, or they can be the brain of the system.

But the MCP and the tools, they really need to both provide the data and then take the action. And I think that plumbing is still being built. That's right.

Yeah. Absolutely. I think that's right.

Let me ask you the money question. Oh. When do you think enterprises are comfortable enough to put their trust in and say, "Okay, we're all in here on agentics.

We're all in here on our MCP servers. " We're done experimenting. We're operationalizing.

I mean, it's there now with a lot. Whether they know it or not, because the reality is they're doing it with almost every application. Their people are doing it.

Right. And almost every application you use, even if you're not consciously doing it. But I think you're asking when are enterprises going to be intentionally saying, "Okay, we are building these workloads.

" There's such a different maturity phase that literally we have some that are there, and we have some that are barely scratching the surface. So I think it's going to be very much like- It's the regular crossing chasm ... cloud and other things.

Correct. Yeah. 15% early adopters, 35% early adopters.

And it always starts, there's a pull from your employees, from your customers, from the market. Yeah. I think there's a healthy sense of paranoia- Yeah ...

in the industry. There still is. Right.

I mean, yes. It's good to have. And customers don't want to be the last person who totally missed the big wave on that.

Right. I think so that's one. Well, there's the FOMO.

Yes. Right? But they also don't want to be the one that has a data breach.

But they don't want to be the first one either. Yeah. Right?

Well, no, that's not true. There is always going to be those early adopters, and you see them. 15%.

Right. Not that much now. There's always 15.

Yeah. Well, that's the classic crossing chasm. Yeah.

And I think that FOMO is then married with just almost an incredible experience that customers are having in their outside enterprise of experiences- That's right ... of working with ChatGPT. Yeah.

This is the ultimate- Or working with Gemini ... consumerization of... Yes.

Yeah, yeah. " That's right. Why do I have to go search through Slack and my Outlook emails, and I still have to put the context together?

It is very frustrating. Yeah. And I think from a production use cases in agentic coding, a lot of the stats we are seeing that almost a lot of enterprises are now doing 50% of their coding- Yeah ...

through agents. So I think that's a plus one on a use case that is getting production level scale. Absolutely.

I would say the other use case is around customer support, customer experience. Right. Agents have started to take a lot of that traffic, at least do the first, second, and third-level calls, and also have a lot of enriched data in the context about who am I speaking.

I know a lot of customers are so frustrated. Again, I have to talk to a bot. He or she doesn't know anything about me.

But at least now, agentic experiences, they already know. They have tied data through Salesforce and ServiceNow. What were the tickets you opened before?

So I think we are definitely making increment- And I think you're making a good point, that it is very much use case driven right now. Yes. You just talked about two of them that were ahead because there was the early promise there.

Right? And it was easier to implement. So I think we're going to continue to add those use cases as we move forward.

I think we've seen this in generative, right? I'll tell you where I see it. Software developers.

Right. That's an early use case. Absolutely.

Right? Absolutely. Most software developers today are not actually writing code anymore.

They're generating code. That's what I call it, anyway. Right.

They're generating. In marketing, I just see- No, I was actually going to bring that up, is that AI is already foundationally changing the way we do marketing, right? Absolutely.

I have agents talking to customers in any digital space. I have agents that are emailing people and setting up meetings. And for the most part, people aren't always sure, is it a human?

Is it an agent? Right? So the lines- It's very hard ...

are blurred. Very hard still. But I will say this, regardless of how mature or advanced or early in the process is, I am still seeing an absolute human desire to still do this.

Like to still do this. Engage at the human-to-human level and engage with content that you know was created by humans, art that was created by humans. And maybe we're getting a little philosophical here, but I'm tired.

But I also believe philosophically that we will always have a piece of that because it's the way we are wired. Sorry, I had this discussion at your party last night. Did you?

Yes. Do you think there'll be a market for human-generated code? Remember, Rolls Royce was a handmade car.

Right. Bugatti- Right ... is a handmade car still, I believe.

Yeah. It might be, I don't know, $800,000. I wouldn't say yes, because the one thing we have learned through technological innovation over and over since the Industrial Revolution is that every time there is something that is that shift, that chasm-crossing innovationWe as humans advance to a different level, right?

Either of creativity, of artistic capability. Yeah. And so I think I'm excited because what are we going to do with this?

Yeah. To me personally, I- Are you more skeptical? I think this gen AI is really democratizing learning.

Oh, wow. And building. And learning and building.

And I think a lot of, in school, we've always had a very standard way of educating. That's great. Now with gen AI, you have so many...

Do you like to read books? Do you want to watch videos? Do you want to see graphics?

Do you want somebody speak to you? Do you want to have a personal tutor? And a lot of people could not afford that luxury.

No. It was a standard operating procedure and until for the first 15, 18, 20 years of your life. I think it's giving- No, I love that because I actually was thinking that the digital divide has already been so great, especially in developing countries that may not have access to internet.

That's why I call this the AI poverty line. And my fear has been that AI is going to have its own gap, right? But you're actually bringing up something different, which gives me a little bit more hope.

Where can AI actually narrow that gap because people will be able to learn and to grow more easily. Which takes me back to this, why we are here- Yeah ... about open source.

Just the push of all the open source models that are really pushing the prices below, are hopefully going to continue to keep the affordability high for a lot of the customers who- Wow ... who can continue to access this. I love that.

Not just affordability, but- Look at you, like full good human ... access period. No, I personally benefit.

No, I love that. That's a good place to end. I do.

I like that. Right, we started our SUSECON- Are we going to sing "We Are The World" now? Because I feel like we're- You don't want me to sing.

Oh. But I'll- We are the world. There you go.

Margaret, look, you've brought it up. You can only sing. I found a hidden talent.

You can't leave them hanging. We're going to do it later because I- Okay. We'll do it later.

We won't video it. Very good. But you don't want me to sing.

But you know what? This is where this conversation started on Tuesday or Monday, was I first started recording about open source. Yep.

SUSE has been an open source company, an open company for 30 plus years. AWS, big open source shop, and user, and supporter- Yep ... since the get-go.

And that's where we're going to end it, right? Awesome. Thank you, Alan.

Thank you. Thank you. Thank you.

Manu, always a pleasure. Yeah. Hey, that wraps up SUSECON 2026 for us here at Techstrong TV on our recordings.

But we've got a lot more written content. There's about eight or nine articles being published. We're going to be doing a wrap up tomorrow as well.

There's one more edition of our SUSECON daily newsletter that'll go out tomorrow. tv starting now. So until, I don't know, Monday I'm back on.

This is Alan Schimmel. We're out.