NVIDIA’s AI Security Power Play: CrowdStrike, Cisco and HPE Join the Stack

Welcome to Security Boulevard, the cybersecurity podcast from the Futurum Group. Each episode explores a variety of topics within cybersecurity and the technologies behind it. com, the Security Boulevard YouTube channel, Techstrong TV, and all of your favorite podcast platforms.

Before we jump into today's fun topic, let's meet the guests, starting with our good friend, Alan. Alan, it's good to see you again. Tom, it's always great to see you, especially right after RSA.

What a great gathering of the industry, and I'm looking forward to talking about the industry today, Tom. Absolutely. My feet still hurt, so that's why I'm sitting down today.

But joining us also is our good friend, Mitch Ashley. Mitch, as always, a pleasure. I'm a little hoarse from RSA, actually.

No, I'm fine. I'm good, but I do need a little rest. What a week.

It's RSA. What can you say? Exactly.

And of course, I'm Tom Hollingsworth. I am related to all things security here at Tech Field Day. So let's jump into today's topic, but spoiler alert, we're not going to be talking about RSA, because the week before RSA, there was an even bigger conference, and that was GTC.

" I would say that maybe GTC has started becoming the destination for people to talk about a lot of other things in the industry. But as luck would have it, GTC was filled with a lot of discussion around security, and that's the topic for today's episode, because it looks like NVIDIA is building a security ecosystem inside of their world domination plans for AI. And it's honestly not surprising given the amount of conversation we've had about AI over the last couple of years, that while a lot of press has been going to companies like Anthropic or OpenAI or Google for creating models and creating images of people with multiple fingers and things like that, the real dark horse in this for a while was NVIDIA.

Because they were the, well, let's call it what it is, they were the arms dealer for all of those people, providing GPUs, providing networking infrastructure, providing all of the things necessary to generate AI everything. And now Jensen Huang and his leather jacket, and yes, his leather jacket is a sentient being at this point. It has been imbued with the power of AI.

They are both- It's the Steve Jobs black shirt. Exactly. It is ...

turtleneck. It is. They are teaming up to kind of create these new ecosystems.

And we saw a lot of announcements from companies that we work closely with, like CrowdStrike and HPE and Cisco and many more. So I kind of want to let the two statesmen of the industry talk about this, because we've seen this before, especially when cloud was brand new. There were a lot of companies that were racing out there to talk about how they were partnering with different cloud providers to offer these services, but it kind of felt scattershot back then.

It really feels like everybody is coming to NVIDIA this time. Well, I'll kick off, Alan. First of all, I just give you the Steve Jobs reference.

It really has become the new Apple Steve Jobs conference, where everything else tends to stop, oxygen leaves, and the event horizon is at the edge of the GTC conference until that's over. " Now, that's actually true, but not in that sense. What's the big issue for getting AI into production?

He spent an hour and a half talking about software and applications before he really got into the hardware side of things. So, somebody gave him the memo about software eating the world. So he knows that they've got to have outcomes out of all this investment in equipment and data centers, et cetera.

So, one of the big ones is securing it, control planes, observability, all of those things. And guess what? All of those companies are trying to figure out, how do we play in this market, not just ourselves, but partner with the big companies like NVIDIA.

So I think it makes a lot of sense that I think that it's happening, natural evolution. You compare it to RSAC, which is more of a kind of inward-looking conference. I don't know if I say that in criticism or not.

RSAC could be like GTC. It just doesn't have a Jensen Huang to kind of ramp them up. Alan could do it, but he's busy these days.

Yeah. Look, I think RSAC still gets more people than GTC. Oh, yeah.

They definitely do. 5,000 people there this past week. But here's the thing, Mitch, I think you're onto something with what you said.

This has become the new Apple black turtleneck, I got the world in my palm with this phone kind of thing. Right? I'm throwing a hammer into the glitches.

But you got to give Jensen and his leather jacket credit, because what they do well over there, they're a great learner of history. They've seen the Jobs script. They've seen the Apple script.

They've also, though, seen, and I firmly believe this, they've seen the cloud rollout. Mm-hmm. Remember, NVIDIA was the gamer and then the crypto chip.

So they're very familiar with those communities. So they're being really smart. And he realizes that as much as we think Nvidia and we think chips, it's all about software.

Mm-hmm. It's about owning the software stack for inference and AI use. And that's what this year's GTC was about, right?

That's exactly what it was about. We got the stack. And by the way, the stack is headed up by, or at the pinnacle, at the top of the stack is the agent that does all of these things.

And we're going to own the agent all the way down that stack. But learning from the cloud industry rollout, A, he can't go it alone. As big as the $5 trillion he has in market cap is, he can't do it alone.

And they did an amazing job in rolling out a full ecosystem, of which cyber is a good piece of it, as it should be. But let's be clear, Tom, you racked off some big names, right? Mitch, you got it, too, right?

Some big names. From cyber, the two biggest ones for me were the CrowdStrike, JFrog, which is now, a DevSecOps or security company as well. Yeah.

Fortinet, Cohesity, there were a number of folks that had direct announcements with them. And then, Cisco was there, and HPE was there, and the usual suspects. So you got to give them credit for assembling this ecosystem.

Not just cyber, but the whole package, right? This is what Apple did. " "Tens of thousands," right?

And Mitch, remember our friend Andrew Greeley, who's over at Armis now. By the way, Armis had something at this GTC. Oh, they did?

Okay. " Right? He has an Australian accent.

He was a big iPhone- Yeah. Australian dude ... That's the key here, guys.

But let's dig in, if you don't mind, Tom and Mitch. Let's talk a little bit about, well, what do you got to do to secure this stuff? What exactly is CrowdStrike, and JFrog, and HPE, and all these folks doing?

And I think that's important. It's coupled too with, because I want to get into that, but they made their own announcements. They announced Open Shell, which is an open-source environment to run your on-premise or in the cloud, wherever the GPUs are, to secure if you want to run OpenClang or run anything else, right?

It's this kind of, I want to say containerized, but it's this walled garden. The goal is it meets enough security requirements of enterprises to get them comfortable to put this into production. So you stack onto that, right, announcements with CrowdStrike, announcements with Cohesity, Fortinet.

So they're piling on, not just with any vendors, but like CrowdStrike especially, who has Fortune 500 wrapped up. They do, right? They're in every one of those companies.

So it's about not just lining up the technology, I think it's lining up who has the customers that they want to make sure their software is playing on Jensen's hardware. So- That's the strategy ... So what's a customer in this case?

Because I find it kind of fascinating that we had, like you said, we had a lot of announcements of partnerships from companies like CrowdStrike, JFrog, HP, Fortinet. But how much bandwidth of that two hours of keynote, that wasn't just them talking about how Nvidia is going to make $11 trillion, was related to OpenClang? Because they had quite a few announcements around OpenClang, creating architectures to run it, creating their own system to accelerate it into what I saw online as Super OpenClang.

And that, to me, is one of the most fascinating parts about this whole thing, is a lot of companies that are out there, in a way, it's when you get any kind of vendor-branded conference, there are a lot of companies that sign up because they want to be seen near the proximity of a successful company, right? Whether it's HPE Discover, whether it's Cisco Live, whether it's RSAC, whether it's Fortinet Accelerate or the Extreme User Conference, whatever it is, we want you to know that we work with these people that you came to see. But with OpenClang, I don't think there was an OpenClang booth.

I don't think that there was any kind of, like, we're going to invite the guy up on stage to talk about it. It was, "Oh, we see that you're using this, and there's a lot of chaos in the industry. " So, in a way, OpenClang brought Mohammed to the mountain.

Like, that's the thing, is why did Nvidia suddenly want to tell everybody-- Actually, it would have been bringing the mountain to Mohammed in this case. Why did Nvidia want to tell everybody- Because the- ... that they worked with OpenClang?

This was the coronation of OpenClang as the de facto standard for your agent- Yeah ... infrastructure. And this is something Mitch and I have dealt with for 25 years.

This is classic, classic open source, right? Classic open source. OpenClang itself, I don't know if you've tried to play with it.

We've got a few people here running it on Mac Minis. Yeah. It's a b***h.

No doubt about it. You got- It's got a lot of claws. It's not just two.

There's a lot. It's unwieldy. You've got to be-- You're going to be working in your CLI.

Typical open source kind of stuff. HoweverOnce it's up and running, and we'll leave security out of it for a second, just a microsecond. " And so we've seen this, you saw it with Linux, what it did to Unix.

Right? It's probably a great example. You saw it with the, for as big as VMware was, it wasn't the default hypervisor for the cloud, was it?

For private cloud, but the public hyperscalers, they didn't use VMware. They used open source. KVM, Linux.

Yeah. Here come in Linux again. Same thing as Linux, exactly.

" The whole observability industry is built on OTel, on OpenTelemetry under the cover, right? Jensen saw this and said, "We're going to build this stack, this agentic stack," and he called it a five-layer stack. " And for people like me who still have a good place in their heart for open source, I all of a sudden now have a good feeling about Nvidia for building on open source and supporting open source.

And a lot of the companies you mentioned, Tom, and Mitch and I mentioned, they also at least give lip service to supporting open source. So again, brilliant move here. But open source is the agent that we're going to build on.

And the good news is it's so raw and so insecure that there's so much to do in terms of these guardrails and the security we're going to need to lock it down, as well as non-security things. Well, what does open source address, Alan? It addresses the number one concern of vendor lock-in.

Yeah. Now, if I'm running this on open source, albeit it may be the Neoclaw or Nemoclaw version of OpenClaw, which they've wrangled some of those claws into packaging it into a little easier environment to install and operate on top of their AI toolkit, which is that open shell, AIQ, and there was another component to it as well. But they're trying to build the environment to run that.

So yeah. And to your point, though, I want to really emphasize one more thing that you nailed, and that is, it was more than a coronation. 7 Richter scale event.

Like, whoa, oh s**t, what happened? Did stuff fall off the shelf, or do I need to go check my house, make sure it's okay? If you want a more recent example, it's the MCP of 2026.

MCP caught fire and was a big thing, and it's this really tiny, little thing. It's not a big thing. But it was the can opener to everyone who had a can that didn't own a can opener.

It was just like, now everybody could get to access their agents, their models, get access to tools and data. That's what OpenClaw is. And you wrote a nice piece.

I appreciate you quoting me on the piece about there will be a thousand recreations of, or innovations of OpenClaw. It will spawn generations of new things, and that's exactly what Jensen announced. That's the hammer that hit the sabotage into the gears of 1984.

It won't be like 1984. Right. It won't be a walled garden.

Mitch, but to MCP or not, that is the question. That's ano- Oh, that's another. Joint article.

You can check that out. com. Yeah.

But let's peel the onion back a few layers here, get into some details for our listeners. I want to talk specifically about JFrog for a second. So they announced something called the JFrog Agent Skills Registry, which is- Mm-hmm ...

designed to provide a secure trust layer, right, for AI workflows in development. And that JFrog Agent Skills Registry actually integrates into the Nvidia Open Shell runtime, which Nvidia announced at GTC. So it's an important part there.

They have an AIQ Nvidia, certified Nvidia AIQ blueprint, as well as automated security scanning now for that. And one other thing in regards to the MPC, they actually announced the first MPC registry, Mitch, where you can now download and check what MPC you have and what is going on there. So JFrog was certainly very, very busy here.

Let's talk a little bit about CrowdStrike too, though. They announced a major partnership embedding Falcon, the Falcon platform, right, directly into Nvidia's agent toolkit as well. So, this agent toolkit is the Swiss Army knife of using, in this case, OpenClaw.

But there's going to be OpenClaw-compatible agents built off the open source. Right? So Alan, let me talk about that real quick because I think that that's a brilliant move by Nvidia.

Oh, yeah. But for the reason of it opens a new market for them. Because one of the problems that we have to deal with a lot in the security space is authorization and risk management, right?

How many times have we heard or seen something in the news about a company that did something, and then it turns out, oh, well, you weren't actually supposed to do that. I mean, shadow AI, shadow IT, we've heard that for years. A lot of organizations, especially heavily regulated ones like finance or government or healthcare, they only buy off the approved vendor list.

How many times have we heard you never got fired for buyingCompany X, whoever that company is, IBM, Cisco, whoever. By embedding Falcon in their toolkits, what they're basically saying to companies that are very on the Herman Cain, nobody jumps out of an airplane because it might crash kind of thing, risk factor of zero. If Falcon is embedded in here and you're already approved to use Falcon as your scanning tool or your prevention tool, well, you're already approved to use Falcon, why can't you just use our stuff because it works with Falcon?

They're trying to open themselves up to those markets that would not have even given AI a second look, or a first look in some cases, because they don't feel that the security controls are mature enough. By having something that is integrated tightly with an industry-recognized name, and that kerfuffle a couple of years ago notwithstanding, CrowdStrike is trusted. " Because, for, what was it my friend Keith Townsend pointed out, on the small business line, it's only $155,000 for the entry level rack to do this.

Yeah. No. But that's an enterprise solution, Tom.

NVIDIA does have... Mitch, what do they call their little box that looks like a Mac Mini? And they have one.

They announced a new one. I don't remember the name of it. The new one is like this massive thing.

But yeah, it's- No, the new one, yeah, is 150 or 170 grand. And I'm not- No, this is like two grand or something like that. Yeah.

No, Keith Townsend has it. Oh, the Jetson. No, it was- The Jetson's the one he had, but I think you're right, they did announce a new one.

Yeah. And that's the Mac Mini. But you know what, Tom?

That's a whole another show we could get into is- ... how Apple, which we thought was an AI laggard, actually is making a very serious play to own the edge in terms of- Yep ... being the AI computer.

They already do own the edge, Alan. Right. Well, yeah.

That's the default choice is your Mac Mini- Yeah ... to run OpenCL. That's what our folks are using here.

Yeah. And this is, again, you got to give to Jensen and that team, they've got everything from this little Mac Mini-looking box that's, I forget what they called it, but whatever. This Mac Mini-looking box all the way up to this mid-range $170,000 behemoth, and then you get into their real stuff, right?

But here's the thing, they said this stack, this stack with the security, with everything, is going to generate a trillion, $1 trillion in revenue in the next year or two. Right? How do you justify a $5 trillion company?

Well, when you got a trillion dollars in revenue coming out of this stack, that might justify a $10 trillion company. I'm not a stockbroker. Don't take my advice.

But this seems to be a real boon to NVIDIA's thing. Here's another play, though. Tom, you mentioned this as part of this Agent Toolkit.

And Mitch, I'll throw it out at you, too. Is the Agent Toolkit another name for a marketplace? Maybe we should get Alex Smith from Futurum in here on.

Mm-hmm. Agent Toolkit become a marketplace. Because it seems like everyone's putting their stuff into the Agent Toolkit, and they'll probably be competing things, right?

You could use this for that or that for that. So I'll let Mitch go first, because I bet you I know what he's going to say, and I bet mine is the opposite. I think it's the start of the marketplace, and it's the open source start.

But you think about where they're going to go next. It all depends on what the channel strategy is, and this is where Alex would kick my butt on it because he knows so much more about it. But do you go to NVIDIA to buy your CrowdStrike software?

Mm. I don't know. Maybe it comes out over time, but I think what I said before, the whole channel strategy of go to the vendors who already have the customers that will get you in, and you're not some startup company, you're NVIDIA, so it's a pretty easy move, I think, for that.

But to your point, it is the start of a huge ecosystem, and you were saying this before, Alan. Everybody already wants to be part of running on NVIDIA's stuff. They want to be part of NVIDIA's announcements.

So I think there's a natural sales channel, partner channel, ecosystem there, for sure. I just don't know enough about how to set one up. I actually wasn't that much different than Mitch's response.

So I don't think this is the marketplace. I think this is effectively GitHub for the AI integration. " Yeah, for us.

GitHub is easy for people who know the difference between a pull request and a push request. But your grandmother doesn't. The people who run boardrooms, by and large, don't.

That's who the marketplace is for. To make it to an app store level, it has to be so simple that it is one or two clicks to do the installation. What will kill their marketplace, and you guys brought this up, is how do the customers approach this?

Is it that I am already in the NVIDIA ecosystem and I want to install this piece of software, a la AWS Marketplace? I'm already here and I'm just going to pick and choose the parts that I want off of the shelf. Or am I looking for something very specific?

And how much is NVIDIA charging meTo be in that marketplace, because we know what the answer is. 30% is too much. 30% cut of everything you get is far too much.

That's what Apple gets, and we have seen massive amounts of pushback because it's not just 30% of the app cost, it's 30% of the subscription cost. It's 30% of any in-app purchases. And in 2009, 2010, there was no other option because Apple created that market.

Now, I think what Nvidia is going to have to do is they're going to have to give back a lot of percentage points for people to want to list themselves in some kind of a marketplace. And it's going to have to be valuable enough to those providers, and they're going to have to see enough attach rate in order to want to list themselves there. Because if you start buying your software directly from Nvidia, then that means that I don't need to open a storefront anywhere else unless I'm selling to the 18 people who are still running Tensor Cores.

There's one thing that Jensen announced or said that I strongly disagreed with, and he made it sound like, "We've got it sewed up. We've figured out how to secure this for the enterprise because we have the AI toolkit. " The third component was actually NeMo Claw.

Yeah, NeMo Claw. And they have NeMo Trime, the model. And that's the typical response we've lived with for years, which is, "Well, let's talk about when it gets into production.

" Well, the world has changed. We use AI to build AI. We use agents to build AI.

So guess what? They're just as vulnerable in the development, in the planning, in the product ideation, in the CICD process, all the way. The whole pipeline is now AI.

That has to be secured. It has to use the same kind of controls, guardrails, governance, all of that throughout the life cycle. It's why I created this concept of observability native.

It's got seven pillars. One of them is that, it's throughout the life cycle. And it also examines, you have to understand the agent's behavior about what it was trying to do, how it decided to do it, what it did, and what was the consequence.

" So I'll give you the prompt to start to do it, and we'll send out the link to be able to download that. I'll have a paper as well describing what it does. I did this to my own environment, dropped it in, and the answer back was, "Well, we do one, three, and five pretty well.

Got those covered. " What do we do for those? Okay, great.

Now, it also implemented an instrumented OTEL, Alan, in the beginning, in the development environment. Oh my God, what a difference those are the things we've talked about, securing the supply chain, a huge step. So that's what prompted me, no pun intended, to say, "Look, I'm going to give this away.

" And of course, they'll improve it. So very cool. It's very cool.

Guys, in all of the talk and all of we've kind of danced on the head of this pin. " Right? And that seems to be their go-to-market.

We're going to give it to you securely. They didn't talk about walled gardens à la Apple, though it may in fact be a walled garden. We'll have to see how this plays out.

And by grabbing the partners they did, they grabbed, by the throat, the security aspect or the security angle of using OpenClaw, which much like when cloud first came out, security was one of the biggest inhibitors to adoption. And we're hearing the same thing around OpenClaw, and Nvidia is looking to own that. So you got to just applaud them and say, "Hey, well done.

" Ditto. Well, I agree that agentic is going to happen, and it needs to be secured, just like LLMs are going to happen and dominate, and they need to be secured. The problem that a company like Nvidia faces is that even though they are the behemoth, the mega behemoth in this market, that we've seen some mega behemoths before.

Ask Lit Bhutan how it feels to miss skating where the puck is going, because that's Nvidia's problem now, is can I pivot fast enough to go where the next advance in AI is? Because while the rapid development of using AI to build AI and coming up with these new paradigms, and it's the race for AGI, is very important, and we know how important it is to secure that. You cannot bet $5 trillion of company on making the wrong decision, because the first mistake will not be a huge problem, but mistakes beget mistakes, and that's how Intel ended up where they are, which, if I'm not mistaken, doesn't Nvidia own 5% of Intel right now?

That's the situation that you find yourself in. One day you are making graphics cards for nerds that want to run Quake a little bit prettier, and the next thing you know, you are on the cusp of a massive revolution, and if you make the wrong decisionand freeze or don't freeze, you can find yourself in a world of hurt. So I think NVIDIA needs to be seen as the dominant player in those markets, whether it's LLMs or agents or whatever comes next.

But they have to play speed versus certainty. I need to be fast to react to these things, but I need to make sure that it's going to be a thing if I'm going to pour millions of dollars of resources in there. Because remember, the other thing that you have to understand, this is a problem that we started hearing about a couple of years ago, you have a company full of multimillionaires doing this job for you.

Every one of the people that owns NVIDIA stock is a multimillionaire right now. " And that's one of the things that they have to be careful of, is the potential for brain drain where someone's going to be, "I'm going to spin out. " Because obviously, they're too big to understand the real problem around here.

Agreed. You know what, though? They've bought themselves $5 trillion buys you a lot of things, Tom.

And - Buys you a lot of friends, too. Yeah, it buys you a lot of friends. You're the popular guy, big man on campus.

It does also buy you a little bit of breathing room, let's call it, that maybe the guy chasing open AI, the Avis of the AI world, if you will, who has to try harder, doesn't have. But look, and I wasn't there. I've spoken to Daniel Newman, who was there, and we had a chat about it.

From everything I see, there was no fumble here. No fumble here. This was damn near flawless, unless two-hour plus keynotes are not your thing, but- It was Disneyland.

There was something for everybody. You wanted robots. You wanted one of the Disney characters.

Even where the keynote was held, the keynote was held in the arena where the San Jose Sharks play. This wasn't your average keynote stage. Conference center, yeah.

It's like where rock stars play, where rock bands play kind of thing. So look, at least for right now, and as far as I could see, which is 12 months maybe at best, six months, they are the reigning past, present, and possibly future heavyweight champions of the world. It's theirs to lose, right?

Get ready to rumble. But security has a way of coming up and biting you in the butt, too. Don't forget that.

If you start seeing a lot of security things, especially even on the NVIDIA stack, it doesn't prove to be appreciably more secure than some of the other open source kind of methods out there. Every plan is great until you get punched in the face, right? Yep.

Until you get punched in the face for the first time, you never know how that stock might going to react, right? And considering who that quote came from, Mitch, we are always on the lookout for the Buster Douglass of the world. Oh, yeah.

So Mitch, what are some things that you're working on that people should check out? Well, with coming out of RSAC, it's a ton of follow-up that I have because people had so many announcements. But announcements aren't as important to me as much as where it signals where people are going, and control planes, governance, agent behavior, agent accountability, and trust, AI trust, is the big thing.

" And I'm going to be launching a framework around agent control plane so you can kind of see where are the pieces, what is Jensen talking about? Where does that fit into this? When I'm talking to this vendor, how does that work?

Give you kind of a reference architecture to talk about agents and agent control planes. Alan, what about you? You know, Tom, no rest for the weary.

I put in a few miles myself working at RSA, though. I do have to admit, I did go up to Napa Valley after RSA for a weekend and did some wining, so drinking, not whining, like Karen was. A little bit of both.

But anyway, I'm home for about two weeks, keeping Techstrong perking, and then I'm heading to Prague, where I'll be there at SUSECON. I think, Mitchell, you're going to be with me. I will be there, too.

Yes, I will be there. I'm really looking forward. I think the whole digital sovereignty thing is going to blow up, because this is in Europe.

SUSE is a big proponent of it. I'm really looking forward to how that plays out. It's going to be interesting, especially given world events and everything.

So I'm looking forward to that, Tom. But we're also, guys, we are eating our own dog food here at Techstrong in terms of adopting AI and agentics and everything else. And I am so damn excited by how fast I'd want us to move.

And the team, Mitch knows this because Mitchell was our CTO for 12 years, 10 years, and he's still intimately involved in everything we do at Techstrong. And we are ready to run and run fast, so- You are ... watch out.

You're becoming an AI native agentic company. We are an AI native tech media company. Go figure.

Yeah. So that's what's going on here, my friend. And I'll point out that we have just posted a lot of videos from RSAC, especially around data protection with companies like Veeam, Object First, and Commvault.

com or the Techfield Day YouTube channel to check that out. We'd love it if you'd leave us a like and maybe a comment because we love to hear those. But the other thing we love is that we love that you listened to this episode of the Security Boulevard Podcast.

And of course, if you enjoyed listening to Alan and I rant and rave about things, and Mitch trying to figure out who to agree with, please subscribe on YouTube if you're not already, or in your favorite podcast application of choice so you don't miss any episodes when you're out riding your bike or mowing your yard. We would love it if you would share this episode with others because that's what helps the show grow. A good recommendation from a friend is the best kind of podcast recommendation there is.

com and the Futurum Group. com, the Techstrong TV website, or if you haven't already downloaded the Techstrong TV app, you should because I am watching your downloads. I can do that.

I'm a security guy. It's available on Apple TV, Roku, or pretty much any smart device. If you are still on social media, may God have mercy on your soul, but you can check us out over on X, Twitter, and LinkedIn.

Just look for Security Boulevard. That's Security B-L-V-D because vowels are for losers. And we have a lot more content out there for you.

Thank you very much for tuning in, and we'll check you out next week.