Microsoft Pushes AI, Trump Resets Rules and Cisco Wants Control

Hey. Good morning, good afternoon, everyone. Welcome to techstrong® Gang.

I don't know if you caught that quick intro of the techstrong® Gang logo in New York Knicks blue and orange, but I got to tell you, those of us from the city area are really excited to see our Knicks back in the finals. What would make me really excited is to see them win the finals this year for the first time since I was 13. And it's been a long wait, and I really feel this is our year, so go New York, go Knicks.

We've got a lot to talk about on techstrong® Gang today. I'm not back in Boca or in the studios. I'm actually out in Las Vegas at Cisco Live, hearing all of the latest and greatest from Cisco, along with actually Tom Hollingsworth from our tech field day team, who's also here at Cisco Live doing a tech field day.

We're also joined today by the one and only Andy Mann. Our man, well, he's our man about town everywhere, John Swartz, and the eternal optimist, hoping this is the year for the Knicks and the Yankees, Mike Szul. Batting left-handed, throwing right.

All right. So gentlemen, and it's all gentlemen today, unfortunately, but we've got a lot going on. com.

And this is the first of the next generation techstrong® sites we're doing. It's the first one we've done since we've joined up with the Futurum Group. We're relying on Futurum Group.

They bring a lot of semi experience and content and depth to it, so that'll really help the site we think. Next Monday, we'll have a special edition techstrong® Gang where we're actually going to officially launch and go into this stuff. But that's next Monday.

We've got so much going on in the world today. It's conference season. We're all scattered to the four winds here.

Everyone I know is at one conference or another. Microsoft had some big announcements that they're trying to redefine or clean up these AI models or make a better model. What do you think?

There you go. Well, the annual Microsoft Build hootenanny was this week, and that's when all the folks come together who are the developer community. And Microsoft, it is the house the developers ultimately built, so they're always showing off what they did here.

And one of the top list items was this family of new AI models. There's six of them, and there's going to be more of them, and they are saying that they are, quote unquote, "clean AI models," which says something about the other AI models that are already out there. So what they're really on about, though, is that these AI models were trained using data that's clean and licensed, and it doesn't have a lot of random stuff in it or stuff that might show up to be belonging to somebody else or might be used to, I don't know, construct some sort of biological weapon.

And I can't help but wonder, you know what? This is a good thing. " Because it just seems logical, and it seems like the thing to do, and I'm hoping other people follow suit.

However, before we applaud Microsoft for all those good things it is doing, we should remember that if it wasn't for Microsoft, we wouldn't have open AI being so pervasive, and we wouldn't have so many of these, what shall we call them, not clean models running around out there. So, on the one hand, they giveth, and on the other hand, they kind of already tooketh. So Andy, I think Microsoft, it's interesting times, and they've already previously had a, I don't know, a separation agreement with OpenAI.

Is that one way to look at this thing, and is this the natural progression thereof? What do you think? Yeah, look, I think you're spot on there.

And there's a lot to like here, right? Especially from an IT leader, from a CTO perspective, right? The ability to build these new models, to be able to create and customize and build, that's very awesome.

And there's a bunch of things here because Microsoft gets enterprises, right? They understand how business needs to work. So things like per session sandboxing, like for the untrusted code.

Being able to have a custom device that builds the tech stack in for development, the Surface, what are they calling it? The RTX Spark or something. All the GitHub Copilot stuff that they're doing as well, providing the application development environment, the AI agents you need.

Look, I am neck deep in agentic AI building code for me right now, let me tell you. And this is actually really cool. Our friend of the show, Mitch Ashley, talks about in the article there about creating a composable integration plane.

Look, I love that. Oh, by the way, this is also a little bit reminiscent of their old SCO Linux non-infringement guarantees. Right?

The idea that our AI model is clean. You can trust it. You won't get sued if you use our AI model.

Because I think the whole ownership and copyright and stuff is still not fully settled. To your point, they created this Hydra, for want of a better word, and we've gone through a couple of court cases, but for the most part, this is not settled law. So if I'm an enterprise buyer, I'm still wondering what my exposure is as a CTO, because to a large degree, this horse has already bolted.

Right? But yeah, I love the idea of pushing that boundary on creation, customization, building, don't just consume. And for Microsoft, they get control of this tech stack.

Like you said, it's their all developers, developers, developers, right? It's what they've always been about. Get the developers.

It's like banks with piggy banks, right? You get the developers early, then you get the rest of the enterprise. So I think there's a lot of goodness here, but I think you're right.

It's a little self-serving, but then again, everyone's got to make a buck Mm-hmm. Alan, do you think that business and IT leaders sent a clear message to Microsoft about all this stuff and said, "Hey, we're not going to play ball with these AI models if we don't have some way of knowing what happens to our content," and maybe they're finally listening? Well, Mike, I got a few things I got to come to on this.

First of all, this hootenanny. I don't remember hearing my hootenanny from Colestomo or Sonny in A Bronx Tale or anything. Where does this hootenanny come from?

Second of all- Well, as it happens, I was at an Irish music festival recently, and well, that was a hootenanny, so there you have it. All right. There we go.

So now we have the hootenanny. You had hootenanny on the brain. Uh-huh.

Next up, we can't blame Microsoft for OpenAI. I know in the tech industry, we've been blaming Microsoft for everything for 30 years. But had Microsoft not financed OpenAI and helped bring that, AI was going to happen.

It was inevitable. The research was moving it. Did Microsoft maybe accelerate it a little bit?

Yeah, but I think we'll be better off for it in the long run. So I'm not here to bury Microsoft or to praise them, but I do think they read a signal in the market that I think others have read, too, which is how much can we rely on these frontier models- Mm-hmm ... without getting ourselves into hot water from a copyright issue, from an IP issue?

And its twin brother of that, which is how much can I tell these frontier models without the whole world sharing in that, too? Right? Those are the two, when you look at these frontier models, those are the two pieces.

They're different sides of the same coin. The information it already has may not be kosher, and the information I'm giving it may slip out the back door to ... And so I think Microsoft saw this as an opportunity, but I think the way we're characterizing it is wrong.

If you read what Satya said in the article, this is a key. This is the path for people to build your own, to roll your own models. So now you could take one of these six or seven models that they swear up and down are clean and start that as the kernel of your own model, not worrying about having to share that in the world.

Right? What Elon's going to do with it, or Slippery Sam or any of these people, and you could build something you could rely on here. And I think if that's really the case, I think Microsoft's onto something.

Right? It will be copied. There'll be others, but Microsoft's onto something, and they seem to have a lead on it.

Tom, let's get Tom in here for a second because that's the next point, right? So won't Anthropic and OpenAI come up with a set of clean models themselves, and they'll just go in and distill that from their dirty models? I don't know.

I don't have another term for it, but we'll come up with something. And this will be the way the world goes. There's going to be these broad, general-purpose AI models for consumers that are going to be full of all kinds of data that we don't know where it came from, and then there's going to be these, quote-unquote, clean AI models that are used by enterprises.

What do you think? I think you're right, but I think the problem is going to come down to how much trust you have in those models, no matter how clean they are. Because one of the things that we see, especially at the enterprise scale, is they're very hesitant to change what they're doing just because something purports to be a better option, and especially in highly regulated industries.

One of my favorite points of discussion is talking to people about why there are still ashtrays on airplanes of a certain vintage. " Well, it's because the airframe was certified to use the parts that were on there, and we can't just randomly switch those parts until we rebuild everything from the ground up. And I think that if OpenAI misses out here on this idea of clean versus unclean models, then enterprises are going to basically pass them by in a cycle before we know it.

John. John- Hey, Mike. Yeah, can I mention something that- Yeah.

Go ahead ... just kind of keeps going in the back, and I think you're going to bring it up but... Well, first of all, I'm going to give props to Microsoft.

Okay. So organization, I think Alan mentioned this, organizations, they have the option now, the ability to move from just merely consuming the frontier AI models to developing them or participating in their development and customization. Okay.

That's good. All props to Microsoft, but it keeps coming back to me, and I keep thinking about the history of Windows. And tell me if I'm crazy, Mike, but is Microsoft kind of quietly building this stack that you can't easily leave?

Is it like deja vu all over again, where it's Windows for AI, or am I maybe thinking too much about that, and maybe this is just some agent control plan? No. It gets better than that, John.

What if OpenAI is essentially the equivalent of Microsoft to, what was Microsoft to IBM, and maybe- Yes ... these new models are kind of like the OS/2 of AI models? I don't know.

I'm starting to think... You know what? That's so funny you mention that because I'm thinking this, we're seeing this history repeat itself, just in a different form and at a different speed and in different players.

I actually kind of see that happening. Yeah. Maybe.

" What do you think? I don't think it's coincidence at all, but it is also just part of the natural evolution of this market. But here's the thing.

I wrote a bunch of articles last week around data center build-outs. Mm-hmm. And look, it's AI that's driving these data center build-outs.

But these data center build-outs transcend AI, transcend technology. They are civilization type of events that require gargantuan engineering, crazy amounts of money. So the pot of gold that we're dealing with here is enough to even make a tech bro egotistical multi-billionaire salivate, including Satya and the Microsoft people.

Mm-hmm. Now, I would also expect, though, anybody who provides a cloud service is going to follow the Microsoft lead here. So we're probably going to see AWS, Google, and then Oracle, and anybody else is going to suddenly start popping up with similar models.

What do you think, John? I think you're right. I think that's inevitable.

And I'm glad you brought up the IPO angle, too, because one thing, last, only a couple of days ago, we were talking about the amount of money that's being thrown into AI infrastructure among the four major players was like $600 billion for this year. Now it's up to $700 billion. And Anthropic is, I think we covered this earlier in the week, I wasn't on the show, but they filed to go public.

OpenAI is coming quickly. So this is all coalescing. There's nothing coincidental about all of these announcements and their timing.

It almost seems as if they're coordinated. Now, I'm not a conspiratorial thinker. Well, sometimes I am.

But it's all starting to coalesce in a certain way that benefits the biggest and the largest players. Well, those people in that Mag 7 are very jealous of their territory, and I don't think they're just going to open the door and roll out the red carpet to OpenAI and Anthropic. They're going to protect their territory.

Yeah. And you can't blame them either for that, by the way. Mm-hmm.

And I suspect Apple will be in this mix as well, right? Mm-hmm. And maybe Meta too.

They're all going to launch something of their own. I think Apple's sitting back as a stalking horse, kind of waiting to pounce. They always do, yeah.

As we've always said ad nauseam, that's what they do. Yeah. They let everybody else establish a market or a beachhead.

They don't squander their resources, and then they move in and offer their alternative and their rabid fan base. And they're good at this, and then they establish themselves as the true leader. But let me say something else, though.

We've been talking about this for over a year on here, that not every query, not every AI prompt, not every task that you're using AI for, requires Claude or Opus or Codex or any of that. If AI is going to be economically viable, we have got to move to a place where we're using homegrown models, small language models, cheaper models, and this is a step in that direction. Yeah.

And I think- I'm going to disagree with something that Tom said, and he can argue back with me about it, but I don't think that we are as tightly coupled to these LLMs as you made it out to be. I think that they are all one API call away, and I can swap them in and out as I need to within an application, and ipso facto, they are disposable. But that's my opinion.

This is this whole harness kind of thing. What's my harness for the different agents and models I'm using? And I think another thing that's really cool about what they're doing here is actually, it's not just about getting cost control.

Absolutely important. We see that Microsoft all of a sudden starting to charge for every token, and my Reddit is full of people complaining that they're out of tokens already. So there's cost control, absolutely, but there's also that IP control, and also just the effectiveness of AI to know my business.

I've been selling software into companies for a long while. Every single data center will say, "Oh, no, I'm different. You don't know my business.

You don't know my architecture. " If I can get an AI that knows my environment, knows my software, knows my business, knows who my customers are, knows how I make money, then I'm going to actually have competitive advantage out of that. So I think it's not just the cost control and the ownership and a bunch of other things, although it is all that.

I think it's also an aspect of getting control over enterprise data within the enterprise in a way that makes my AI really an employee, not just another consultant. Well- So- ... back when Alan and I were much younger and you ran out of tokens in New York City, you hopped a turnstile.

Yeah. I guess the turnstile in the age of AI is open source. I think a lot of people are going to find a new way to hop that thing, and they're going to go to an open source model and go, "Screw it.

" Yeah. Under the turnstile, not over. Use that in a story.

That's a good phrase. Guys, but wait, we could count on the government maybe to help us enforce this on AI, and legalities and so forth. And that brings us to our next one.

The government, we've got a new executive order that these AI companies have to do a pinky promise- ... if they're going to release something. Mike, what's this nonsense about?

So yeah, the United States government is saying that they would like people to voluntarily submit their AI models for review to the United States government to make sure that, well, they're not going to be giving away nuclear secrets or whatever else that might be leading to something like, I don't know, a weapon of mass destruction or whatever it is. But John, here's the backwards thinking about it. " They're self-regulating.

" It's basically a tell, right? Yeah, this is big tech's version or their interpretation of self-regulation. Trust us.

This sounds almost like a rubber stamp program. Right? So Trump, this has been going on for a while, but he signed this executive order establishing this framework for the leading models systems from specifically OpenAI, Anthropic, Google, before they're publicly released.

And if we remember, this is a 30-day review cycle, a 30-day review process versus 90 days, which was an earlier version, which David Sacks, the former AI czar, objected to. He wants to rush this for national security purposes because he's worried about China and about our ability to compete with China. So, it's absurd.

This idea of having any type of government agency review something in 30 days. How do you do a thorough review in 30 days? It just, to me, sounds like window dressing for rubber stamping and a ceremonial attaboy pat on the back.

Cutting through all the b******t, that's what I think this is. But wait, can't we use AI to review the AI in 30 days? Because that's the whole point of this exercise.

That might pose too many difficulties for the government because they want to probably rush this through while they can, and so does big tech. The oligarchy is moving as fast as they possibly can. They're breaking everything in their way, and this plays really well into their plan.

You know- All right ... I have a different conspiracy theory. Because you watch this crap long enough, and I think you start to get a little flavor of how this administration works.

Right? And I think what happened here is someone from the rump CISA that's left, whatever's left of CISA and our national cybersecurity infrastructure people said, "Look, we have got to get in front of these AI models. We got to make sure that people can't build a nuclear bomb from them.

We've got to make sure that they're not telling people to commit suicide. We've got to at least look like we're in charge here, and we recognize the issue. We need a review process.

We need to say these things need to be reviewed before they're put out in the public. That guy from Anthropic with Mythos was onto something. " And our administration says, "What a great idea.

What a great idea. We're in charge of this AI. " And then Sam, Darren, Elon, one of them call up their buddy sitting over there behind the- It was- ...

desk in this case. In this case, it was Elon, for sure. Yeah.

And called up and said, "What are you doing? " "Oh, no, not crazy at all. " So they cut the baby in half.

They took what was a well-meaning thing that we should review these AI models ala Mythos and Glasswing, and cut that baby in half and said, "Well, if you let us do it and if we pinky promise to give it to us, but you're not under any obligation. " "That's good, Donnie. " And off they go.

So this is unfortunately what happened, and this is how government works. There's compromises, give and take. Mm.

And what comes out is often a bastardized nonsense versus what the original intent was. And I think this is a joke. Well, there are other governments around the world, so will they implement something- Yes ...

that's a little stricter, and maybe EU will take the lead on this thing, or who knows? Even our friends in Australia. No problem.

Yeah. They will. Yeah, but- I think other governments are going to do something a little bit less corrupt, I think.

Look, this is the trust me, bro executive order again. I love that phrase, John. By the way, we already had an executive order on AI.

We had a real policy document. It was multi pages. It actually set up actual policy and implementation through federal departments, like the boring stuff of government, which is what politics should be.

Of course, we rescinded that January 20th, 2025. I wonder why. Look, this is so toothless.

I'm reading from the executive order here. " Right? " Look, that's it.

That's it. It's pointless. It is confirmed.

But it's not completely meaningless. This is not just another performative exercise for the TikTok generation, although it's also that. Right?

Read deeply. There's a bunch of things where they establish departments and so forth, but this is the kicker, team. There's an AI cybersecurity clearinghouse.

Once again, I'm going to read from the executive order. " That's it, team, for me. There's the griftThat's it.

Mm-hmm. Find the money and establish this BS AI security clearing house. Yeah.

Use that to pressure AI people to bend the knee and to give them our money so to get them in favor. I don't know. Look, maybe I've been living in this- No, you're totally right, Andy.

" I think you're totally right. I think we need a blue ribbon committee on that one. I say no taxation without AI representation.

Where do we get some shares in these companies because they're using our tax dollars? Well, Bernie Sanders actually is trying to do that. That's an idea of his, and it's not going to gain any traction, but he has brought that up.

I actually got a note about that a couple of days ago, so. Yeah. He wants to nationalize the AI industry and take 50% ownership for the government and do it as a sovereign wealth fund, much like, I think it was, Denmark owns Nokia and the rubber factory, and now it's all oil money and anything.

But yeah, to establish a fund for UBI to pay for all the people getting put out of work by AI, it's sort of crazy. It has no chance of happening in this country, but it's a very provocative idea, I think, John. But you know what?

To be fair, Sam Altman said something similar a couple of weeks ago, maybe two months ago, where he said, "We need to set aside a certain amount of revenue from AI- Yes. Right ... " Yeah.

I'm glad you brought that up. That deserves mention. Yeah.

He did bring that up. Yes. Hey, Alan, is he going to be taking that out of the profits from his company?

Because if he is- The profits ... we're going to be in a lot of trouble. We will be.

Well, but don't forget, it's a not-for-profit foundation with a for-profit entity kind of sucked in there. But I'm not saying... And I'm not a huge Sam Altman fan.

I know John, he's one of your buds. But the fact is- He's as well-liked as Mark Zuckerberg and Larry Ellison are in Silicon Valley. Absolutely.

But the fact is, guys, here's the deal. I'm not a Bernie bro, and I'm not a socialist, and I'm not for this plan. I wasn't for what Sam said.

But what I do take notice of is not when politicians and not when snake oil salesmen, but when real AI people, the people who are the scientists, the sociologists who are looking at this, the experts tell us that we've got to do something to short-circuit the disruption that this could make into too many people's lives. We've got to do something, or you just going to wait? I think it's a serious question, though, that's being brought up here or implied.

Is this AI stuff just too dangerous to be left in the hands of a bunch of for-profit maniacs who don't care what happens to the impact that they have on society? Oh, gee. Ask the guys who run the media.

Wait, wait. But we go back to, was steel too valuable to leave to Carnegie? Was oil too valuable to leave to Rockefeller?

Was Wall Street and banking too valuable to leave- And didn't we create antitrust laws for the specific purpose? Yes, we did. And I think that's exactly what we're going to have here.

It may take 10 years because now it took us 20 or 30 in the last go around. It may take five, 10 years in this crunch time, but eventually, you're going to have some sort of anti-monopoly, some sort of trust buster. You have to.

Hey, John, can you put a prompt in to ask ChatGPT what it thinks we ought to do about all this? What do you think? That's interesting, though.

I think you're right, Alan. I think it's going to happen. This is going to take years, and it's going to- Yeah.

There's going to be damage that's already going to be done, but that hasn't changed- But that level of carnage is going to be what it takes to get people to- Exactly ... do something. And I think about younger people, and I think about anyone who's looking at AI now.

They're engaged, and they see all the harmful effects, and they are actually vocal about it. And, so I actually am pretty optimistic that we're going to address it within a few years, within five. All right.

There you go. We'll find out. There are clearly political campaigns wrapped around it, and I know Alan wrote about this, but here in New York, there's now millions, maybe even a billion being spent on one congressional district because one guy wants AI regulation, and the people supporting him belong to a PAC that belonged to Anthropic, and the people spending money against them belong to a PAC that's led by, well, OpenAI.

Interesting times, don't you think? Yeah. Well, there was in the governor's race, really quickly, the results, they're not final, but Tom Steyer, the billionaire who's running against big tech, is in third place.

He's not qualifying for the runoff, and Steve Hilton, who's a Republican, has the most votes. And the conspiracy theory out here is that the big tech paper tiger or whatever you want to call it, stalking horse, Matt Mahan, was put in to siphon votes away from Steyer and that big tech was behind Steyer perhaps finishing third and not being part of a runoff. He's a threat to big tech.

That's one of the dirtiest political tricks in the book since time began. Yep. The people have spoken.

Let's move along. As I mentioned at the top of the show, Tom and I are out here at Las Vegas. We're at Cisco Live.

Tom, they told me somewhere between 20 and 30,000 people. I don't know if it's that many people, but there are a lot of people here. It looks busy, but it feels light.

I've talked to a lot of people who said that they had a lot of traffic on the World of Solutions Expo floor yesterday morning, but overall, things feel a little more spread out. One thing I did notice, there's a lot of new people here, a lot of first-time attendees, a lot of folks who've never been to Cisco Live before, and a lot of people that I personally know that have been coming for years said that they were taking this year off. And I don't know if that's just because work has been crazy or if it's a sign of something bigger going on.

I just don't want to go to Las Vegas. I don't know. It could be.

Well, it is 100- Too many shows in Las Vegas ... it's 104 here. But it's also a-- Look, there's too many conferences going on at the same time.

Yes. Right? You've got Snowflake, you've got the Microsoft one, you've got Cisco.

That's just this week. Last week was something else. So in San Francisco, there was Build, I believe.

I think Mitch was out here for that. Yes. There was Upload, which I have no idea what the hell that is.

Bloomberg has their conference starting today, and then there's a fourth one that I keep getting emails for that I've never heard of. So there are at least four of them here. Yeah.

There's vague-- I don't remember a conference season this intense. I think they're going to differentiate themselves by one of them is going to start calling themselves a hootenanny. You watch me.

Isn't that right? I will sign up for Hootenanny Conference. And honestly, I would've thought 20,000 odd would be quite small for Cisco Live.

No, I think they said 20 to 30. I was- Can I come back to what actually Cisco's talking about before I move on from here? Okay.

All right. So Tom, as I read all the coverage, it kind of seemed to me that Cisco's trying to make a play for the one control plane to rule them all in the age of AI, and this is how they're going to be relevant going forward. And I'm sure they're not the only people with that idea, but what are the odds on Cisco winning that argument, Tom?

What do you think? I feel like they are positioning themselves to be in a better shape to do that. But like you said, they're kind of playing from the back of the ice, so to speak.

There are other companies out there, HPE, through their acquisition of Aruba and Juniper, is probably in a little bit better position, especially if you go look at some of the magic quadrants that have come out recently. Trust me, that's gotten a lot of play around here. A lot of the things that they're talking about around this new cloud control idea feel very reminiscent of things like Aruba Central.

We're going to hear more about that actually, as we're live right now, they're talking about it in the keynote, the technical keynote, and going over some of the big details. There's still some things that are yet to be released. We actually recorded an embargoed session yesterday that we're going to post later, with some of the announcements that are coming out.

But I feel like Cisco is starting to see that people want to have a unified system for doing all of this management, whether it's the network underlay, whether it's the overlay on top of it that enables applications to work, whether it's the security that ties it all together. Because that's another thing that I feel has been a huge undercurrent of everything going on here is I think Mythos and Glasswing have a lot of people on the production side very scared about what AI can do to their networks. And so a lot of the tooling that Cisco's been talking about on and off for the last year really came to the forefront here because some of the things that are being baked into NX-OS right now that will allow you not only to detect these things, but to create mitigations while you're waiting to patch things.

That's been getting a lot of conversation from the on-the-ground people here. So I have a little different take on it. I put up three articles.

Two are on Techstrong IT, and one is on our new Techstrong Semi site around it. I've been in technology 30-plus years, maybe 35 years now. And Cisco's been a constant all through those years, to the point where there were times, especially early in the commercial internet era, where I thought they were a monopoly, right?

Juniper finally gave them a little bit of a run for the money, but let's be honest, Juniper was never Cisco. There's a reason why Juniper got bought. There's a reason why Aruba got bought.

There's a reason why all of these contenders became pretenders, and Cisco is still the network king. But Cisco doesn't jump into a market haphazardly. Cisco, Dell does this, too.

They wait till a market is more clearly defined. They wait till it's a big enough pot of gold to jump into. And I think this Cisco Live signifies that they're saying, "Hey, this AI market is mature enough, big enough.

" But what's interesting about this plan is for a long time, we only saw the tip of the iceberg. We only saw the top of the mountain. We didn't see the mountain coming up out of the water.

That mountain starts on custom silicon. Right? Cisco's always had custom ASICs in their stuff.

Who gave a s**t, to tell you the truth? Because none of us cared about silicon back then. But silicon's sexy now, and that's why we have Techstrong Semi.

And so this whole mountain is built on this Cisco Silicon One kind of-... family of processes. They got almost a half a trillion processors or transistors, whatever, on one chip.

A half a trillion. So they've got the hardware, they've got the chip, the silicon for this next era of AI networking, and then they're building the control plane, the harness on top of it. The cloud control is part of it, but cloud control also has a lot of the security and the Splunk stuff and observability built into it.

Look, they're Cisco. They want to own the network. They now call that network the AI infrastructure.

Right? So NVIDIA may have your GPUs and your servers and all that, but that information has to move from one data center to another, from one server to another, from one place to another, and Cisco wants to own that. They are going to butt heads with NVIDIA on that because NVIDIA has their eyes on that too.

I'm reading it slightly differently because to me, it's becoming a battle between Cisco, HPE, which now owns Juniper, and Dell, and they're all pulling together server, networking, storage infrastructure and selling that as a single motion. And if you look at their numbers for HPE and Dell especially, and Cisco, the amount of infrastructure gear that appears to be going into on-premises IT environments is off the frigging charts. And some of that may be a response to the fact that I don't want to put all my data in the cloud anymore in the age of AI.

And I think that Cisco, and whether it's Dell or HPE, they all have an opportunity to be the proverbial tail that may wag the dog here because people are going to say, "You know what else I don't want in the cloud is the control plane. " And those three might be the better option to do that. Just saying.

Yeah. Look, I think that's actually a really strong point, and I know personally and from my business, that on-premises data center and infrastructure still matters, and it absolutely matters for the AI infrastructure. But I tell you...

Oh, by the way, Alan, I always cared about chips because mainframes matter. Anyway, but look, this to me is, how do I put this carefully and gently? Absolutely absurd.

Look, Cisco, you're right, Alan. Cisco will own networking. They'll talk about it as ops.

But look, they are awful at ops. They're great at net ops. They've bought their way into sec ops.

They're still awful at IT ops. This idea that there's one pane of glass, that there's one control plane is... Look, how many times do we have to go through this absurd argument that we're going to build the OS for the...

Look, remember the cloud OS? Remember the virtual OS? Right?

This is great marketing, and I praise the marketing team that came up with the PR for this because it rings really nicely. It's agentic ops. It's the holy grail of we're going to get rid of all IT ops, right?

This is the holy grail of no ops again, and yet this is not going to happen. They'll get really good at the network. I think you're right there, Alan.

That's where huge amount of money to be made, a huge amount of capability, specialization, the chipsets, all of this stuff, really positive. " No. But I will tell you who to really watch here, and I wrote this article as well, I think it was published, called PAX NVIDIA.

Whether it's HPE, or Cisco, or Dell, or any other conference that I've gone to in the last eight or nine months, the marquee star, the marquee partner, the company that gets to go up on the keynote stage with the CEO- Mm-hmm ... even if it's some junior guy there or junior person they're walking out, is the NVIDIA people. They are omnipresent.

$5 trillion goes a long way, but they have huge presence at every single one- Yeah ... of these shows, including Cisco, and the message you get is if you want to be cool in AI, you got to be doing something with NVIDIA. Yeah.

And just by the way, when did people start calling it NVIDIA? I'm sorry. All the mainstream press- I haven't seen it.

Is that what they're calling it? I haven't seen that. NVIDIA.

Yeah. Yeah. They're pronouncing it NVIDIA, and it drives me nuts.

Same here, Andy. I can tell when you got into tech by whether or not you realize that this company used to make graphics cards, and the N is pronounced E-N, not N. But it's a lot of up and coming people who think that they've got this AI thing figured out, and they think it's just a cool, trendy name that doesn't have a whole lot of vowels in it, and it drives me up the wall because me being a grammar supremacist, I'm like, "No, no, no, no.

" Yeah. Branding 101. And it's a really good tell for what people know about this industry.

It's a good point, Tom. This is a European debate, right? Adidas versus Adidas.

Is that where this is going to come out? Wow. That's a European thing.

But you know what? Let's continue in this NVIDIA thing for a moment. I wrote another, or it was in this PAX article.

What a balancing act they're playing here. They cozy up to the US government. They buy 4% of Intel.

" Yeah. And gets on the plane. They're promising this.

They're trying to do deals in China with that, though they haven't sold any of these H200s there. And now they just committed to $150 billion annually to Taiwan for their headquarter and a whole bunch of thing there. So are they building chips in Taiwan?

Are they building-- You can't be everybody's friend all the time. We talked a little bit about that last week, Alan. We talked about how this guy, Jensen, I think he's the best tech CEO, if not the best CEO, period, in the world, because he's played this perfectly.

He hasn't bowed down or gone down on his knees. He's maintained some sort of integrity while playing hardball, but presenting to the public this kind of omnipotent, good guy image. Everywhere.

And everyone wants to partner with him. And I think- Wait, Tom has something to say. But he gives the same speech everywhere he goes.

Of course he does. " And then next week, he's probably going to be in Liechtenstein saying the same damn thing. No, I think it was Luxembourg.

But Tom, what do you got? I think, to bring this back to where we're at right now, I agree, NVIDIA is the king of the hill right now when it comes to AI accelerators, when it comes to the development that's happening there. And yet, when it came time for them to start talking about the network, even though they have SpectrumX, even though they have all the stuff happening with Mellanox, what did they do?

" And Cisco's got two very different approaches to this. They have Silicon One on the one hand, and they do have the SpectrumX on the other. " In the networking space, it's, "Nobody ever got fired for buying Cisco," because they still are the 850-pound gorilla.

They are the clear number one in the space, and they cannot be dislodged from a lot of environments- Right ... by an upstart trying to sell you a little bit of networking with all of these AI accelerators. You know what's remarkable is this company that is considered, quite frankly, a boring company for many people, they are the one company that is a borderline monopoly player that still is dominant as they were 30, 40 years ago.

Yeah. It's remarkable. And that's the truth.

It is. It just doesn't change. Interesting, Tom, I don't know if you're going to be able to go to this because I know you have a Tech Field Day coming up, but I'm heading down after this.

There's a special session with G2, chief product officer at Cisco, president, with OpenAI. It seems they built Codex in. And Cisco, I heard from Cisco people here, they're using Codex for a lot of their software development.

Not Anthropic. Not Claude. So it's going to be interesting.

Bringing it back to our opening discussion on Microsoft and OpenAI, Cisco seems to be on the OpenAI bandwagon as well. Go figure. Anyway, we're about out of time today, guys.

What a great discussion, though. I feel like I could stay up here in my hotel room and not go downstairs and have just as good a time. But we can't do that.

Mike, Andy, John, Tom, thanks for joining. Tom, I hope I'll hook up with you later, maybe down on the show floor. Just text me or Slack me or whatever.

Again, hey, we soft launched Techstrong Semi today. com. Stay tuned Monday for that.

Go New York, go New York, 8:30 Eastern Time. We've got to root for my Knicks. But in the meantime, Mike, you'll have the ship tomorrow.

I'll be on a plane home. No worries. We've got more Techstrong gang coming at you.

We have a lot more Techstrong TV. As always, thanks for joining us. tv, the Techstrong TV OTT app, just about any screen you want to watch on, or any of our other great Techstrong sites.

Tom, we'll have the Tech Field Day on Techstrong TV too, won't we? Yep, we will. We've got videos that we recorded here on site with Cisco and a couple of other companies, and we'll have all that ready to go next week.

Fantastic. All right. Until then, on behalf of the Techstrong gang, we're out.

Thanks, everyone.