Cisco AI Infrastructure Moves at AI Speed | Cisco Live 2026

Hi, everyone. We are back here with our final interview of Cisco Live. It's been a great couple of days here interviewing some of the leadership of Cisco, some of the great customers, couple of SEC universities we had in here talking.

But this is a great one to finish off with. I want to introduce you to Will Etherton. Etherton.

Yes. Will, thank you so much for coming on Techstrong TV. Thank you for having me, and this is my last chat for Cisco Live also.

So too. All right. There we go.

So let's go out with a bang here. Will, let's start with you. Yes.

I don't know if everyone knows who you are out there. Maybe your family and friends watch. But beyond that- For my kids For your kids.

I do it for mine as well. Will, tell people a little bit about yourself, your journey. Sure.

We'll start there. So I've been in networking since 1998. Did silicon design architecture.

I was acquired into Cisco in 2000, the first time. Really? The first time.

I've been acquired twice. I know a lot of people like that, to tell you the truth. That's not a bad way to come into company.

Oh, yeah. So came in, was doing architecture for large multi-terabit systems that would go into an AT&T. Did network processors, all those kind of things.

Really enjoyed that, and then at some point, got pulled into the dark land of management. And so, I worked at Juniper. I did a second startup that then got acquired in.

And so, all through that, the big themes for me have always been in critical infrastructure, large software systems, the boundary of software, silicon, hardware, optics. Those are things that I've always enjoyed. Fantastic.

So your relationship with Cisco, on and off, 25 years. Yep. More.

27, 28 years now. Yeah. I'm feeling old.

You don't have to tell me. Because I was there. I was in this industry for all those years.

Not behind the camera, more in front of the camera, but in tech, doing startups myself. So what are you doing now, Will? So I've, at some point, run engineering for most all of Cisco networking.

And right now, my big focus has been this whole AI infrastructure. So I'm doing software engineering teams and systems teams for the hyperscale, so the things that we sell into that set of folks, across silicon optics, software systems. And then service providers and sovereign clouds, data centers, and enterprise.

And so, yeah, that's been a big focus. And obviously, the last two or three years, that has just been crazy with the whole explosion of AI infra, so I've been a lot busier. I get a lot more questions and requests to be on camera than I did a few years ago.

So, yes. Is that a good thing or a bad thing? You will tell me at the end of this, John.

Oh, you're doing fine. Depends if you like being... I've gotten comfortable, but you're right, though.

I tell people, especially people who aren't in tech, like friends at home and stuff, these are really historic times. Yeah. Look, we've been around- Yeah ...

around the same amount of time, you and I. Living through the dot-com. I sold my first business into the dot-com.

Yeah. Helped take a company public. Rode it all the way up, all the way down.

I've been in tech ever since. Yep. And those were crazy times.

And I try to think back, was it that crazy compared to now? Or is this just that much more crazy? Yeah.

But these are historic times, and I think when we look back with the historic hindsight, this might be crazier than the dot-com. It is crazy. I will say, in some ways, this is crazier.

In some ways, I do think there's more fundamental technology underpinnings to this crazy- Yes. Right. Yeah ...

than there was last time, but there is- Well, but- ... there's crazy on top. How's that?

There's crazy, and then there's crazy. Yes. I remember in those days, if you...

com. Yeah. Yeah.

com in, you got funded. ai on your thing and you're getting funded. ai after their name, if there's really a business there, or are they features that- Yeah ...

get soaked into the next product? I think that's a question we'll see, but I think that's why you and I spend a lot of time on the infrastructure because the infrastructure definitely has some strong technical underpinnings. Absolutely.

Yes. And here's the thing. I wrote a few articles already from my experience here the last few days.

Did you? When you look at all of the announcements at Cisco Live, you could take each one in its own sort of silo and say, "That makes sense. " But you really got to take a step back and look at how they're connected.

Like for instance, this Silicon One. Yep. You look at it and say, "Well, it's Cisco.

They make ASICs. " Yeah, they do. But this particular generation, this particular brand of silicon really is the foundational underpinning- It is ...

of a mountain that's being built on top, right, of this whole AI infrastructure game. And on that, if you look back, and to your point, when I came into Cisco in 1999, 2000, we were doing silicon. The difference is we had silicon for each domain and each area with separate architectures.

ASICs. They were ASICs, separate architectures. purpose built.

Purpose built. I remember selling that. Yeah.

But they were different. Yeah This, for the last 10 years, Silicon One, I would say two key parts of what makes it a platform rather than just Cisco making ASICs, is we have taken on the full silicon process, the series aspect, the full custom, the back end. Yep.

We're taping out to TSMC. That means we have supply chain and all those things. So, that's one part.

Second part is architecturally, it is a common silicon architecture with P4 programmability built in, even at these crazy high speeds. " And we did, but the rate of bandwidth was low enough that we could actually put risk processors into those things, and we could still meet the line rate. Now, we're at a point where when we're talking about 25, 50, 100 terabit, 200 terabit, and this is moving every 18 to 24 months, having any programmable actually is uncommon.

So, that is a key substrate. Yeah. And then it builds from there.

We have a common SDK for those silicon that we're using. So, that's what makes it a platform. And I think when you talk about themes, I would say our view, a lot of things, is around Cisco, what can make a platform that helps our customers be more than the sum of the parts.

I agree with you 100%. What I find interesting is, to me, AI is putting pressure at every step of the way. Yeah.

At every station. It's forcing us to go at AI scale, at AI speed. Yes.

And human, what we did before just doesn't, it won't hold up. Yep. It'll break.

And so now you've got that AI silicon, and you're right, Cisco has to own their own silicon. Hardware has never been as sexy- Absolutely ... as it is right now.

Yep. We just launched the site today. We have eight different sites.

Mm-hmm. DevOps, and security, and platform, and AI. We launched Techstrong Semi today.

Okay. Because the semi industry, it always underpinned it- Yes ... but it was never this front and center.

Folks like to believe it was going to be a commodity where there was no differentiation, and it could all be... I think that has so flipped back. Absolutely.

Yes. Who would've thought that? Yes.

Think back to 2000. Yes. Who would've thought?

It was boring. From 2005 to 2015, maybe. Well, I think the crux was the software eats the world.

That was the point of flipping of- That was right ... all this hardware stuff is just commodity and so on. Right.

It's all- Cloud was a big thing, but I think we've, again, come back. The hardware is very differentiating. Well, this is- Yes.

Just between you and I, who's paying attention. Yeah. The fact of the matter is we never moved more than, what, 25% of our workloads to the cloud or something like that?

It was a little bit higher than that. But yeah, there is still a massive amount in folks' data centers today, distributed globally, sovereign, telcos, and so on. Yeah.

Absolutely. But again, with this AI infrastructure, we want to make it, whether it's in a public or private, a little of this- Yeah ... and a little of that, it all has to move at AI speed.

Yes. And that's really the thing. But then if you look at what Cisco has done here, and they've announced on top of this great new silicon- Yeah ...

it's still software, too, right? Yeah. It's this AI infrastructure, cloud control, everything they've announced.

From where you sit, because you're in the network. Mm-hmm. You're sitting right in the crux of it.

How does that all play in? Is it I need to keep developing my silicon to keep up with this, or is that putting the horse before the cart? I've had a few chats where the way I think about it is a lot of my engineering has a day job, which is the box, getting these boxes out.

The new silicon's coming on. We have to power that up. We have to get that working.

And the topic of AI speed, we had customer chats back at Cisco Live Amsterdam about, hey, you know what? There's these multi-tenancy features in InfiniBand. It'd be nice if we had that in Ethernet.

The team took that back, and this is what we call co-design. We worked from the bottom of P4 programmability in the silicon all the way up. How do we add that type of capability into Ethernet?

And we actually are demoing that here- Really? four months later. And that is very stressful.

That's also about AI speed, though. That is AI speed. No one wants to wait two quarters.

They don't want to wait. So, you have all the hardware, you have these features like this kind of thing, and then on top of that, we've had the Cisco Cloud Control that you mentioned. I helped co-start that within Cisco from an engineering standpoint.

We've been working on that 9 or 10 months. And that was something where as we went along, it became clear that for everything I'm doing in the data center and all this AI infra, for the things that the teams are doing with Catalyst and Campus and SD-WAN, the security team, that if we each are doing that independently, that it is still not providing the full value. So, we got going on the Cisco Cloud Control, and the notion that it was more than a UI.

It's around bringing together common APIs, common user, and then the most important part is AI Canvas. Of course, it has to be AI. And some of the demos we have where you can ask for debugging that might be across security and data center.

You can ask for updates. It's not just observability, it's also making changes to your infrastructure. So, this elusive closed loop automation that companies have been looking for, that was a project that we did in parallel to all of the day-to-day work.

So, it has been stressful. It's AI speed. It is AI speed.

And scale. But here's the rub. Yes.

You know how hardware and software have this relationship. Mm-hmm. The faster you can go here- Yep ...

the more they seem to put on here. " Yeah. How do I make the next breakthrough?

How do I get through to that? I'm not asking you to give up anything confidential, but what's next here? So, it's interesting you say that, because one of the things that's also been a big topic all week has been the glass wing mythos, the use of- Yes ...

LLMs to find, and I think one of the things that, and again, there's the day job, the night job, the weekend job, we've actually been saying just in the last month that as we're getting a feel for the vulnerabilities that these models can find, all this other feature stuff is awesome, but actually bumping to the top of the priority, we need to get out releases that are addressing these types of vulnerabilities that's been found very quickly. That has now become what all the engineers are working on as their top priority. It has to.

It has to be. And so that is what's next for the next three months, six months. And then in addition, obviously we'll keep working on the architectures and so on.

But this brings a whole new discussion to how will upgrades be happening, when will they be happening. Interesting enough, I've had more discussion this week also about deployment models. We've had SaaS has been beaten down in the market.

The SaaS apocalypse. The SaaS apocalypse. But now interesting, in infrastructure, we have offerings from Cisco, for instance, like Meraki and- Mm-hmm ...

wireless and so on. We have Hyperfabric, which is a SaaS for data center. We've had customers that six, nine months ago were saying, "Yeah, where possible, I'm going to stick with on-prem.

I might experiment. " Now this is coming back, "Hey, you know what? SaaS managed infrastructure means that from the time we see an issue to the time we can fix it is might be two weeks.

" Versus we see it, we go through testing, we publish a- A cycle ... advanced CCO. A customer goes through testing in their lab.

The customer deploys it. And I think this is a broader discussion that SaaS managed infrastructure should have a place. And I'm not saying it's everywhere.

No. But I think this makes a big case for where it can be a benefit. Yeah.

So here, from where I sit- Yeah ... I think we're going to be in a place where nothing is necessarily dominant. Yeah.

Because if you look at it over the last few years, it's a pendulum. Yeah. It swings from here to here.

I think we're coming to a place where that pendulum might be in the middle. Yeah. And it may be I want some of my stuff SaaS and I want some of my stuff not SaaS.

Yep. But I want that to all work together. I don't want three- Absolutely ...

throats to choke or- Yes ... two butts to kick. I want a vendor who's going to give me that model, this model, and I can blend it as I need, and doggone it, I want to be protected.

Yes. Right? And I was talking to Peter, Peter Banks from Security.

Our security lead. Yep. Yes.

This mythos glass wing thing, I think it's a temporary thing as we adjust to AI speed. Mm-hmm. Right?

Because the endgame of this is it's going to force us to make better software quality. Yep. It's going to force us to make more secure software so that it's not every month is a new fire drill- Yeah ...

running out, patching, and new IT. I do believe now is going to be, is a higher level. But it's the pain point.

And then it will come down. Right. That this will be a painful period.

We're in the pain point. But as an engineering leader, I am actually excited about the fact that this gives an umbrella for engineers to spend time on refactoring code that maybe they've wanted to refactor for a while. They can use AI to help them, but this is the time.

This is the opportunity. They have a license- This is their time ... to spend on this now.

I like that. Yes. That's a great way of saying it.

It's a license to do this. Because look, I've been in security 25 years. Mm-hmm.

Everyone in security knew this. I think everyone in engineering knows this. We took a lot of shortcuts with software.

It is often that. " We never caught up to it. You acquire a couple companies.

You integrate those together. Technical debt. And yes, it is very difficult.

Yeah. com. In DevOps we call it technical debt.

We're drowning in technical debt. It's time we cleaned the house out, clean the attic, the basement. Let's get it on a secure plane and go forward.

Because at AI speed, we can't afford anything less. Absolutely. Because right now, it's mostly the good guys who are using mythos.

Yeah. Not for long. No.

I think we've had historically nation states had an upper hand, I think, with these technologies, the small attackers and the defenders, but we've got to move fast. Every two guys in a garage. Yes.

That's all you need. That's what we're competing against. Yes.

Yeah. I agree with you. So you can no longer pretend the technical debt isn't there.

No. All engineering leaders need to appreciate that, and I think this is the right time to make it better. And we got to take down the debt in more ways than one.

Yes. Anyway, Will, we're about out of time. You're our last guest here at Cisco Live.

Yes. What should we leave the audience with? I think we should leave with the first of all, just on the last topic we hit, now is the time to make software better for the whole industry.

Now is the time, and I think that's going to percolate beyond critical infra like Cisco. I think we're going to see this across enterprise software, IoT systems. A lot more attention's going to have to go into how we operate and manage software.

I think that's a really large one. AI infra at Cisco is all in on AI infra at every layer, but it isn't just about selling compute. It is about integrating in security, simplicity, operational management, and that co-design from top to bottom.

And we're excited to be working with a lot of great customers, and it's easy to say, but the crazy has never been higher. That's what I'll end on. Absolutely.

Great. Hey, man. Thank you.

Thank you. All right. All right.

Hey, that's going to wrap up Cisco Live 2026 here in Vegas. I hope you've enjoyed. tv or our OTT app on any screen you watch videos on.

Check it out. There's a lot of great stuff here. tv.

We're out.