Chuck, thank you for being here.
I wanted to ask you, in 2016, shortly after you took over as CEO, you signed off on this acquisition of Liba, which was an Israeli chip company that didn't even have a product yet.
You know, that became Silicon One.
Well, you say it that way. It makes me sound pretty stupid.
Wait, wait to see where I'm going with this. So, that became Silicon One. You know, fast forward to the AI boom. It's right in the middle of this AI data center stack, connecting server racks and making you a lot of money. And I'm just wondering if there's anything that we can learn from that bet that you made almost a decade ago.
Yeah. Apparently, I'm a visionary genius and didn't know it. Don't let this accent fool you. You know, I think that in seriousness, some of your best decisions in retrospect, there's a lot of luck associated with it because we bought it for a different reason. I mean, sort of a different reason. We bought the company because we wanted to control our destiny and our silicon, but it was a four or five year investment cycle before we got anything. You're right, they didn't have a product. They had a design on paper and they had a history of success. And I think we paid $160 million for them with nothing. They had nothing. And now we're underpinning the GPU training models and running the enterprise AI stacks. And it turns out it's been brilliant. I think the thing to take away from it is, first of all, you never know how important a decision is when you make it. And if you have conviction and your team has conviction, even on something we had to invest in for probably five years before we begin to see any ROI, you just got to stay with it. And, you know, when you're going through ups and downs and economic cycles, it's easy to say, 'All right, we got to stop funding that. It's just not helping right now.' But you got to stay committed.
I don't even know what I was staying committed to, but it worked out.
So, one thing you've been getting asked a lot about, I know at our event, is the Claude Mythos, this AI model that Anthropic released that's too good to release to the public. It's too dangerous. And Cisco is one of 11 companies that have access to this. And the idea is you're going to fix all these vulnerabilities before it's too late, before everybody. Is that the idea?
That's the idea, I think. Correct me if I'm wrong.
But you know, if this is so important, why only 11 companies? Shouldn't every company be getting access to this?
Well, there are more that have it now. We were just, we've had it two months. And so we were one of the first 11 and I think it's reflective of the role we play in the critical infrastructure in the technology world and the relationships that we have with our chief security team. And so yeah, in theory, you go and you run this model on your code and you find vulnerabilities and then you fix them and then you go through this neat process and you put out a security bulletin and then your customers see it and they say, 'Oh, I need to apply a patch' and then they upgrade all their infrastructure and the world's beautiful. Never happens. What happens is we find, before Mythos, we find a vulnerability, other tech companies find vulnerabilities, we put out bulletins, the customers are overwhelmed, they triage and say, 'I can live with that for a little while. I'll get to it next week' and next week never comes and then you land with Vault Typhoon and Salt Typhoon and all these other things that we've seen. And what Mythos does is it just massively multiplies the number of instances where that occurs. I mean, I was just over at the White House and it's like, this is what should happen and then the customers upgrade, but given the volume of what's going to be discovered, we got to figure out how we get from here to there much faster and cleaner. And that's the work that's going on right now.
But by the way, long term it's great because having those models that we can run our code through before we release it, you're going to find vulnerabilities that we wouldn't see today. And I say we as an industry as much as Cisco.
So in other words, your code will be cleaner when it's first developed.
And just another security question for you. You know, I think you think of yourself as a security company, you said, and you know, you're also dabbling in quantum.
Um, and partly as a way, dabbling a little quantum here and there.
Quantum, a little quantum over here. Partly as a way to secure your network, your equipment in a post-quantum world, which seems to be rapidly approaching. So, are you worried that quantum will get figured out before we figure out post-quantum security?
No, I feel pretty good about where we are right now on that one. I think there's been a lot of work. We've been doing research on this for, this is another one of those invest before there's anything there. Wasn't as big as a silicon investment, but we've been working on this for almost seven years. And we released the first prototype of quantum networking silicon. Didn't release it, we built it and demoed it. I think the industry's been moving towards this and been prepared for Mythos. No one, we weren't prepared for. Even though I don't know why we weren't because every two months something monumental happens. And Dan Schulman said something earlier. I don't know if he said it in his session earlier. He said, you know, these models are the worst they're going to be right now. It's like, oh god. I mean, we know that, but that's sort of an interesting way of saying it. So, I think we're on the right track with quantum, but I could be surprised.
I got someone who works for me in the room that could answer it better than I can. I think Liz is in here somewhere.
Who started this? Liz started the whole thing six and a half years ago. Raise your hand, Liz. You can talk to Liz later.
We all know Liz. So, the industry is dealing with this crippling shortage of memory. You're dealing with it. You said it's going to last about 18 months I think. And the industry seems to bump up against these supply constraints pretty regularly now.
The last five years we have, not before that so much.
So you're on this Business Roundtable. You're talking with the, you're running it actually, CEO, talking the most, chair.
There is a CEO who runs it.
Okay. Well the website's Josh Bolton. Okay. So you're conferring with the most powerful CEOs in the world, including the big tech ones, a bunch of them. Does this topic of shortages come up and what do these people think about it? What do they say about it?
Memory specifically, or other?
No, memory, but also just all these supply, silicon shortages, etc. Like, is there something we can do?
Yeah, I mean, the AI buildout has just exacerbated all these things. And so right now memory is the issue. Who knows? It could be PCB assemblies. It could be wafers. It could be power supplies. I mean, we, so I think what we're learning and we're continuing to build capacity. And I think that, you know, I think we coming out of the COVID supply chain or getting into the supply chain shock that we saw post-COVID was not anything that we'd experienced really before, at least not in my lifetime. And so we learned a bit from that and we did a lot of work around resilience of supply chains. I think now we have to do a better job of capacity planning because building fabs to generate more memory as an example. I think the current estimate is 75% of the capacity over the next year or two is what we believe we have as an industry. And I think the only way to solve it is unique approaches on the demand side, like we're looking at products where can we redesign and get by with less memory in the short term. Everybody saw the Google announcement about some of the things that they're going to try to do. And you could just see demand on consumer-oriented products slow because the prices are going up so rapidly. You may see smartphones, demand for smartphones slow down if you, you know, if somebody's got to pay another $100 for a phone what they paid before. To a lot of people in this room that may not be a big deal, but to most people it's a big deal.
Yeah, it is. And you know that's another question about the sort of view of AI in the world, right? I mean we're seeing, you know, attacks on, you know, CEO's homes in San Francisco now because people are so outraged about this. Do you think that these rising prices are going to exacerbate that? Is it something you worry about?
Um, I don't worry about it exacerbating that. I do worry about, and I don't know that this is definitively going to occur. So don't say that I believe this or not that I fundamentally believe this. But should we experience meaningful job loss as a result of AI, I think you could see more civil challenges.
