I am Dan Costa, the chief content officer at Worth Media Group, and I'm here in Los Angeles at the Milken Institute Global Conference 2026, and I am joined once again by John Chambers of JC2. John, thanks so much for taking time to talk to me.
Dan, it's a pleasure. I'd like to do this every year with you, and I love your questions.
Let's do it. You've been to many Milken conferences. Is there anything this year that feels different?
Yes, I have, and Mike Milken's a very good personal friend. I come here because of the audience and how much you can learn. Speed of change. AI is clearly front and center. It's crossing the chasm both in terms of every industry and every company moving to it, but people don't grasp it's going to change productivity at a whole different number than what you read about in the press where they are concerned 2% productivity growth. I'm predicting 3 to 5% productivity growth in this country being the norm.
Yeah, it's an extraordinary thing. Of course, it's the thing that's running into some bottlenecks. We've got to build the data centers to continue to grow these models. Those data centers need power, and it seems like the US energy infrastructure is just not up for that task.
Well, you're giving me a fastball down the middle of the plate. As you know, Dan, I'm on the board of directors of Bloom Energy, and KR is the CEO there. He might be one of the best CEOs I've ever seen. He literally is a rocket scientist, and he studied how do you put satellites into space and go to Mars. He's built an architecture for on-site power generation. I think we'll lead the whole industry, and I think that's the solution. The grid cannot handle it. And so, what will be the limiting factor on these data centers is power plants take you what? 3 years to build if you can get the permits. Then you have a concern about local cost of electricity versus what's added by AI. Bloom can stand up a power plant beside the data center in 3 months. They can do it in a way that the data center operator, if they decide to do so, can literally pay for all the energy themselves or maybe give some back as well and they can scale. So, I think you're seeing a segment of the industry where power is as important as the semiconductor piece, as important as the models, as important as the cloud players. I think it's something people would write about being exciting. I think that's what you're going to be talking about more each time you're here. Let's see if I'm right or wrong.
I think the reluctance, a lot of people are worried that they're not going to hit any regulatory blockers. I mean, a lot of people, I live in upstate New York. A lot of people don't want any power plants up there. They don't want these things to be built. Are you worried at all about just governments and small towns saying no, we don't want to have it here?
Well, I'm worried more that when I did the internet as part of Cisco and other companies, 98.2% of America felt the internet was good for our country and good for them. That's not where we are with AI. So, we've got to say how do we address the legitimate concerns of the citizenry? And part of it is you have to put power generation in a form factor that is acceptable, that doesn't generate pollution, that doesn't drive up your local electrical cost or cause other problems. So, I actually think this is one that on-site power generation is going to be the answer. Clean on-site power generation that will scale. So, I'm actually not as concerned about that as I am some of the other questions you're probably going to ask me today about who survives out of this because the internet, it raised all boats for a period of time. Cisco was lucky enough to break away from the majority of our peers, but it raised all boats for almost a decade. Not going to be true with AI. It would literally break down that bell-shaped curve that you and I studied in school where if you were in the middle you were safe. No longer, it's going to flatten out. Even the Magnificent Seven will see instead of all moving to the far right, they're going to hit a bell-shaped curve, too. I'm not sure which ones will not make it. A couple will not in a meaningful way. So, that speed of change is what people don't grasp. It's just not going to be technology company. It's going to be automotive. It's going to be healthcare. It's going to be finance. It's going to be retail. And whichever country is leading AI will lead in standard of living increase. Will also lead economically in defense. So, this is about national security as well. But also productivity. If you grow productivity, when I said originally 4 years ago 7 to 10% productivity for a well-run company is a given, you said no way. Now, it is actually too conservative. My average startup, and I've got 11 unicorns to decacorns, so I've been very lucky, but the average one is growing productivity at 50% a year.
And is that through embracing AI on every level of the company?
Every level of the company. It starts off first in engineering. If you're not getting a 60 or 70% increase in your code generation as a company, whether you're in the CIO position or in the technology position, you've got a bigger problem. But it will go everywhere. It will go legal. It will go finance. It will go supply chain. Big data. Every aspect. Customer service, etc.
It's fascinating because one of the concepts I was introduced to just a couple of months ago was token maxing. Where companies are setting up goals and saying and actually evaluating employees based on how many tokens they are using because they're trying to find those that can really take advantage of this technology and identify those that don't want to be part of that process.
Well, and maybe even to take your question a second step, is it shifts very quickly between how companies consume technology. The software as a service model is really struggling as you've seen. Charging people on a per seat basis probably does not have a long life. If you have the capability to do it off of tokens and measure who's is it, who's getting the utilization, and what outcome do they get? So, the transition that you're just now starting to talk about tokens being a good first step is people pay for outcomes. So, all of a sudden I see my customers willing for even a small startup to be able to say, 'If you do this, I will give you 10% of the amount of money you save me for making this project work.' And this could be 10% of a million or it could be 10% of a billion. And so, you're going to see the models change, and the companies who emerge out of this will focus on outcomes, will focus about disrupting themselves, and realize there is no entitlement. Now, that's why you're seeing so much money, Dan, being thrown at this. If you don't play, you get left behind.
