What the AI Boom Looks Like from Inside Cisco with Sam Badri

Steve Eisman

Hi, this is Steve Eisman. Welcome to another episode of The Real Eisman Playbook. So we’ve been doing a lot of work on AI on this podcast for the last many, many months.

We’ve gotten different perspectives. We’ve had on Gil Luria and Dan Ives, who are very positive. We’ve had Gary Marcus on, who is negative.

We recently had Ed Zitron on, who is really negative. But it’s also good to talk to people who are not commentators, but are in the trenches. And so today we have, as a returning guest, an old friend of mine, Sam Badry, who is head of investor relations at Cisco.

Sam, welcome back.

Sam Badry

Thank you, Steve.

Steve Eisman

So for those who don’t know you, other than the fact that you are head of investor relations at Cisco, let’s tell people a little bit about your background. When did you join Cisco and what were you before that?

Sam Badry

Yes, and thank you for having me back again. Maybe I’ll do the funny plug first. Okay.

One of my family members recently asked me, they said, hey, Sammy, you’re involved in this tech finance stuff. Have you heard of the Einstein playbook? And I said, do you mean the Eisman playbook?

They’re like, yeah, yeah, that. I said, yeah, I’m familiar. It’s kind of funny.

So congrats on the success.

Steve Eisman

Thank you. I think I may change the name. I think it’s a better name.

Sam Badry

Okay. If it’s getting to one of my family members who asked me this question, it means you have reach and distribution. So I started off, I graduated from undergrad.

I started off in consulting at PwC. I did that for about a year and eight, a year and nine months. And then I started my career really on Wall Street at Credit Suisse, working as a macro researcher.

I did that for a very short period of time before going into the equity research department, where I started off really at the kind of the lowest part of the totem pole as a junior analyst.

Steve Eisman

So you were working for another analyst.

Sam Badry

Correct. I can even call him out. His name was Kolbinder Garsha. He was the guy who covered Apple, IBM, HP, HPQ.

Steve Eisman

Good coverage.

Sam Badry

Really big coverage, big cap. And it was while the sector was changing.

Steve Eisman

What years were this?

Sam Badry

This was going back to the 2013 plus time period.

Steve Eisman

Got it.

Sam Badry

So I worked in the trench as a junior analyst for quite a bit of time. And I was a research associate and a junior working for a senior analyst. And then I did that for a number of years.

I think it was about four years. And then I got tapped to take over data center infrastructure coverage, which included companies that at the time weren’t even well known. They were Equinix and Digital Realty.

And I helped out on an IPO for, say, a company called Switch, Switch Inc.

Steve Eisman

Which got bought out, didn’t it?

Sam Badry

It went private. And there are already reports that it may potentially go public again. But we’ll see how that ends up happening.

Okay. And then I started off with data centers. I then absorbed networking.

So companies like Cisco, Arista, Juniper at the time. It also included companies. Sorry, in addition to that, I started growing my coverage into large cap software.

That’s when I started bringing in companies like Oracle, Microsoft, and some others. And then Credit Suisse was acquired by UBS. And in that window of time, I made the decision to leave Credit Suisse.

And while I was on guard and leave, and then soon after I started at the new bank, around that window of time, the former CFO of Cisco approached me and said, hey, would you like to come and take on this head of investor relations role to help explain our story to investors analysts. At the time, it seemed like a really good opportunity to finally make my way into industry. And now I’m here.

I’ve been at Cisco for three years.

Steve Eisman

Great. So before we dig down into what’s going on at Cisco, because my audience has all different levels of understanding. If I said to you, oh, you’re head of investor relations, Cisco. What does that company do? Let’s explore that simply first so that we can really dig deeper for everybody.

Sam Badry

It’s a really good question. And the majority of our business is stuff that people do not see.

Steve Eisman

Right.

Sam Badry

It’s the stuff that allows connectivity, like our phones to send a message from one phone to another phone or to connect networks together, which was the original foundation of the whole company was helping one database in a university connect to another database within the same university. So that network connectivity. In addition to what we do, that would be a switch or a router.

In addition to that, we produce Wi-Fi access points. So allowing people to go from their devices like a laptop or a phone and connect to the internet through a Wi-Fi access point. And we mainly sell to major businesses or enterprises.

And that is the bulk and the majority of the company. And since then, we’ve gone to many more things. We’ve gone into silicon, at least in the last decade.

And here’s like an example of our G300 Silicon One chip that we are currently marketing.

Steve Eisman

Now, is that a GPU, CPU, memory chip? What is this?

Sam Badry

This is a network processing unit, an NPU.

Steve Eisman

NPU.

Sam Badry

Yes. So a GPU would be considerably larger than this.

Steve Eisman

A GPU would be bigger.

Sam Badry

A GPU would be larger. How much bigger? It depends on the GPU or the accelerator. So it could be something like Cerberus, which is another kind of accelerator chip company. That would be a much larger device.

Steve Eisman

How about an NVIDIA chip?

Sam Badry

That would actually also be larger than this considerably because when NVIDIA puts together the products, they are essentially a large closed system with many pieces that kind of look like this.

Steve Eisman

Okay.

Sam Badry

We sell one piece in addition to the switch. Other companies can also sell all combined systems, almost like an entire rack of compute.

Steve Eisman

So what does this chip do that a GPU does not do?

Sam Badry

So this actually allows you to transmit information from point A to point B. this becomes the processing layer of light to transmit said information.

Steve Eisman

And this is new?

Sam Badry

This specific chip is very new. This was announced in February of 2026, but the actual silicon program is not new at Cisco. It’s been around for about 10 years.

Steve Eisman

Okay. All right. So let’s dig down into what’s happened at Cisco. So you joined Cisco what year?

Sam Badry

Three years ago. Yeah. So 2023 timeframe. Okay.

Steve Eisman

So let’s say the very first quarter that you were there, roughly, what was the revenue growth of Cisco at that time?

Sam Badry

Actually, we were completing a year where the revenue growth profile of the company was following a pretty big shipment schedule. So it was actually fairly attractive, but you could think of it as almost like normalized single digits. Single digits.

Mid-single digits. And then shortly after I joined the company, we went through a bit of a inventory digestion period where we shipped a lot of gear and our customers need to absorb it. So we had some patchy or rough quarters that followed that.

And then we guided to the street to grow around mid-single digits around that window.

Steve Eisman

This most recent quarter, your revenue growth was 18%.

Sam Badry

In the quarter? Yeah. In that range, yeah.

Steve Eisman

When was the last time Cisco had revenue growth normalized that was that level? Would that be during the dot-com boom?

Sam Badry

You’d have to go back quite a bit. But there could have been some quarters which saw a little bit more activity, right?

Steve Eisman

But not consistent.

Sam Badry

Not consistent. So normalized, not in this range.

Steve Eisman

Okay. So, let’s talk about, you’ve got a company, three years is not that long, whose normalized revenue growth is, call it 5%, and now it’s high double digits, call it 15% to 20%. Why?

What’s selling? Why is it selling? Let’s dig down.

This is not a small company. So to go from 5% revenue growth to 18% revenue growth, that’s pretty dramatic. So why?

Sam Badry

Well, I’ll tell you something dramatic as well. When we guided to 5% top line growth for essentially kind of a multi-year long-term guide, the sell side analyst reports essentially described this guidance as, here’s how we think this growth could be achievable.

Steve Eisman

Achievable.

Sam Badry

It was a very glass half empty type of framing.

Steve Eisman

Right. So in other words, the people who covered your stock, you said 5% and they said, eh.

Sam Badry

Yeah. They’re like, yeah, we’ll see. It was a we’ll see kind of reaction.

Got it. Okay. So what’s changed today?

So our fiscal year 27 guidance is actually 15% revenue growth at the midpoint. So what changed so dramatically? And it’s not like our last 12 months was slow growth.

Our last 12 months of growth was also low double digits growth. So we have an acceleration for the next 12 months versus the prior 12 months. So two big things have changed.

Big thing number one is our hyperscale business with clouds. So think of your household hyperscalers. Our business with them is growing significantly.

Steve Eisman

How much?

Sam Badry

Almost doubling in FY27.

Steve Eisman

Doubling. Yep. And what percentage of the revenue of the company is that?

Sam Badry

It’s still, it’s not a material piece of the company. It’s still below 20%.

Steve Eisman

Below 20%. But it’s doubling.

Sam Badry

It’s doubling. Okay. And that roughly doubling in FY27.

Steve Eisman

Okay. So that’s one thing.

Sam Badry

Yeah, that’s right. What else? Thing number two is…

Steve Eisman

But by the way, before we explore that, so your hyperscaler business, what are you selling to the hyperscalers just so that people understand why that part of the business is growing so rapidly?

Sam Badry

So we sell them, it comes down to really two main products. One of them is a switch, which is designed for hyperscalers that have chips that look just like this that are in them that allow them to transmit massive amounts of traffic through their data centers.

Steve Eisman

Okay.

Sam Badry

So that’s kind of the first product we sell. The other main product we sell is something called a coherent pluggable, which is essentially like an optical-derived type product, and that’s to connect a lot of fiber connections or Ethernet connections to switches. So this is all in the connectivity type spend envelope of a hyperscaler.

Steve Eisman

Okay. So that part of your business, the hyperscaler business is doubling, but it’s less than 20% of your business. What else is growing a lot?

Sam Badry

So this gets us into something called the campus refresh cycle.

Steve Eisman

Campus refresh cycle.

Sam Badry

So if we go back over the last almost two decades, we have shipped what I would describe as almost like less capable campus switches. So maybe I can zoom on and describe what a campus is and help explain the differences.

Steve Eisman

We’re not talking a university. Or are we talking about a university?

Sam Badry

A university could be a campus, but let me kind of maybe describe Define

Steve Eisman

what a campus is. Exactly. Okay.

Sam Badry

So everything we sell to a hyperscaler predominantly goes into a actual data center. And if you searched what a data center looks like on the news, that’s essentially where the hyperscale type products are going that include chips like this and that coherent pluggable I refer to. Okay. So that is the data center kind of opportunity.

Steve Eisman

And they get the best equipment.

Sam Badry

They get the best equipment, the highest performance stuff, which is typically very demanding stuff. Okay. Then there is something called the campus switching market and opportunity.

A campus switch or a campus could be an office building in downtown Manhattan. It could be a university campus. It could be an airport.

But really the simple way to understand it is a campus is where people like us, like end users, are walking, talking, transacting, and being productive. So that’s where the end users are very close to the switches. That is a campus.

Steve Eisman

Okay.

Sam Badry

So in the Cisco business, the data center switching opportunity and the hyperscale business that’s growing, it’s doing its own thing, separate to that, we have a campus refresh cycle where we’re essentially telling customers the old stuff that you used to have, we are no longer going to support it. because some of these devices have actually been deployed out there for 15 years plus. So Steve, imagine having like a phone, like an iPhone that’s 15 years old and still using it. There are still Cisco devices that are sitting out there that are in production for enterprises, still running today.

Steve Eisman

Okay.

Sam Badry

So we are no longer going to support the patching and the updates from some of this older gear. And as customers make these rip and replace decisions, they’re submitting new product orders to Cisco for the newer stuff. And then we are shipping it accordingly. And that’s what’s driving our 15% overall revenue growth in addition to the hyperscale opportunity.

Steve Eisman

But the hyperscaler part of the story, would you call that the AI part of the story? Yes.

Sam Badry

Short answer is yes. But we can kind of, as we talk, you’ll kind of find that the campus opportunity does draw in some of the AI dynamics as well. And we can get into that.

Steve Eisman

Okay. So let’s broaden the conversation a little bit. From your seat, when did this whole AI story really start to cause Cisco to inflect?

I’m sure there was not a day. Maybe there was a day or an hour, but there must have been a timeframe where you said, okay, things are really changing here. When was that?

Sam Badry

At the company specifically, right?

Steve Eisman

Yeah, I’m just talking about Cisco.

Sam Badry

So for us, the overall sector trend was already moving. So companies like NVIDIA were executing, communicating. The hyperscale CapEx was kind of already being spent. So those trends were already in motion. The point in time it changed for Cisco when we started seeing significant demand for specifically our Silicon One-based systems.

Steve Eisman

What is your Silicon One-based system?

Sam Badry

These are the switches that take this chip, which is under the Silicon One portfolio. As we started seeing product orders come in and much bigger size than we historically have seen them come in, that really became the AI relevance factor of Cisco. That was like the inflection window.

Steve Eisman

And what was this?

Sam Badry

This was probably between six and nine months ago. So not that long ago. No, I mean, we were shipping to it.

But I think the pushback from the investment community is, good job on your silicon. You guys are clearly winning. But it just wasn’t a big piece of revenue.

So today it is larger and it’s growing at a faster pace. And we’ve been able to demonstrate multiple use case wins in different parts of a hyperscaler network. And this is what’s making it a little bit more real to the market.

Steve Eisman

Okay. Can we do a comparison between Cisco and Arista? Because Arista is your biggest competitor.

It’s a newer company, smaller, but faster growing. For people who are not that familiar, I mean, other than knowing that the two biggest companies in your sub, little, any meaty subsector are you and Arista, how do you differentiate between yourself and Arista? What are the differences?

Sam Badry

So if it came down to hardware and for hyperscalers, it almost really does. We have our own silicon. So silicon one goes into our switches and form factors.

Whereas for Arista, they typically are buying someone else a silicon and putting it into their switches and then programming it and shipping out to customers. So their additional argument to what makes their product offering so compelling is they have very advanced software engineering, which is required to be in this business. We have that too, but that’s not the flagship differentiating factor of Cisco versus Arista when we sell to a hyperscaler.

Steve Eisman

Okay. So why are they growing faster than you? I mean, they have been growing faster for a long time, even with your acceleration in revenue. Why is that?

Sam Badry

So their revenue mix is much more indexed and weighted towards hyperscalers as an entire company. Whereas for us, it’s essentially like a fraction, right?

Steve Eisman

And that’s because you have many more legacy businesses.

Sam Badry

We have many other businesses. We have security. We have collaboration. We have a services business that’s one fifth plus of the company. So we have many other things that are indexed to slower trends than Arista, which the majority is indexed to hyperscalers or faster moving markets.

Steve Eisman

Let’s broaden the conversation a little bit more. So, look, from where you sit, things are great. Your revenue is accelerated.

It may even accelerate more. But there are a lot of people out there who are very negative on the AI story. Partially, if I could, and this is where I want you to comment, I mean, A, NVIDIA had an insane quarter.

You know, the largest company on planet Earth who revenue over 100%, which when you think about it, it’s even hard to get your mind around that. But 70% of its accounts receivable are from five accounts. That’s from their 10Q. And it seems like the business is incredibly concentrated.

And at the end of the day, in some ways, some people would argue very dependent upon the health of anthropic and open AI succeeding. Because so much of the, you know, if you go right down the food chain, is dependent upon them. How do you see, I mean, I’m sure you’ve read this.

You’re in the industry. You were a cell site analyst. How do you see this issue, if it’s an issue at all in your head?

Hi, Steve Eisman here. Most meetings end with a bunch of action items floating around, but they’re not documented anywhere concrete. Someone thinks they’re handling something, but it’s not actually written down.

And a week later, nothing has happened and nobody knows who owns the task. That’s why I switched to Granola. Granola is an AI-powered notepad that makes your meeting notes useful.

It captures what happens in your meetings and turns into clean, structured notes with decisions and action items pulled out and made easy to find. And the best part? Granola integrates seamlessly with how you already work.

No setup, no configuration, no friction. It transforms messy meeting calls into something you can actually reference weeks later. Your notes become searchable, organized, and actionable.

You get better notes in the central place while you get to stay present in your meetings with no extra work. I use Granola, and it has improved the usefulness of my meetings dramatically. Now I know exactly what’s happened during my meetings.

Once you try it on a first meeting, it’s hard to go without. You can try it totally free. to granola.ai slash eisman.

That’s granola.ai slash eisman to get your time back. Go to granola.ai slash eisman and try it on your next call.

Speaker 1

When you need to build up your team to handle the growing chaos at work, use Indeed Sponsored Jobs. It gives your job post the boost it needs to be seen and helps reach people with the right skills, certifications, and more. Spend less time searching and more time actually interviewing candidates who check all your boxes.

Listeners of this show will get a $75 sponsored job credit at Indeed.com slash podcast. That’s Indeed.com slash podcast. Terms and conditions apply.

Need a hiring hero? This is a job for Indeed sponsored jobs.

Sam Badry

So I appreciate the question because two companies are really kind of commanding so much of the narrative, right? So it’s a very fair concern. The way I would really kind of compare this to or really kind of layer it and break it down is what is driving these two companies to be as influential or as commanding?

Steve Eisman

Correct.

Sam Badry

What is leading that? What’s getting us to that point where we are anchoring on these two companies being so influential or commanding of this entire cycle? I think the answer is consumer demand.

By consumer, I mean consumers of frontier models looking for tokens. You’re seeing a very widespread collective demand acceleration for tokens to basically use what OpenAI and Anthropic have created. This is essentially consolidating so much traffic activity to two companies, and this is manifesting to these two companies being so pervasive in the industry.

As I think through this, it’s almost like imagine if Google search was a private company. That’s kind of like what’s going on here with one company being so capable or two companies being so capable, but being so compelling to so many end users, right? This, I think, is almost like a metaphor or a parallel, is these two companies are the creators of these, really of these categories, and they’ve created these products that are being used so widely with token consumption that, quite frankly, is accelerated and further model development that’s just making the productivity or value increase.

So that’s really to speak to their commanding position in the sector. But how that ends up manifesting to the supply chain, I think, is probably a different type of consideration.

Steve Eisman

What do you mean by that?

Sam Badry

So how companies are figuring out how to stand up the capacity for that, that is like a mechanism that is playing out in itself. I personally don’t really have a comment on how it’s all being stood up. What I would always track is what is the quality of the demand vector that’s coming in? The quality of the demand vector seems to be very positive, which is token consumption or token growth.

Steve Eisman

And what about the fact that token maxing seems to have ended and you’ve got big competition from open weight models? How do you think about that, if you do think about it?

Sam Badry

So I do think about that. And I actually have a funny question for you. Sure.

Steve Eisman

I could use some humor.

Sam Badry

Okay. So on average, right, or people you meet, when you ask them about these AI tools, how do these people typically explain what they did with these tools? Like, have they said, I’m a power user.

I have uploaded a skills file. I’ve deployed my first AI agent. Do they ever say things like that?

Or did they say, hey, I submitted a prompt?

Steve Eisman

Well, the person who I speak to the most about this is my son-in-law. who John, who I therefore trust, he works at a hedge fund. He’s enraptured with this stuff, and he uses it for research. He uses agentic agents. I mean, I don’t think there is a single tool that he doesn’t use. And the way he describes it, because he understands this much better than me, obviously, is that prior to using these tools, maybe he could cover 40 stocks. This is the buy side. He basically says that his capacity has doubled. That’s how he describes it. So I take that very seriously. My issue with the reliance on anthropic and open AI is they’re like the Mercedes of the LLM world. and you’ve got these Chinese and open weight models, which are considerably cheaper, and you don’t need Mercedes to do every single task. Maybe you need Mercedes to do 5% of your tasks. So, I mean, my view of the whole situation is, I don’t know how this is going to turn out, but if one of these guys gets in trouble, we’re in trouble, because so much of the business is dependent upon them. That’s kind of how I view it.

Sam Badry

That’s a fair assumption. And I want to get back to the, it sounds like John is a power user. Yes.

If his workflow is set up. But Steve, it would really shock you. How many people have not gone past simple prompts on these products?

Steve Eisman

What do you mean simple prompts?

Sam Badry

So let’s just take a simple prompt. Like what did Steve talk about today on his podcast? Produce a summary. That’s a pretty simple prompt.

Steve Eisman

We’ve done that.

Sam Badry

Right? Okay. But has anyone ever uploaded every single transcript of every single pod you’ve ever done?

Steve Eisman

One person did this for me.

Sam Badry

Okay.

Steve Eisman

One person. It was wonderful.

Sam Badry

Right. And then it can become almost like a It’s

Steve Eisman

like a book.

Sam Badry

Like a book and it learns you. Right. And all of a sudden it’s got memory and it’s got capabilities and you can even tune it, right?

Steve Eisman

What do you mean tune it? So this gets us into the conversation of a skills file, right?

Sam Badry

Okay. A skills file, let’s just say you’re trying to research a stock and I’m sure your son-in-law understands exactly what this is and any kind of power user or more sophisticated user understands this. When you’re using these tools, there are different stages. There is the beginner stage, there’s the intermediate, and there’s the power user stage.

The fair majority of people that use these tools today are in the beginner stage. It’s probably 90% of the users, and it’s considerably smaller for power users. But a skills file is, say, every analyst, you used to be an analyst, I used to be an analyst, we think in a very specific way, and we look for certain things.

the way we can increase our bandwidth of research, kind of like what John is doing, is to give the skills file very specific explanations of what to look for when analyzing a specific stock company or trend, right? So let me tell you how I built my skills file to understand how this gets built. I found some of my old research reports.

I uploaded all of them into one of these AI harnesses, which is like Claude Cowork or something else. And I said, Create a skills file using all the different research and valuation techniques that I discuss in these research reports and then create it in a way that also allows me to edit it so I can keep modifying the skills file as I go.

Steve Eisman

So I don’t understand what that means, skills file.

Sam Badry

It’s literally a Word doc of instructions to basically steer the LLM and the harness on how to research and how to analyze a stock.

Steve Eisman

Okay.

Sam Badry

So it’s just like if you had a teammate on your team and you said, hey, I want to look into X company, that person probably already knows what to do, right? Because they’ve been working for you and they understand what to do. The skills file is essentially equivalent to that.

It is a set of parameters, angles, terms, explanations for the LLM and the harness to really interact with to enrich the output that you are going to get so you can accelerate your research process. So I’m sure your son-in-law, John, has a very capable skills file, probably 30 to 100 pages. My skills file is 40 pages, and it could be longer.

Steve Eisman

So when you have this skills file, what do you use it for?

Sam Badry

Today, I just upload. So say there’s some big, long industry research report, like a blog post, right? And I’m on the run.

I copy the whole thing. I paste it into my workflow and say, look at the attach, like the attach copy paste. Tell me if there’s anything I need to care about in this actual posting and then map it to my skills file.

The output is stock by stock, company by company. It says, hey, Sammy, here is what you should care about. And it just accelerates my productivity.

I know I don’t feel like I have to go read that full blog post anymore.

Steve Eisman

OK, so it basically gives you a great summary in the manner in which you want it.

Sam Badry

Exactly. I see. Now, it comes down to I want to tie this into the open weight versus the frontier models as well.

Steve Eisman

OK.

Sam Badry

So there will be work processes that do not need the latest and greatest from Anthropic and OpenAI. These can be open-weight models. And even at Cisco today, we do have workflows that use open-weight models and that are much less costly than a frontier model. But for a lot of executives that are operating today and a lot of kind of team leads or I would describe a very demanding team. So, for example, the Cisco Silicon engineering team is a very demanding division of the company.

Steve Eisman

What do you mean demanding?

Sam Badry

Very intensive, very competitive, can be very time-consuming because of all the runs and all the analysis you need to do to come to the final product. You want best-in-class intelligence to accelerate productivity. And for that, you will pay up.

And we acknowledge that. And even at the company, we have a whole range. We have open weights.

We have frontier models. I almost think of it as almost like, think of like a team at a bank. You have like the team lead, and then you have a bunch of team members that report to that team lead.

You don’t need tier one first string players on the entire team. Right. You need two probably top guns in there.

You need two middle of the line with some specializations, and you have two other members that maybe they’ll figure it out. The open weights are like the, you’ll figure it out. But then you need that real competitive intelligence sitting in those other four seats.

Because you’re going to be running your operation on that. Those are like the real rails. So maybe the hope is over time, the open weights get better.

And then the frontier models remain as competitive. And that may dictate how things land. But from where we are today, it’s looking like quite a distribution in favor of the frontier models as far as main usage.

Steve Eisman

Let’s turn the page for a sec to a topic that I’ve written about a lot. And I’m curious, given your sell side bracket, what you think. companies like Google, Amazon, Meta, where I would call them capital light businesses that threw off so much cash that they literally didn’t know what to do with that cash. I mean, they’re like flooded. And today, because of all the things that are happening and what we’ve been talking about, these businesses have become incredibly capital intensive. The fact that Google raised equity capital, I almost fell off my chair. I couldn’t believe it when it happened. Um, well, you know, this is not a Cisco question. I’m just curious, you know, from, from your seat, what do you make about all that?

Sam Badry

So for these companies to go from free cashflow positive to potentially free cashflow negative, the management decisions is that they must really believe in this forward looking opportunity, right? That has to be the operative assumption. But Steve, I want to go back in time, right?

Hyperscalers doing what they’re doing today. The only difference is the scales at which they’re doing it with CapEx. They’ve always kind of done something like this in the past.

They’ve always been these almost like frontier spenders of networking or IT hardware technology, mainly to push compute into such a highly scalable distributed way to monetize the applications or products that they’re trying to monetize. So this kind of behavior is actually in their DNA, and they’ve actually always kind of done this. The difference today is the scales are bigger, the fundamental models are kind of changing, but if these companies are willing to do this, they must really believe in the journey or the destination they’re solving for.

And I think the way they really view it is they need to do it today. This is where a lot of the key structural positions in this market are being set up today. And this is like the risk that maybe they want to take.

And that is just kind of the new dynamic that I think everyone is trying to adjust towards, right?

Steve Eisman

From your seat, are you seeing pushback on the build-out of data centers?

Sam Badry

So we don’t see it because we aren’t really like the frontline builders. We’re shipping gear to–

Steve Eisman

You’re the shipper.

Sam Badry

Correct. But okay, I’ll tell you, I live in Texas now, right? And there was the longest period of time, it was like Texas is a data center friendly state. Do whatever you want. Do whatever you want, right?

Permits fast, power available.

Steve Eisman

Right.

Sam Badry

It’s actually not like that anymore.

Steve Eisman

Really?

Sam Badry

No. There has been pushback. There is now grid allocation audits.

Steve Eisman

What is it? What’s a grid allocation? I never heard of this grid allocation audit.

Sam Badry

So whenever there are, when you’re submitting a request to the grid for power, that goes through a process, almost like not an auction, but just a queue, a submission queue. That queue has gone so big over the last couple of years because of all the data center construction. Now they’re actually auditing what’s in that backlog, that queue.

And they’re trying to figure out how much of this is double build, double capacity versus not. So they’re trying to figure out what’s real and what’s not. Now, the size of the backlog is enormous.

It is like multiples, the size of what’s actually under construction. So there’s an audit to just check what’s in this number.

Steve Eisman

Okay.

Sam Badry

Yeah. Now, the bigger… So this has actually been kind of normal.

This phenomenon has existed in places like Northern Virginia, which is a very big data center market. It’s existed in other areas too. It’s happening obviously in Texas, given the land, the power, and all the other kind of state policy that’s made it attractive to build in Texas, but there actually is pushback or friction now to building in Texas.

So

Steve Eisman

It’s shocking. It shocked us. Given Texas.

Sam Badry

Correct. It was really a data center friendly state. So just that change, that kind of speaks volumes to where we are, right?

There are probably a lot of different kind of theories out there to why this is happening. Part of it is sound. Part of it is, There’s this theory about water, and I have a fun number for you.

It takes 38,000 prompts to make – sorry, the water usage equivalent of 38,000 AI prompts is equivalent to making one almond. Did you know that?

Steve Eisman

No.

Sam Badry

Yeah, so that’s the magnitude of efficiency we’re talking about. Right.

Steve Eisman

38,000 prompts equals an almond.

Sam Badry

One almond. That’s how much water usage is going in. To make one almond of water, the water taken to make one almond, that amount of water equals 38,000 prompts.

Steve Eisman

Which is not a small number of prompts.

Sam Badry

Not a small number. That is massive. Okay.

Steve Eisman

So these things are, from a water perspective, are very efficient.

Sam Badry

Very efficient. And they can cycle out. So that was like, that’s one form of community pushback.

There’s all types of various community pushback points. Are we surprised to see this? No. I mean, I covered companies like Equinex and Digital Realty going back seven, eight years ago.

And this kind of like not in my backyard or this criticality of data centers, I used to hear about it back then. Today, the data centers are just much larger.

Steve Eisman

So people understand, compared to the data centers that Digital and Equinex built, when you covered them, let’s call it circa 2022, how much bigger are these data centers today? Just in terms of like physical size.

Sam Badry

Yeah. So what was considered big around that window, I would even say like, let’s just say 2021.

Steve Eisman

Okay.

Sam Badry

A big data center facility. And this was like multiple halls. So multiple kind of, you know, buildings.

That was like 100 to 200 megawatts. That was like big, a big deal. You know, we just signed, if there was a company that a victory lap about a big deal, that was like 20 megawatts.

Okay. Today.

Steve Eisman

100 to 200 megawatts. 200 megawatts was a big thing back then. Correct. And today?

Sam Badry

Today, you’re talking about gigawatts. So we’re talking about 10X, 20X, the size. 10X, 20X. Correct.

Steve Eisman

And how much bigger in terms of physical size would one of these new data centers be compared to these 100, 200 megawatts?

Sam Badry

So the sizes of the data centers are not going 10X or 20X bigger. They’re probably doubling, tripling, or 4XXing in size. But the difference is the electrical density, the power density in the facilities are really going up.

So back then in the 200 megawatt world, you’d hear about 35, 40 kilowatts per rack, right? Today, you have roadmaps that are talking about 150 to 650 and eventually like megawatts at the rack level. Wow.

Yeah. This is – Yeah.

Steve Eisman

The engineering complexity to do this must be astonishing.

Sam Badry

You’re talking, you even have different types of power currents, like the 800 volt direct current that’s being discussed. This is like an area I’m not the most well-spoken in, but you even have different power architectures that are being discussed to push even more power into the rack and therefore increasing the density another step function higher. Got it. So, yes, the answer is this is like electrical engineering at peak on peak as far as a field.

Steve Eisman

Got it. Let’s finish up with something about you. So you were a self-sign analyst. You became head of IR at Cisco. What do you think someone with your background brought to Cisco’s IR that might not have existed before, given your background?

Sam Badry

Yeah, I think every company makes some operative assumptions around, or even like guidance assumptions or messaging assumptions around what they think the street wants to hear or what investors or analysts want to hear. until you, you know, they may think they know, until you really have someone from the industry, like working in a bank or even on the buy side, come into the company, you really will not learn all the subtleties and all really kind of like the art that it takes to get corporate communication aligned with shareholder and analyst engagement. It just takes, it’s a process.

And we went through a pretty big process where we made some changes in communication. There are things that we stick to. There are things that we stopped sharing.

There are things that there’s disclosures that we increased in. There’s disclosures where we decreased in. But it’s all about that continuous calibration and not being shy to ask the buy side or analyst community, what do you guys want to hear?

What can you live with? What can you live without?

Steve Eisman

So you would call analysts sell side by side and have conversations about what they would want.

Sam Badry

Correct. I’ll ask them, like, what do you want to see from us? And I’ll have to pitch it internally, see if this is something we can do. We’ll have to assess it. but you can have confidence in me communicating this internally.

Steve Eisman

Okay.

Sam Badry

I actually live with like a very simple quote. I represent Cisco externally as head of IR, but internally I’m a shareholder advocate.

Steve Eisman

What do you mean you’re a shareholder advocate internally?

Sam Badry

If there are concerns, questions, or kind of probing areas that shareholders are, you know, they say to me, hey, can you make sure that, you know, your CEO and CFO knows about this factor? Like we want to know more color about this area. Whenever you can share it, I will communicate that internally.

Steve Eisman

Okay.

Sam Badry

Or are there things that are explicitly said to me like, hey, we didn’t like this or we really like this. Can you do less of this or do more of this? I will layer that into my workflow internally. And as we go through the quarterly reporting, when we’re ready to release material information, we will try to factor and take into account the feedback.

Steve Eisman

Okay, so I’m going to give you the last word. If you had just top of your mind, like what message would you like to get across to viewers about the company and what the environment is like?

Sam Badry

Yeah. So maybe first on Cisco and then maybe as a sector. On Cisco specifically, a lot of things have changed, right? In fiscal year 26, which is the last 12 months that we’ve reported, we had record revenue, record operating margin, and record earnings per employee.

So the DNA of the company has changed and we’ve had to accelerate and become more productive given the opportunity, but also because of the kinds of tools that we are now working with. So this has kind of changed the DNA of the company, and it takes time to completely change the perception of the company. But this is, as you know, an ongoing process, and shareholders need to see quarterly reports, progress, and we have to deliver and execute accordingly.

So that’s the big thing is Cisco has changed as a company in the very short period of time I’ve been there considering the life of the company. I’ve been there for three years. The company’s been around for 40 years plus.

So in a very short period of time, we’ve evolved quite dramatically and we’ve introduced new products and we’ve prepared accordingly. As a sector, what I do is I go kind of around the world and I meet with shareholders of Cisco or prospective shareholders. And they ask me very kind of critical questions like, how do you know your demand is real?

How do you know the demand is durable? Or what do you guys see? Or what do you have to believe?

I think for AI, because so much of that, so much of the tech trend hinges on specifically AI, you need to believe that we are still early from a token consumption perspective. And if I take the token data that’s readily available, some of it is on open router, there you can go on X, there’s hundreds of publications. So much of the token kind of charts or frameworks are showing just explosive growth.

And then when I poll everybody as I travel around the world, I asked them, like, what kind of user would you describe yourself when it comes to AI applications? Are you a simpler user? Do you have a skills file?

Have you deployed an AI agent into production? And just from the questions I ask and the answers I hear, it’s just shockingly people haven’t even gone to the skills file stage.

Steve Eisman

So from your perspective, it’s early.

Sam Badry

Yeah. I almost describe these tools as like when Microsoft Office was available. Some people kind of fought it. They didn’t want to learn Excel or the functions. Then at some point That

Steve Eisman

takes me back.

Sam Badry

Okay. So you remember that discomfort in learning the new tool. That’s what an AI harness and a skills file and eventually an AI agent is like.

You’re learning, you’re developing a new tool that you’re going to use to become much more productive. That tool uses up a lot of tokens. But with this said productivity, you can just get a lot more done.

And today, the main power users are software programmers and developers, which is actually not surprising at all. These kinds of professionals have always been able to At the forefront of this type of thing. They’re able to take things that are kind of difficult and clunky to work with, and they just make them work because that’s essentially their domain.

I think the next cohort is financial services. People are going to adopt this because they have to out of productivity needs. And then eventually it’s going to go into other places, into healthcare, into other parts of the world, into science.

And everyone’s going to have to learn these tools. And the trend basically says, if you are going to use these tools and you’re going to use them in the right way, that equals a lot more token usage. Hence, the greater activity that you might see in all these frontier companies.

Steve Eisman

Sam, thank you. That was great. I appreciate you coming back.

Sam Badry

Absolutely. Steve, thank you very much for having me.

Steve Eisman

And we’re back. So, that was a really great interview because it was an interview with someone who is in industry and sees things from his company’s perspective. So, a couple of points here.

First of all, Cisco’s revenue growth has dramatically accelerated. I mean, I remember when Sam Badry joined Cisco, the revenue growth of the company was 5%. And people were questioning whether they could even achieve 5%.

And in the most recent quarter, the revenue growth was 18%. And much of that,

Steve Eisman

not all, but much of that is from Cisco’s hyperscaler AI-related businesses. I mean, I’ve been saying there’s no question that the growth in AI is for real. How could it not be real? NVIDIA had revenue growth of over 100%.

Steve Eisman

The question is, how durable is it? You know, there are a lot of critics out there. From Sam’s perspective, it’s very durable as long as token growth keeps going up.

And it is going up. And so I think his perspective is that as long as token growth continues to accelerate, everything is fine or everything is good. And we’d have to have something happen that would cause token growth to dramatically decelerate.

But I don’t know what that is at this point. And I think the AI story is real. To me, the risks really are, will there ever be a price war between anthropic open AI and the Chinese models?

And if that were to happen, maybe the story would change. But until then, the story probably continues. Hope you enjoyed that interview.

I certainly learned a lot, and we’ll see you soon.