AI’s ROI Question and the Future of Software Engineering | TSG Ep. 969

Wanna buy a data center? You're watching Text and Gang. Hey everyone, happy Tuesday.

Welcome to this Tuesday's Textron gang. It is, I gotta tell you the truth. It's good to be back in my studio here at Textron Studios doing this.

Unfortunately, this'll be the only day this week. I get to do it from the studio. I'm back out on the road later.

So, um, but nevertheless, the show must go on, as they say in showbiz. Uh, we've got a great text on gang for you today. We've got some of my favorite people with me on the gang.

Let me introduce you to them. We have Mr. FoST, Steven Foskett back from his couple of road trips as well, the one and only JP Morgan.

com or Morgenthal? That's a 20-year-old blog. I just hit my 20th anniversary.

Very good. But I, look, I wanted to give you a shout out. A simple thank you.

Would've suff surprised. Appreciate it. You You Should started writing.

It's like 20 years of writing new, new to Alan, new To me. New to me. Yeah.

So, so you're saying it only took Alan two decades to notice? Is that what you're asking? Exactly.

You know, Alan Took this thing better. Never Gang, right? Yeah.

You know, you know, move fast and break things. Uh, and then next up it, and I'm glad she joined us down in Atlanta for Q Con, but she's back here with us on a regular basis. Again, the one and Only Hope Lynch, and of course, Mike Vard, gang members.

Welcome. Happy Tuesday to you. Thank you all for coming on.

So, Mike, we just can't get enough data centers, it seems. Now we're up to 5 trillion. Let's put our heads, let's let wrap your heads around that.

What does $5 trillion mean? I think it's something like $35 from every single iPhone user. I, I saw that bandi about in the article.

Well, we can't seem to get enough of these reports either. So, JP Morgan Chase is out with one now, and we had, um, you know, Wharton the week before, and MITA couple of weeks before that. But JP Morgan Chase is pointing out that, um, AI investments would require somewhere in the neighborhood of $650 billion in annual revenue just to generate a 10% return on income.

And that would be, or investment, sorry. And then the, that would equate to something like 35 bucks for every iPhone user and 180 bucks for every Netflix subscriber user. But some folks are like saying, well, we'll do better than 10%, and saying, this is good news.

And other folks are saying, this is patently ridiculous. But JP Morgenthal, let's start with you. What's your take on this whole thing?

I look, the, the race right now is about infrastructure. It's, uh, there is a, you know, uh, right now, the way the LLMs work and the demand that people are, are placing on the requirements for ai, it's, it's hungry and it's consuming, and it needs, uh, significant, uh, and powerful infrastructure, especially GPU based. And so this, you know, that to me, when I look at, and I see, and they people talking about bubbles, right?

A big part of this is, is there a bubble? Um, you know, and they're like, no, there's huge upside because we haven't even met the requirements for, uh, what the infrastructure requirements are for the demand. Uh, you know, I, I'm looking at it and saying, you know, there, there is some runway here.

There is definitely a need. But the race is, is, it's unfortunate to see that the race is so centrally focused on infrastructure. Uh, and, and once again, you know, the, the software side of things is, uh, of lower priority, right?

It, it seems like we know, we know from various, you know, contributors into the space that this hardware may not be necessary to the degree that it's being built out. If we focused on the software and making the inference engines, you know, more, uh, better, you know, a greater opti, more optimized, Better efficient optimization efficiencies. Yeah.

Yeah. More efficient. Thank you.

More efficient. Uh, and we've seen that it can be done. We, you know, we had deep seek come out year, you know, months ago now it seems like it's old news, but still demonstrating that this can be done.

So, uh, you know, why, why are we chasing the infrastructure game? And it seems to me, and somebody went, put, put out a, a chart showing, oh, all the money's actually being passed around among a handful of players in the space, and you can actually follow the money. It's the same money.

And so you begin to wonder, and this is where people are starting to question, you know, A, we could be more efficient. B, it seems like the same players are really just trying to ratchet up their profitability. Well, it's, it's a bit of a circle jerk.

That's why we should release the entire Epstein files. Um, just zinger thing. You throw that in there, throw that in there.

But, um, but no, seriously, I am serious. But, you know, I wrote an article, I guess maybe two weeks ago now, about all your data centers at debt. And, you know, un you never know what's gonna go viral.

That article got 38,000 views on LinkedIn, something like that. 37, 30 8,000 views. And so I've been involved in a lot of conversations around it as a result.

The, the story is this, yes, jp, you're right. Is that 5 trillion number based upon doing a lot more training? And how the hell are we gonna do a lot more training when we've already consumed all the data in the internet out there?

We need to do synthetic data and all of these things that are gonna keep that training locomotive humming. Is it based more on inference, which can be more optimized and efficiencies added and is, you know, is, is, is gonna be less intense and from a power point of view, because the problem with data centers quite, you know, this $5 trillion data center thing, it's not about building buildings. We got a lot of swamp land here in Florida.

We got a lot of swamps in Louisiana. We've got plenty of open spaces in Ohio where we can, we can build mega factory data centers. It's getting the power to these data centers, right?

You know, legacy data, data centers. Steven, you know, the numbers here better than me. 'cause I always mess up megawatts with gigawatts with watts.

But, you know, legacy data centers, I believe, uh, run on 10 megawatts or something like that, where one one of These, yeah, there's definitely, uh, my understanding is that a profitable AI data center needs as much as one point 21 gigawatts of power, um, to travel back to, That's back to the future already 25. And, um, you wondering Where you were going with that, Steve? Yeah, exactly.

No, I, it's, it's a tremendous amount. But, you know, the thing that gets me, and this is what we're focusing on on the New Techstrong podcast, you, Leslie, do you need me to do my Doc Brown? It's all about the profitability to me.

And, and I think that that's the interesting thing about this JP Morgan, sorry, not Morganthal, uh, Report. He wishes. He wishes, yeah.

Is that, you know, they're, they're not anti ai. They're not pro ai. This is a very objective report.

And essentially they're just looking at the numbers and saying, look, we're putting all this money in, we've got all these data centers, all this hardware, all this power, all this environmental impact, and, um, and where's our $650 billion of annual recurring revenue? And that's Just on the initial spend, Steve, that's not on the 5 trillion. Exactly.

And they're not questioning, uh, for example, the depreciation problem or the circular financing problem, or even the environmental and power problems. They're just saying, we put this money in. Where's our profit folks?

And that's where, you know, I think that gets interesting. I, I, I don't wanna dodge the question about power, but I do wanna say there are productive applications here, but is there enough applications that as, as it says in this article, for example, that every iPhone user would pay the equivalent of $35 a month for this new service? Maybe, um, or maybe not.

I, I think to put it in perspective, right? Open AI is scheduled, or what are they making about $20 billion now, is their run rate annual, that's revenue, which annual Very optimistic number. That's, you know, the way they calculate that is suggested to be as ridiculous as multiplying their best day by 365 or something like that.

You know, You know. But, but again, I'll come back to it. At the end of the day, I don't know if we could spend 5 trillion on data centers without a good chunk of that going to power generation.

And, and that's gonna require regulatory approval. It's gonna re require some technology improvements, you know, because they, you know, they, they, they wanna knock solar and and wind. But that's our best bet to, to do this short of putting a nucle, you know, a, a nuclear fusion plan in every DeLorean vehicle and parking it outside, you know, the Stevens point about gigawatts, I think it was seven something.

Well, and, and for that point, I'll just say that one point or 122 gigawatts of power, which is what they say in the article, is the equivalent of twice the annual solar power generated in the United States. So could we do it with solar? Yes.

In fact, China installed that much solar this year, really? But, but if we're opposed to solar and opposed to wind and not willing to do it, This, This whole thing is cray gray to JP P'S point. We are gonna build more efficient LLMs, and we're probably gonna use smaller LLMs and software developers.

We'll look at the cost of this and somebody will beat them with stick and says, don't write, don't spend so much money on this stuff. Find a better way to write this stuff. So hope can software save us from ourselves.

What do you say? I, I really don't think software is going to save us from ourselves until we get to a GI maybe, right? Um, but one of the, a a few things that come to mind earlier, you know, we were talking about they're the same players, and I think they're now calling them the super six, right?

Nvidia, Microsoft, apple, apple, alphabet, Amazon, and Meta. And they are around half of the NASDAQ right now. So the money that is being followed, they're not necessarily paying attention to the infrastructure investments.

It is, um, how much of, uh, stock market return are we getting? How much more investment, how many people are continuing to pour money into our coffers, right? So ai, since chat, EPT, um, has made up about 75% of the returns of the s and p 580% of earnings growth, 90% of capital spending.

So now we're looking at what's gonna happen. So I think one of the big debates here, if we not only look at the data centers, but the chips that are being built, both of those are going to be tied to depreciation schedules. So if everyone gets it right, um, this is the build out of the century, it's going to be amazing.

But if they are wrong, earnings are gonna be overstated by so much that Yeah. The market crash will, will be huge. It will be, it will be amazing because all of that excess investment is going to destroy the returns.

And once that happens, everyone is going to pull their money out. We already see people starting out, you know, um, Hope I agree with you right now, Stephen, Mike, you know, Daniel Newman, CEO of Fu. It's one of the most bullish AI I know, right?

He, he really, you know, and he and I, I respect his opinion and expertise when it comes to markets and stuff like that, but even he was on TV last week, I saw it on LinkedIn and I, he, he didn't say a bubble was gonna burst, but he said there are some, some speed bumps, some, some minor things on the horizon. And, and I commented it on his LinkedIn. You know, I, I'll say it again.

Bobby Bakala said it onto the Sopranos. Tony, do you think you hear it coming? You don't, you like that, Jamie?

You don't hear it coming when a bubble burn. And we've all, look, we've all been around the corner once or twice, right? And we know that what you hear when that bubbles burst, so you hear it, it's a momentary, it's a momentary correction as they, they regroup.

It's a, it's just as little speed bump on the way up, you know, to the next level. We're in a new paradigm. It's a new reality.

Can, can I point out one more speed bump that we will surely hit? Sure. That I don't see people talking about is the, it's not like the price for AI is gonna stay stable.

There's gonna be competition, there's gonna be pressure, there's gonna be, you know, laws of diminishing returns. So, you know, even these numbers from JP Morgan or, or shall we say, uh, irrationally optimistic because it's not like I'm gonna sit here and go, well, I guess I'll just pay more for AI because, you know, investors need to be happy. I'm gonna sit here and say, who's got the best deal on ai?

And I'm gonna force people down on pricing time and time again until, well, but, But the problem is, data center build out power generation capability. These are not the kind of things you could turn on and off like a faucet handle. Want a little more hot water?

You Not my problem. I'm just gonna try to force the price down as a buyer. I'm gonna have an AI agent that's gonna help me do that.

But I also think it is interesting though, reading. Um, there are a few warnings in the report though. They, JP Morgan says the path isn't gonna be, you know, straight up and to the right.

And that eventually it may be a winner's take all market that will have spectacular losers. So what if I am all in on, let's say chat, GPT, let's say something turns for chat GPT, right? And suddenly that faucet is turned off and I have to pivot to another provider.

Um, that shift, no one really has talked about. How do you migrate all of your agents and everything you have built around one LLM right? To another provider quickly in, in case there is, Well, in, in all fairness, hope, I mean, I think that's actually a better problem than it was with the cloud, because OpenAI API has become a standard.

MCP has become a standard. Um, doesn't mean they're the best, but many organizations are building to these specifications now. And I, I, you know, I do this, I work with it, I work with these LLMs on a daily basis, switching between them.

I actually have a blog about, uh, or a LinkedIn post about this, you know, open ai. API allows you to switch quickly. The responses from the each of these LLMs do differ.

So you are not necessarily getting, um, the exact same structure or, uh, or data in the same way. But even that's gotten better. Um, uh, many of the LMS are now responding to structural requests, uh, for return data and adhering to it.

So it's saying, Hey, when you answer me, I want A-J-S-O-N, and I want it in this structure. And, you know, chat GPT and Claude or Anthropic and, uh, uh oh, oh, you know, you can use this name. You know, they, they, they all, they're all adhering to that, right?

And, and that makes moving much easier now given the newness the, uh, early phase we're in, in the market. That's amazing. 'cause if you remember, I mean, this is the point in cloud where you had the service meshes and companies like that trying to homogenize and, and nor, you know, the ability to switch or have mul multi-cloud, and you couldn't do it.

You just, it just wasn't working. Agreed. But here, you know what, a couple of people on that LinkedIn post I did came to me with this and said, look, the, the, the training and inference part of this whole equation is, is limited.

It, it, it's a, it's a narrower band. When you look at the entire of AI powered applications, we're gonna run and beyond that narrow band of training and then maybe inference, most of the AI applications don't need that kind of horsepower, don't need that kind of power consumption. And a lot of our existing data centers and and cheaper to build data centers will house these AI apps of the future.

And, and maybe what we'll see is, is training and inference instead of hockey sticking, begin to kind of plateau. I'm not saying that that's gonna happen, and I don't know how that would figure into the JP Morgan 5 trillion number, but it, it, you know, few people did suggest that, You know, when you go visit the wild, wild West out there in the back in the day, you would find, and you still see them today, all these ghost towns and the ghost towns existed because somebody thought that the railroad was coming through that town, and then the railroad went somewhere else. Well, I got a feeling there's gonna be a lot of AI data centers that are gonna be ghost towns.

Could be, could be, maybe we'll convert 'em to condos like Wall Street. But on the flip side, you know, you got towns like Chicago that, uh, that bet on the railroad and, and, and made out big time. So, you know, I, I think ultimately, I, I talked to Daniel actually right after that interview.

And, um, and, you know, I don't wanna misrepresent what he said to me, but, uh, you know, I think he feels like there is a, there, there, and that there is a market for these things, and he's not sure what it's gonna look like, and there's gonna be bumps. But ultimately, I think this technology is too transformative to, uh, not have a huge impact on everything we do. Yep.

Alright. Hey, we gotta take a break here. We're a little over time.

We're gonna come back and we've got, uh, well, we'll keep talking about ai, but this is about doubling down on DevOps and ai. You're watching Textron gang, You've earned it. The spotlight, The responsibility, the weight of teams, companies, and entire industries fall on your shoulders.

Lives depend on your decisions, your home life included that work. You are protected physically and digitally. Nothing gets through your team without a fight.

But in a globally connected world, everyone sees you, including those who mean to cause you and your organization harm. And now home your sanctuary attackers see an opportunity. Your digital front door is wide open.

And what compromises your home can breach your boardroom. Because the devil's greatest trick isn't targeting your workplace firewall. It's convincing you that your personal life isn't at risk.

Black clerk, digital executive protection, defending the new attack surface your personal life. Hey folks, we're back in. There is a lot of debate in the land of DevOps about what the impact of AI is gonna be.

There's a report out from GitLab talking about how, um, well, at least the people they surveyed think that there's gonna be a need for more software engineers, largely because as developers write more code, there's not only a lot more volume of code moving through the pipelines, but a lot of that code is, shall we say, overly verbose, hard to debug and may contain more vulnerabilities than ever. So I guess hope, what's your take here? What's the impact that we're gonna see?

And will we need more software engineers? Or will the software engineers than themselves become? Well, AI engineers, I, I think it's nuanced because we've already seen that there are glut, there's currently a glut of available, uh, software engineers in the market right now.

So many have been laid off from different organizations. Some of the hyperscalers have cut staff significantly. But for the remaining staff, uh, if you consider that, um, let's say I am, I am creating the, I have the raw ingredients for a pie, and I am able to assemble pies much more quickly, flaws and all.

And then I ship them down the line to the team that has to bake these pies and they must bake them. They have no choice. Um, suddenly we need more people to bake just because of that output and more people who may quality check.

But then can AI take on some of that? So it, I, I don't necessarily see it as, um, a straight line to the need for more developers. I think for organizations that plan really well, perhaps they can, uh, no one's I, I still don't think the 10 x developer is, is going to happen anytime truly soon.

But I think, think it does empower developers to create more output, but the quality of the output really is the question. So if, if I am always getting, you know, more pies to bake, and these are, and these are terrible pies, I'm gonna look for another job pretty soon. So, uh, I think the quality of life for developers is probably gonna take a hit as well.

Jp, you wanna jump in here? I know you have an article up on your site that talks about some of this. Yeah, last week I, I blogged about this topic in a, in a roundabout fashion, right?

What I, what I mentioned was, I, I think there's a different pattern and approach coming to how we build and think about software. Uh, uh, you know, the old world was you came up with an application concept, uh, and, and it had to have a specific payback value in order to invest in building that out, whether you're inside the business or whether you're gonna sell it as a SaaS application on the open market, right? It has to be a justification.

I think that goes away. I think, you know, we're, we're entering a tools economy where I want a tool to help me get something done at work. Uh, and I'm gonna have AI build that.

More importantly, what I wrote about in the article was the, uh, the, the value of the zero shot subject matter expert. The ability for somebody who's not a software engineer to take their expert knowledge and codify it, which has, I mean, if you think about that, my favorite example is office space, right? The Bobs, right?

You have a team come in and question and, and capture all the requirements, and then they're gonna turn that into a specification, which is, it is the worst game of grapevine that was ever cursed upon the, the, the software industry or, or the, you know, software engineering industry is that we had people playing grapevine. And, and, and by the time you got done building the software, you know, the fact that they missed the boat that isn't apparent to the user until it's too late and you're now six months down the road, right? And hence, why do we have agile?

Why did it even come about? 'cause people got tired of waiting six months to see the big uncover and go, that's not what I asked for. So, uh, you, you know, what I see happening and what I, and I've experienced this.

I have real world e examples of this, uh, you know, where I, I have taken and coached people who've never built software in their life. They're using Claude and they are building their subject matter expertise into an application. Now, that application is a single page React app that is not reliable, cannot be hosted, should not be used by anybody but that one person.

So if, if that's for me, great, marvelous, it's, oh, everything's stored in your browsing, you're good molded. If it's something you want to share, if it's something you wanna expose, if it's something you want other teammates to use, if it's a SaaS application, it's gotta be hard. But what I've, what have I, well, again, I have real data here.

What I've analyzed is it takes one engineer to do that hardening process and one person QA to test it. And both of those are supported by AI in their activities. Okay?

And, uh, so I, I, I, I did see that there is an opportunity for software engineers to be applied if we see the volume of tools growing and each engineer is applied to one gets one or two tools that they're going to great. Now you're building an ecosystem of tools. And I have a pool of software engineers that I'm using to harden them and make them usable and, and available to others.

And, um, and unless without that volume, I just don't see the market, uh, doing anything that's shrinking for software engineers in the near future. Alan, you know, hope said, you know, this can all work out for organizations that plan really well, you know, well, I'm not used to such a funny knee slapper early in the morning, but, um, what's your take on what's happening here? Because I haven't seen any organizations plan very well in a long time.

Yeah. So JP, what you are saying is, we are going to have more software jobs or some software jobs. If a volume of tools emerges, then the opportunity to grow software engineers to harden those tools also grows.

If the tools market does not emerge, then no, I think we go the other way. I Think it's gonna create more platform engineering roles. Look, I, I've said it before.

I'll say it again. I'm trying to be consistent. 'cause cons consist, consistency matters matters.

Um, AI is going to create more jobs than it takes away. There'll be different kinds of jobs, even within software engineering. It may be software engineering jobs, like you're talking about JP hardening and, and that kind of thing, you know, but to hope's point, it's going to create more jobs because creative people will come up with creative ways of using the tool to make money, to make things happen, to solve problems.

It's, that's human, that's humanity in a nutshell, right there, folks, find a tool, do it better, use the right tool for the job. And that's what we've built our whole civilization on. This isn't gonna be any different, I think, I hope.

Um, but, you know, but what's gonna change is maybe the day to day of what people do, like what the, what a software engineer does now may be very different than what a software engineer does in a year or two or five, right? But there's going to be software engineers, there's going to be people using this tool, and not just in software, but in testing, in security, in ops, in platform SREs and everything else. I I think we, you know, so everyone's on the, the a ai bubble busting now, right?

Everyone's saying the bursting of the bubble and, and yeah, even maybe a little burst. And, but we can't, we can't have this herd mentality where, oh, it's gonna take all these jobs away and software engineering is dead, and this is, it is dead and security is dead. And, you know, what is a, what is AI killing next?

I I don't believe it's gonna work like that. I, I I irrational exuberance. And then, and then you, so bubbles are a unique herd mentality kind of thing.

And you know, that's like saying what, well, what, what, what moves the herd, right? Right. Because, you know, a herd of cattle will jump over a cliff in the stampede, but yet, what is it?

Like a, a bird or something could turn around a herd of something else and make it go in another direction? I wish I knew jp. I wouldn't be sitting around here doing the Textron gang with you guys.

I'd be off on an island. So, Well, I think, I mean, there's a bub, most bubbles die because demand dries up. Everyone who has bought in has bought in.

Everyone who is interested has already taken their stake. No, But so you're talking about the financial, so, and, and that's an important thing here. There's two pieces of this.

There's the financial bubble, which hope, I think you nailed seven, seven companies responsible for 50% of the GDP, 80% of the stock market, blah, blah, you know, all of these things. But then there's the technology underlying that financial bubble. And that could also, you could have a technology bubble.

AI's going to take all the software jobs away, AI's gonna, you know, take your job away. But underneath, in, in, in, in my lifetime anyway, financial bubbles come and go. Technology sticks around.

It may not happen as fast or as big as a bubble predicted, but the internet did change the world. The cell phone did change the world. The cloud, in many ways changed the world.

AI in a lot of ways is gonna change the world. It may not happen with $5 trillion at data centers in the next five years, but it's going to change the world. It's gonna change the software developers' job.

It's going change the op person's job. It's gonna change the CIO's job. It's gonna change the CEO's job, for sure.

Make no doubt about it. To JP P'S point, I think, you know, it's gonna be another instance of where people are gonna perceive that Silicon Valley overpromised and under-delivered once again. And eventually they'll get it right.

But it ain't happening in any timeline that justifies a bubble investment. Fair enough. All right, if we have nothing else, Steven, you have nothing else on this one?

Well, I, I guess the only thing I would add to this is, um, we recently had our AI field day, and I was quite impressed and, um, surprised by the enthusiasm that the software development community has for AI tools. And we had people like Calvin Hendricks Parker, who is a, you know, one of the best and most, uh, you know, on the forefront. Developers I know, uh, enthusiastic, really downright enthusiastic about these AI coding tools.

Not because he thinks that they're gonna replace his job, but because he uses them and makes incredible use out of them. I mean, you know, and, and I, I suspect that there is a lot of on the ground enthusiasm for this technology, even though there's a little, maybe some trepidation that it's gonna hurt, uh, employment or that it's gonna somehow change the, uh, the, the picture. Um, I don't think it's so bad.

I think, uh, you know, at the end of the day, this is some useful technology that, uh, people are gonna use to, uh, do their job better. I think there's a lot of software engineers out there who are aggressively looking for AI slopping code and kicking it out before it ever gets into production environments. Great.

And that, and that's part of that hardening kind of thing JP spoke about. Let's take a break here. Let's come back.

We're gonna talk our next C block, Discover Textron Group, the epicenter of tech innovation. We are your go-to for reaching IT leaders and practitioners worldwide. Our secret impactful content that sparks awareness, engagement, and top quality leads with us.

You'll access editorial websites, streaming videos, virtual events, custom content analyst research, and more. Join our satisfied clients. Let's revolutionize your tech journey.

Contact us today and tell your story to the world in the most powerful way with Textron Group. Hey folks, we're back. And we talk a lot about emerging technologies on this show.

And one of the subjects that we seem to always not quite give enough of its due is networking. Because, well, there is this thing called latency, and it tends to impact everything we do out there. So, Steven, I know you guys just had the networking field day.

What are our good friends in networking saying all about this? 'cause I suspect maybe they're having a good laugh. Well, yeah, I, you know, I, I think that, uh, the big upshots from networking Field day, uh, a few weeks ago was the fact that, uh, networking is, uh, it remains an incredibly important component of the new data center.

Uh, there are a lot of companies that are working on some interesting things. You know, we can pretend that it's all about ai, but AI is built out of the same nuts and bolts that everything else was, and it all remains really important. Um, I've had the same conversations with storage companies as well.

Uh, you know, you can look at these industries as having been dwarfed by the, uh, purported AI industry, but at the same time, AI is driving these things forward like crazy. So, you know, for example, companies like Cisco have benefited massively from the AI infrastructure build out, uh, as you might expect. You know, you gotta connect this stuff somehow.

And even though companies like Nvidia, uh, Broadcom, you know, are in there trying to build, um, let's say, um, less, uh, branded networking solutions, and, and they've had a lot of success at that too. Uh, Cisco's still selling a heck of a lot of hardware into these AI data centers. And of course, everything else still requires an awful lot of networking hardware too.

So, so that was one of the big takeaways from, from networking Field Day. Um, but the other aspects that were interesting were, um, sort of not net infrastructure for ai, but AI for infrastructure, essentially the ways in which AI can help networks be better and solve some of the problems. So things like, uh, data governance and building security and building secure networks and operation, you know, operating these networks and monitoring them, managing them root cause analysis.

How do you figure out what it is that broke the network? All of those things are greatly benefited by ai, to my point earlier in the show that, um, there are really practical applications for this technology. And sometimes I think we, we get blinded by these huge numbers, trillions of dollars and the world's, you know, first $5 trillion valuation company.

And we forget that, uh, that $5 trillion valuation company is spinning off, um, useful technology that's helping build new companies and helping rebuild companies. Even names you've heard of, you know, Nokia was there not talking about, uh, you know, those old cell phones because they don't make those anymore. But talking about, uh, network management and network operations where they've become a leader.

Um, you know, we had, you know, companies, you know, there you graph in there talking about building network as a service with security and data governance and, and maybe these aren't the things you would expect from networking field. A maybe you think it's gonna be all Cisco and, and high speed switching and stuff, but there's an awful lot more networking and there's an awful lot more spinoffs from AI technology. And, and those are all having a big impact too.

To your point about that, we were at CubeCon last week, and, um, ISO was there, and they were an arm of Cisco, and they were talking about a, you know, a study they did. It wasn't huge or exhaustive, but one of the things they pointed out was that networking people today have on average, like, you know, six plus tools that they're trying to navigate. And like, their number one issue was observability and troubleshooting followed by distantly everything else.

And so, you know, to your point, I feel like the networking people are kind of flying blind a little bit with all these different nodes, and we just keep adding nodes out there. But, you know, how much stress are they under Steven? Yeah, anytime you have upheaval in the, uh, infrastructure stack, you have a need for, uh, processes and tools and capabilities to try to get your hands around that.

And, and, you know, we're seeing it in storage, we're seeing it in networking, we're seeing it in servers, we're seeing it in data center management power, you know, all of these spaces, you know, in many cases, these systems were built out by people who aren't infrastructure domain experts were they're built out by people who just needed something to get the job done. In many cases that was supplied by a partner or a vendor and, and these things, they may be great. In fact, I would, hes, you know, I would wager a bet that the canned network that you buy as part of your overall, um, you know, AI solution from a major provider like HPE or you know, an NVIDIA partner or somebody, that's gonna be a pretty good network.

But that doesn't mean that it's manageable. It doesn't mean that network is operations are ready to handle it. And that, to your point, is really their concern.

How do we get in there? How do we manage this? How do we observe this?

How do we extract metrics from this thing? How do we integrate it into the rest of the processes? And that, of course, opens up a huge opportunity for, for folks in these industries.

Yeah. Hey, hope. Um, you know, one of my favorite stupid jokes is, what's the one thing that an IT admin and a developer can agree on?

It's the networking guy's fault. Yeah. Can we, um, you know, bring in networking in the platform engineering and maybe have one big happy family?

Or are we always gonna have these networking folks in a silo somewhere? I think, I think there has to be a bit of overlap just because of how AI will become part of the plumbing, right? Uh, the platform engineering team can't be divorced from that.

They need to have an understanding of, um, how it runs the capacity of the network. Uh, there are so many concerns there that I, I just think that they can't. So even if we look at, um, how they might now be looking at traditional AIOps as, as a platform engineer and understanding, uh, what those tools can do, then you're looking at ag agentic ai and the capacity that that requires, um, that is a platform engineering problem in search of a solution in collaboration with the Network.

I don't know, jp, what's your thoughts? I mean, are you playing around with networking in your role these days, or is that something for somebody else to worry about? Yeah, I just haven't had the opportunity to around, you know, I, I focus more on the software side of these things these days.

My, my struggles with, with networking is, uh, you know, being places where ports are closed and I can't get out. That's it. And looking for ways to get around it by tunneling out.

That's about my experience, uh, with networking, uh, is, you know, is, is home networking and, and, you know, just trying to overcome people who are shutting stuff off on. Yeah, Steven, I think this is gonna get worse before it gets better. And this is my theory is that a, we are playing around with more data than ever and shipping it across these networks that are probably getting a little bit clogged in terms of the amount of bandwidth that we need to consume.

But two, we're pushing more workloads out to the edge. And, um, latency is more important than ever. 'cause we need to process that data at the point where it's being created and analyzed and consumed.

So are the dynamics of networking changing? Yeah, absolutely. And you're absolutely right that it's all about na net latency.

Uh, the more data you have, the more, uh, latency starts impacting, uh, user experience and performance, which means you have to start shipping data out, which means you need better data management tools, better data movement, uh, better data security, uh, better network management, more network performance, all of these things. Unless you can invent a time machine that allows you to transfer, you know, gigabytes or terabytes of data around instantaneously, we're just going to need to, uh, have better tools and better technologies to enable us to meet the demands. And I, I think that what's gonna happen ultimately is that we're gonna see a greater adoption of distributed applications, um, as we talked about earlier, you know, more and more of this AI processing is gonna happen closer and closer to the edge, closer and closer to the application or the end user.

And that's going to drive, uh, demand for really more nuts and bolts tools. Not these sort of big high profile, uh, LLMs, but, uh, you know, real nuts and bolts about how do I move and manage data? How do I build networks, how do I monitor them?

How do I secure them? And that's a big opportunity in this industry. Absolutely.

You know, to a certain extent, it's still about the picks and shovels, man. Yep. Right.

Well, that's where the money is. You know, uh, look at Nvidia. They, they, they, they are the world's greatest manufacturer of picks and shovels today.

Mm-hmm. Well, It, but Cisco, but Cisco's no slouch. Oh, no.

Yeah, There you go. So at the risk of starting in another bubble, are the networking vendors gonna be the ones that enjoy most of their revenue benefit from all this? Alan, what do you say?

It's, it's picked the, the, the guys who sold the picks and shovels made all the money in the gold rush. And, and, and I think it, it, it's still true today and you know, more true than ever. I I do, but here's what I think we need to be cognizant of, just because the networking guys aren't talking about the latest, greatest LLMs and some of the other things, you know, that this AI revolution brings, doesn't mean there's not real innovation going on in the networks, right?

There really is. I mean, you know, whether it's, you know, fully implementing 5G and talking about six G and, you know, uh, fiber and, and hundred gig Ether, and, you know, and all of these things we are, these guys aren't standing, still waiting around for AI to kick their butt or, or to give 'em a shot in the arm. They continue to innovate constantly.

Well, you know, somewhere I once wrote, you know, at the end of the day, uh, the network owns the Keone for all of this because it's your, it's your limiting factor moving data and the speed at which you can move data will control exactly what you're able to accomplish. Here's my DevOps stream for all of this. I think that we wanna get to the point where you build the plane app and you through a natural language interface, describe your intent for the latency, and then the system automatically configures it and makes sure it stays that way.

Steven, is that ever gonna happen? Well, We can, we can keep working on it. And that's, uh, gonna give us all a job for the rest of our careers and all the people listening, We'll end it with this.

Do a Eyes dream, and if so, what did they dream about? You'll, you know what, this is a great text on gang JP Hope, Steve and Mike, thanks for joining us. Thank you for joining us.

Hey, Steven mentioned that networking, uh, field Day, Steve, that's up on, on, uh, the YouTube channel already. Yes. Oh, yeah.

Every, uh, every one of the presentations up there, uh, along with, uh, now presentations from, uh, CubeCon as well. Absolutely. Or you could catch 'em on text Drunk TV on our OTT app or the website.

Uh, we've as usual got a full day of text, drunk tv, not a full day, but a good couple hours of text Drunk TV following the gang today. I think some of our CubeCon stuff might be up on there as well, if you didn't catch it, live streamed until tomorrow. On behalf of all of us here, this is Alan Shimmel.

We're out.