The Hidden AI Layer Nobody's Talking About

What is going on investors hope the guys are doing well out there welcome back to the equity empire podcast where like look the topics and the conversations on this show really come from you so if you have a topic or conversation that you want me to have send me an email team at my equity empire.com and we'll be sure to talk about it now today's video actually going back to a podcast that I recorded a week or so ago on the trader Merlin show I peered on that show is a kind of a the broad topic was give your outlook on 2026 and on there I said like look right now as we sit early 2026 software is in the dumps everybody on Wall Street is convinced that you know everybody's going to be able to vibe code their own software the vibe code their own CRM the vibe code their own calendar everybody's going to just be using software that they create themselves not just from the kind of the consumer level but all the way up to the enterprise business level where if you look at this enterprise software accounts for a very small percentage of spending at a large organization so if they were to undertake the task of replacing software at the organization and enterprise level that actually wouldn't be saving a lot of money will put all of those arguments aside one thing that I believe about 2026 is everybody is looking at software the exact same way and that there are opportunities going forward as it relates to software and one area one specific area I'm going to be talking about on today's show is what I'm going to be calling the infrastructure layer of software and this is a layer of software that most investors and certainly most consumers in the average person listening to it might not interact with in a direct way now when you are on Instagram or you are on even listening to this podcast or doing anything on your phone you're interacting with the infrastructure layer of the software you just might not be aware of it things like database things about cybersecurity things that help the app run better and run more smoothly and handle different types of processes this is a very exciting layer of software that is because what we are seeing in the age of AI is more and more of this layer of software is required because now it's not just humans interacting with the device or the software interface it's not just a human opening up a Google spreadsheet or a PowerPoint slide presentation now the AI is making the PowerPoint the AI is going into documents and sheets and PDFs and writing files making changes saving those to a database so we have exponentially more opportunity and opportunity means demand for this layer of software which again most investors don't understand and it's not to their fault unless you're into computer engineering or designing websites or designing different types of mobile applications unless you're on that side of things you probably don't know what data warehousing is you probably aware of cybersecurity but are we all aware as investors as more and more agentic agents and AI doing work rather than the human it obviously can scale to a number and to a degree where most investors I don't think are necessarily pricing this in now when I talk about the infrastructure layer I want to lay out something that Jensen Wong said earlier this week in Davos at the world economic forum he was there and he gave like about a 30 or 40 minute interview and I thought one of the more interesting things that he said was he condensed the AI compute layer down to five separate layers and I'll tell you the five layers that he laid out I thought that was interesting that's great that Jensen Wong did that he simplified it in my opinion down to the five simplest layers well I think his investors though there are many layers in between the five layers that he laid out and the layer of software I'm going to be talking about today and that's infrastructure software it actually sits in between one of the layers the main we'll call it the like the fab five or the starting five layers of AI infrastructure software those are great from a starting point and most investors these days are aware of those five but I'm here to tell you there are different layers in between them and I believe one of the most critical ones of them all is actually the infrastructure layer which we'll talk about today so let's get into it so layer one that Jensen Wong laid out in his interview was really power so these are the utility companies these are power companies maybe nuclear if that's available in the country or or the region I would argue to there's a layer zero where you have the raw materials that go into these things the precious metals the earth rare earths the supply of the energy whether it's natural gas or it's coal or it's nuclear or it's solar there is a base underneath the utility where obviously as layer one C significant upside demand so do the layers underneath of that that'll be a common thing is we move through this but layer one according to Jensen Wong were the utility companies there's certainly great investments there now layer two is where Jensen Wong's company sets this is at the chip level the Nvidia's the AMD's the broad comps you could probably throw in as well a layer just underneath them the chip manufacturers like Taiwan semiconductor and then you also have the equipment makers that go into the manufacturing of these chips now would be a sml applied materials even the memory companies like micron then there is a layer three so layer one you have the power just the raw power required for these things layer two or the chips layer three Jensen Wong says is the data center this is Amazon's AWS Azure Google Cloud Oracle core we have many of the smaller neo clouds that have popped up all fall in this layer three that is because they are buying the chips then they need to power their data center using layer one and that is the utilities then comes layer four this is something that I think the average investor is starting to wake up to in the early days of AI we weren't really sure where companies like open AI and in the topic and where these large language models ultimately stacked themselves up inside of the AI ecosystem more and more as we move forward we are starting to realize that layer four is the large language model that sits on the cloud and that is being built upon and that the large language interface whether it's the chat bot or if you're on cloud and you're using kind of a coding environment whatever that happens to be or if you're inside of an application that is essentially a user interface to interact with the AI layer four is the large language models it is part software but it is also tightly integrating inside of hardware so you have layer one the power you have layer two this is the chip layer the Nvidia's layer three you have the actual data center layer four you have open AI and a topic in the large language model and then finally layer five is the software stack that we ultimately interact with and this is the area of software layer five that has been absolutely decimated over the past year or so with sales force service now many others not performing well now there are some exceptions to this layer of software Palantir probably being the shining example however if you really dig into what Palantir does Palantir actually sits a few layers below just layer five now it'll be come something to really think about and understand when you start thinking about software companies like Adobe and Salesforce has really seen their share prices decline over the past year or two in a lot of ways is because those companies are layering AI onto some existing services and yes I think they're making them a lot better and in a lot of ways helping those companies continue to keep up with different types of competition or even the vibe coding version that you could potentially make of their software and services and so there's an area of software that has been really beaten down however if you look at the performance of like a Palantir and one that doesn't get talked about as much is like a Shopify where they are essentially kind of a software layer of ecommerce that allows people that want to sell things online to use Shopify and there are other competing software in the marketplace but for the most part Shopify has a dominant position if you want to set up your own ecommerce website and sell stuff that stock has been doing particularly well Palantir obviously doing quite well And I think we can have a broader discussion about more of the software layer, layer five, if you will, of the infrastructure layer of AI. I think we can have that discussion another time. But right now, I want to talk about the infrastructure layer of AI, which I don't think gets enough discussion and enough attention in this new age of AI and what it's going to do to the demand for the services that a lot of these companies happen to offer. So where I'm putting the infrastructure layer of AI is between layer three and layer four. I'll remind you, layer three is the actual cloud computing layer. This is the Amazon AWS. This is the actual data center. The data center obviously contains chips made by Nvidia or AMD or Intel or usually all three combined. And then layer one, there's some electricity powering the grid and powering these chips to do whatever I'm asking them to do. So that's layer three. Sitting just above that is layer four, the open AI's anthropics. Many of these companies are investing in their own chips and investing in their own data centers, particularly when you get up to like the Googles and the Amazon's and the Microsoft's there developing an entire stack. So between layer three, the actual physical data center, you have the large language model sitting on top of that. But I believe there is a critical infrastructure layer in between three and four. We could call it layer three point five if you want to. This is an infrastructure layer, which as I continue to dive into how people are using AI on a daily basis, not just end users, but also enterprises and startups and all the way down to your well established companies like Costco and Bank of America. I'm constantly trying to find data points and trying to understand how AI is making its way into business because that's ultimately is where most of the profits will be made, at least in the shorter term. The more I'm digging into this and the more I'm experimenting with these tools and also seeing these tools continuing to evolve a year ago, AI was largely just a chatbot that you interacted with and it would do one task at a time. Now we're getting much more into where the infrastructure is sitting on the device level. So a co-pilot is a good example of that. That is sitting on my Macintosh desktop with access to my browser to my files and I'm sure they'll eventually open this up to the entire computer. Apple is likely wanting to integrate something like this Microsoft with their co-pilot and their Microsoft Windows PCs as well. We're giving AI more and more autonomous control to where they are able to write and read files and just execute things in the cloud or on the device that is creating exponential more data and files and things that can be stored and remembered and called back and just the amount of code that is being created, just the amount of work that is being done by the AI is like multiplying exponentially by the month and the day. This is why when you look at memory companies, memory companies are absolutely stacked up for the next year or two in terms of orders and demand because this data needs to go somewhere it goes on to memory. When we heard from Intel earlier this week when they reported their earnings, one of the interesting things that they said on their conference call was that six months ago they weren't necessarily expecting huge amounts of demand for what I'll call cloud CPUs. These are your CPUs that have been running the cloud, the Amazon AWS's before you had AI, before you had AI, you basically ran them with what we would call maybe souped up computer chips inside of the data center and that could deliver anything from a Netflix video to the cloud that we knew it before AI. Well six months ago the cloud services providers were like, yeah, we're spending so much money on Nvidia GPUs and I'm paraphrasing here but this is what was going through their minds. Six months ago was we're spending so much on these cloud AI GPUs that maybe we don't have as much money to invest into the traditional club but what they are seeing is that through agente coding and through all these agente workloads where you're going to your computer and you're saying, hey, research this, do this press enter and then it goes off and do this does some of that work. A lot of that work is running through general compute cloud. So your CPUs it's utilizing that on the device as well. And so what Intel saw was there's this massive expansion now at what I would call layer 3.5 where this is not the Nvidia GPU. This is not the AMD high end MI 350 X GPU. This is coming all the way down to your normal memory, your normal CPU and that is showing me that the underlying demand for what I'm calling infrastructure 3.5. This is going to be off the charts and this is the software layer that is going to surprise to the upside in 2026. So of this group of software, I can kind of break it down into four pieces. The first piece is what I would call kind of like database and maybe data warehousing. This is where all of the data is stored. So as you interact with something, your inputting data, your writing articles, your taking photos, it all needs to be stored somewhere. And this is common. We understand this as humans and in a lot of ways we take it for granted. But one thing that I think Wall Street is now taking for granted is it's not just humans creating the data anymore. Now you have these agents out there that are creating vast amounts of data and they're doing it very quickly. They obviously can work at the speed of multiple humans. Work is getting done that used to take humans weeks and months. It is being done in minutes or hours. And so we are stacking more and more of this data on top of itself. So the key players in here in terms of kind of public companies and maybe maybe one major private company. Is obviously Oracle, Oracle, you know, basically the king of the database, if you will, certainly on premise. So if you are a hospital or something like that, you are almost guaranteed to be using some type of Oracle system. And obviously Oracle's business over the years is really expanding. And then are obviously getting all the attention on the data center as well. I tell you what, if they hadn't gone so aggressively into data center and maybe they kind of picked their spots there. And I'm not saying that that's the right move, but let's imagine that they did that. Well, I think you could be getting much, much more attention to their database and what's going on over there. Oracle remains interesting, although again, based on the amount of tension the stock has gotten and kind of its rapid rise and then pull back, it's in a slightly different category as the rest of these. Now in terms of kind of data and data warehousing, you have more pure plays in the market. No flake is one, Mando or MongoDB. And then you also have Databricks on the private side. What I'm noticing on the private side with Databricks, Databricks is having no problem raising more and more money. In fact, they just did a recent fundraising round as well. So the smart money continues to funnel money into this area of software. Whereas I'm not really seeing that when you look at the big deals across software, it's all like AI native stuff. It's the maness acquisition that met him made Nvidia has made several acquisitions, even ASMR made an investment. It wasn't an acquisition, but they made a major investment into kind of a year, we'll call it a European AI company. That's where I'm seeing the major investment. It is happening at layer four of the actual large language model layer of software where I'm seeing private money continue to flow in in the software category is not at the end use space. So the Adobe's anybody trying to compete at that level or Salesforce or service now not seeing money flow there and you're seeing it in the public markets as well. But where I'm seeing the private money continue to funnel into is what I would call layer 3.5. And part of that layer is data and data warehousing and Databricks sits there and they are getting funding and that's an interesting layer. Now you have another layer here which is harder to conceptualize, but you need to monitor all of this. Certainly if you are a solo entrepreneur or you're a team of two or three, you can monitor all of your agents at once. But once you start to expand this across an enterprise and by agent, I think you know some of you might be listening to this and maybe you don't have a job where you're utilizing this. Or it hasn't been implemented or you just have something where you are not going, maybe you're a teacher. Okay. And you've got a room of 30 or 40 kids. You're probably not going to you know implement a ton of a, a genetic workloads into your work environment. So I can totally understand that. But what's happening across these enterprise? So let's say you have a business and you've had you get a bunch of customers that kind of come in and out They're not super sticky so you constantly have people coming in you have people leaving you have people that are kind of warm to your Products they've maybe signed up for a lead form or they've attended a conference that you had so you have all these emails You got a half a million emails and you know you could sit down a human and have them go through one by one on each email But you know by the time you got through the half million emails the whole year would have gone by and who knows if you would be profitable All of that well you can set up AI agents to start emailing and having conversations with all of those 500,000 people all at once and putting them in different segments and saying this guy just needs a coupon code This guy needs this feature added and then he'll sign up and so you can have these AI agents doing all this work for you in the background Now you have to set all this up you have to understand how the system all works And that's one of these layers that I think is being underestimated by the stock market is this kind of Observation layer of the agent layer this sits above the database layer Where all of the data is just being collected all of the emails all the interactions that's being stored Well, now you need to observe the AI and make sure it's not going off the rails And maybe make tweaks and changes to the process a lot of that is going to have at the kind of the monitoring level You've got data dog in this level. You've got Splunk, which is a part of Cisco There's several companies in this layer now to layer number three within layer 3.5 is you have the security level And this is where most investors think when they think about kind of this in between layer of software Where it's sitting on the database and it's protecting for hacks and different types of malicious injections into software and identity and all those types of things it is security and we're seeing the big players here the crowd strikes the Palo Alto networks even octa and Fortnet cyber arc There is a consolidation in this industry because I think what they are seeing is they're all competing for the same kind of maybe 500 to a thousand clients and there's so many players in cyber security because there's all these edge cases and use cases and Different price points and different capabilities that the big guys the crowd strikes and the Palo Alto's in particular are seeing that they could Could solidate and create a more complete product in one package and the larger enterprises are spreading out to where yeah They have sales organizations. Yeah, they have a cloud computing division. Yeah, they have a consumer and a B to B division They have all the companies at the top are just getting their hand like think about Amazon How many hands it has into so many different things so you can imagine cyber security over at Amazon needs to touch both the physical and the digital and the business to business and business to consumer and government to business and all these types of things And so it becomes more and more difficult to serve those customers if you're the Palo Alto and the crowd strikes if you don't have a complete offering and so that's what we're seeing There's certainly some opportunities. I think in cyber security. It sits in its own I would say its own category here where it's not it definitely is benefited from AI But I think when we look at the Genetic workflows that are happening and the fact that these computer systems are like a hundred or a thousand people working in tandem creating so much more data and production That yes, I think you'd comfortably say that the database data warehousing layer of AI is Here to stay it doesn't matter what software wins the software is going to need database data warehousing in those types of things Cyber security sits in its own special place because that obviously will evolve with AI but it obviously has Some benefits that are not directly tied to more and more AI usage in some ways it creates more risk and liability for these companies Because if the AI gets so good and it becomes really really good It could tarnish the reputation of a cyber security company And so that's always one of your risk if you own them if there is a major outage and we've seen these across all of the cyber security companies There's always some type of headline case It doesn't typically bury them but from an investment standpoint it can create a cloud over the stock for a while now The other layer of this software that a lot of people don't understand is Now you have your data and so you have all your customer data and maybe you want to tie that with something some other piece of data over here Maybe it's statistics about where all these people live So you have all these all these customers you know where they live You know Their address you know their phone number you know their history with you But let's say you want to layer some data on here and they're from different countries around the world And you want to target the users that you have in your database that live in a particular region of the world maybe it's they live somewhere where it's sunny All-large percentage of the time you have some product coming out some suntan lotion or something that benefits people that are in the sun Well, you obviously wouldn't want to email everybody on your list if they lived somewhere where the sun didn't shine very much And they lived in Alaska or the sun actually shines a lot in Alaska But somewhere where the sun doesn't shine a lot maybe the north pole or something Where they maybe don't need a bunch of suntan lotion you probably don't want to be emailing those people as much as the people that live in Miami or in the Caribbean and so been able to layer on some of that data which you don't have you have to go to weather data or demographic data on another website you connect those two pieces of data usually through what's called an API now There's other types of things that are emerging in AI and that's other types of connectors Which I think are called model context protocol or mcp I've played around with them a little bit. They worked really well very similar to an API where somebody has data Sitting over here and it allows you to connect to that data service and pull the relevant data and with the right computer code You could model that and map it on top of your customer data And then you would know all of your customers that are sitting in Sunshine soaked regions of the world so you can send them marketing materials on your new product That deals with just sitting in the sunshine. So there's this layer of software that I would call kind of API integration now some of this is being done at the large language model level Some of it or a lot of it is being done also on the data level So if you are a website with a lot of data on it then you are Integrating your own APIs and the models make it very easy to kind of call this different types of data But a Salesforce acquired mule soft which does this and in a lot of ways These APIs in these connectors just make it more and more easy to develop AI applications That can kind of run autonomously and combine and stack skills if you will And it just creates more and more What I would call this infrastructure layer that is rooted in the database And so those are the four areas Within what I think is going to be the key pillar of software infrastructure That a lot of investors are just tossing out right now They're looking at companies like Salesforce and ServiceNow and Adobe And they just think that is all software that every software operates the same and they all will perform in tandem But we've seen that's not actually true You do see the volunteers and the Shopify's and a handful of other software stocks that have actually done Particularly well over the past year That is because there are aspects of the software trade that are going to be massively benefited by AI I believe the area of software that people are not paying close enough attention to Is what I would call the infrastructure layer This sits on the data center Underneath the large language models Allowing the large language models to store lots and lots of data It allows you the user of Salesforce of Palantir of ServiceNow or any of these offers It allows you to input your files in your photographs and all of the different things that you might have from a resource perspective Onto a database onto a system that organizes it for you And then now we have agentic workflows really coming to the mainstream and I know agentic workflows gets tossed out there And it is a terrible buzzword But what I am telling you and what I'm describing to you is happening across organizations And it is going to just exponentially get more and more spread out between a wider number of companies Companies that had tasks inside of their organization that just simply were not doable a year or two ago It was too time consuming to put a human on it. It wasn't profitable affordable enough to put a human on now because the computer works at a much faster speed and it's a lot cheaper. Well there's a lot of work that can be done in an organization that simply couldn't have been done a year or two ago. Things like cleaning up code. I have heard from a number of different businesses, both private and public, that they've finally been able to go into their code base and rewrite a bunch of their code. That was too expensive and too time consuming in the past. Now they're able to do that. I saw stats for the App Store, the Apple App Store. For many years, the number of apps being published to the App Store had actually gone stagnant. That is because mobile app development was very time consuming. It was very expensive and that limited the number of apps that could actually be made. It had to generate some type of profit or the person building the app had to sacrifice a lot of features in order to keep it under a certain budget. That is no longer the case. If you look at the trends, certainly over the past six months, exponentially more apps are being published into the App Store. It has or AI has created and benefited the development of apps has shrunk. So instead of costing hundreds of thousands of dollars and weeks or months or a year to develop a mobile app, that now all can be done in a week. And I know this because I literally just did it over the past seven days and it only took me seven days and eight hundred dollars to develop the equity empire mobile app. This is a mobile app that contained all of the content. It integrated some live pricing. It has all of the videos and embeds those inside of the app. It pulls up AI summaries of news. It allows guests who don't have a premium subscription to have an option to log in. There are some features on the app that I'm very happy with. I have never designed a mobile app by myself and I certainly wasn't in the market to do it. But I saw some people achieving some amazing things with AI. So I decided to give it a try and in seven days and under eight hundred dollars total, I was able to submit the equity empire mobile app to the Apple App Store. Now I still have to get improved. There's still some testing. The last 10 to 15% is sometimes the hardest hurdle for an app. But given the fact that I literally only had to spend a week on this app and yeah, there was a lot of hours and man hours into it. I spent many, many nights staying up till two, three, four in the morning. I was working on this essentially in the after hours, if you will, after my kids go to bed, after the stock market has closed. There were a lot of hours in there, but it's just one example that the price and the time to develop software and to launch things has compressed so much. I don't. I know. It's not that I don't know. I absolutely 100% know that Wall Street doesn't appreciate this. They don't understand it. The average retail investor, particularly one that doesn't understand software, never is designed it himself and there's nothing wrong with that. We're not degrading these people. I am just saying that a large percentage of Wall Street and a large percentage of investors have no idea what is about to come that you can develop mobile apps me myself. Not alone a team of five, six, seven, ten capable people. I don't have a computer science degree. Imagine if I did, you can develop and launch these things and ship these things in seven days instead of prototyping and having meetings in the past. You would prototype and have a meeting prototype and have a meeting. You'd have a roadmap. Now you can in the time it takes to set up the meeting. You can ship the product. I don't think people have any idea what this means for software. And so yes, the Salesforce and the dobes, they've got a lot of work to do. But I describe to you the layer of infrastructure software that I think is going to have the best year in 2026. This is the infrastructure software layer that encompasses database. It encompasses monitoring those databases and the agents. It is cybersecurity and it also is what maybe is called API or integration of all of these systems. This layer of software is going to be doing absolutely fantastic. I don't think it's priced in and I think it creates an exciting opportunity for 2026. That was the equity empire show. Remember that if you have a topic or something that you want me to talk about on the show, you can send me an email team at my equity empire dot com. Would love to hear from you. Love your questions or your show ideas. Send me an email team at my equity empire dot com. Hopefully you've enjoyed today's podcast. We'll see you again very soon in the meantime. Good luck with your investments.