The SaaS Situation Is INSANE

Added: 2026-05-15

[00:00:00]The SAS situation right now is insane.

[00:00:02]Most SAS stocks are down over 50%. While the rest of tech is at all-time highs today, we're ranking the biggest names in SAS. For each stock, is AI a threat or a tailwind? Drew Cohen left Goldman Sachs, left Capital Group to start Speedwell Research, and now he manages money as a registered investment adviser. I'm Verun. I sold my startup to Google. Let's get into it. The first company is Adobe. I think Adobe is one of the more interesting names because there are clearly a lot of risks to the creative space and creative asset generation. But Adobe is not in the business of creative asset generation.

[00:00:39]They're in the business of creative asset modification and keeping that framing in mind. It's actually a little harder to disrupt them than many people think. And so we can broadly break out their business into an enterprise business where they're serving Fortune 500 companies. It's brand focused firm stuff. People like Nike, Coca-Cola, they're going to all use Adobe and a lot of them too also use Adobe Experience Manager which is not just about creating uh the assets which happens in the media business but it's actually about delivering those assets AB testing them all of the customization that goes on and delivering that to the journey and analytics thereafter. And so keeping that framing in mind and we don't actually know how much of the business is going to be this consumer segment which is more at risk and also keeping in mind that these are not recommendations. This is just off of the AI risk itself. I'm going to put it in the middle. I'll put it at a B.

[00:01:33]>> Yeah, I think it's Btier as well. I think there are risks especially on the consumer side, but that's not the business. It is more of an enterprise play. I do worry that there'll be certain aspects of the creative process that get taken either by competitors or the large models, but I think from an AI risk, it's probably in that B or C >> kind of the risk that I see is a lot of the end markets they serve could go to different functionality basically. And so if you used to make advertisements using Adobe, maybe now you're going directly to Meta's ad manager and it's creating you a Gen AI ad right there.

[00:02:08]And so that's how that you know journey is kind of replacing the old workflow.

[00:02:12]>> Next one is Service Now.

[00:02:14]>> I think that they're in a pretty good position to capitalize on AI. Enthropic NVIDIA themselves actually use Service Now. And so, you know, as a result of that and the fact that they've been able to accelerate a lot of their product development, they have a a decent sort of strategy with this AI control tower idea. They're going to be the guard rails, the audit, the governance on all of these uh AI agents. The real risk is that these other software incumbents, they're starting to compete against each other. And I feel like it's just up for grabs again as soon as we kind of transition to this AI agent world.

[00:02:51]Microsoft moving into AI agents.

[00:02:53]Salesforce is as well. And all of this is coming alongside them going out into more categories. Service Now used to specialize in ITSM. Salesforce moving into ITSM. uh Service Now moving into CRM which was Salesforce's old kind of stronghold in Salesforce in turn as I mentioned moving into its I think they're probably going to be in an A tier in terms of AI disruptability.

[00:03:15]Again, these aren't recommendations. I don't think there's a huge AI specific risk for them.

[00:03:19]>> Yeah, I speak with IT directors and they all seem to love Service Now and it has a lot of data of how their IT infrastructure is set up with like CMDB for example. So I feel like they have this data and this layer that they could kind of be that governance and uh for AI agents in the future. Okay. So next up is Salesforce.

[00:03:40]Salesforce kind of a similar bucket is Service Now. They've been built with a lot of acquisitions. So they have a lot of tech debt that they're now having to go back and kind of undo. That means they're kind of going to be moving a little slower. some of the advents with AI and all that does advantage them in that respect, but it just doesn't seem like as clean and unified of a platform as service now just doesn't seem as quite as tech forward in that same way.

[00:04:04]Not that there's like an imminent risk in any regard to AI, you know, fully replacing them. They're still the system of record. They're still used by the majority of all Fortune 500 companies.

[00:04:12]They'll actually use both Service Now and Salesforce. But just in terms of kind of the AI risk and all that, I just and especially the pricing risk too, that's another aspect of it. When the seats go down, you know, can you offset for that with usage or can you maybe do an enterprisewide contract and all that?

[00:04:28]I just think they're going to be a little bit behind Service Now. So, I'm going to go Btier.

[00:04:32]>> Yeah, I think that's the right spot. I think Service Now is a higher quality business here. I think Salesforce has a lot of the enduser data, but I think you'd rather have that IT layer in data.

[00:04:40]So here's one we may disagree with and that is in it.

[00:04:45]>> So in it has three main businesses. They have QuickBooks, they have Turboax, and they have Credit Karma. Credit Karma is kind of a weird business. It's basically helping people budget out. They'll check their credit scores and it's used as lead generation for a lot of these different financial apps. the the advantage Credit Karma has is it already has preloaded a lot of your finance information and so that the leads they're able to generate for a financial institution are more likely to convert and that business has actually been doing pretty well and been doing a little bit better with uh AI as well because they could potentially help figure out who's a better credit risk or not and a better lead. The last bigger two businesses is you have QuickBooks and you have Turboax. Turboax is going to be on the individual the consumer side. This is going to be consumers going to the internet saying how can I, you know, go ahead and fill out my uh taxes and they'll go to Turboax and they can use that kind of flow to go ahead and fill them out. You would think this would be ripe for AI disruption because can't you just drop your taxes into your favorite LLM and it would figure it out for you. The problem with that is LLMs are going to be probabilistic and so the any sort of math they're doing, it's going to always be a probability. And you know, Google's gotten a little bit better with this and they've tried to create some guard rails whenever you're doing a math calculation, but it's still not at a level where you're going to want to trust it for your taxes. That aspect could get better in the future.

[00:06:08]The other thing though they need is they need business logic and tax logic wrapped around the LLM. So, this means that the LLM provider has to actually decide, I want to get into tax code and I want to write rules and then I want to offer that to my consumers and then I want them to go ahead and submit their tax returns with kind of an open-ended liability. So, whenever you usually submit a tax return that's uh prepared by someone, there's going to be a signed preparer who takes liability for that.

[00:06:34]That's if you have a CPA for it. into it offers that service. They could actually have someone uh a full-ervice CPA help you out, someone sign your taxes for you, so they're going to be liable in the event something's wrong. So, it's a business I think LLM could get into if they wanted to. I'm still not fully convinced because there's a whole another layer to Turboax where you actually speak to a professional. And they're going even more physical now by opening up um hundreds of different stores, checkpoints where you could drop off your physical returns, you could talk to a human, and then they're using a on the back end to actually do your tax returns. And I still think we're at a place where people like that human element. And then QuickBooks, I would say, is less at risk. That's pretty entrenched in a lot of small businesses workflows. And also, you're talking about a kind of customer that doesn't really want to change that much. They're I use QuickBooks, for instance. Like, it's good enough for me to just get my bookkeeping done. I don't have to spend much time figuring it out. So, the net of all that is I I would probably put it as a B as well.

[00:07:31]>> Okay. So, we'll put it in B tier, but I would have it in C tier. And the reason I would have it in C tier is I think Turbo Tax is at larger risk because I think your average person's filings are not that difficult. And I understand there is this trust element, but I think over time we are going to trust specific AI tools to do these types of things for you. I don't think they actually have to write the rules if you're an anthropic or an open AI. I think OpenAI could just have an agent that goes through all of into its web pages and does filings automatically and creates this synthetic data and they can kind of reverse engineer this type of tool. Next up is Constellation Software. So, Constellation Software is a conglomerate of vertical market software companies.

[00:08:20]And it's interesting because I think in some respect to actually build the software for a lot of those businesses, it's probably the easiest of any of these businesses that we're talking about, but since they serve such small markets and it's so hard to actually find these people and sell them and then it's mission critical and it's a small portion of the overall all cost. I have a video out on constellation software where I say a lot more. Uh I think in short I'll probably put it in a tier.

[00:08:47]>> Yeah, I think it's also a tier because I think you're getting a lot of diversity.

[00:08:51]I think you're getting these kind of mini monopolies and I think people are just less likely to want to compete in that space because the addressable market's not large and in these smaller markets distribution can sometimes be harder as well. Yeah, I think that's all right. And I think that once you actually have kind of the customer on boarded, the real question becomes what value does the customer get by switching? Y >> and in a lot of markets, you're going to just say price for an answer, right? And if you're talking about Turboax in a consumer market, that's a pretty good answer. Some of them will switch to save 60 bucks and they'll take the risk and that makes sense to me. If you're talking about something that is missionritical for your business and it's a relatively small portion of your overall cost, one to two percent of your total expenses and if this thing messes you up, you're out of business and you cannot have revenue collected for a week. I don't see that.

[00:09:43]>> All right, next one. With only Mag Seven on this list, Microsoft, >> it's a tricky one to be honest. is a little trickier I think than people realize too because they have a a lot of businesses and a lot of different seats basically that they sell and as more of this work goes to AI agents they're going to undergo a pretty big transition in of them themselves going from seats to AI agents there's a lot of things changing in that business model in short though I guess it's probably going to be an A.

[00:10:15]>> Isn't it an S tier? it's not clear that they actually are a better business at the other end of this AI transition because if you're using AI agents to do a lot of your work and you think about what we use Microsoft for we're using it as like a word processor Excel and all of that right now it's like the AI agents are kind of just in the word processor and Excel and all that. I'm not convinced though that workflow doesn't shift to actually being an anthropic chatbot eventually or OpenAI's chatbot where OpenAI can say, "Hey, I have, you know, a word processor. Just go ahead and edit right in there." It's not that hard to make something that looks very similar to Microsoft Word.

[00:10:51]It's just a matter of changing that habit. And once you have that user already there and at least in my workflow when I'm researching, a lot of times I'm in something like Claude, I'm reading something, I'm going back to Word, I'm pasting it, I'm going back and forth. And so they have a very good opportunity to just get me to stay in Anthropics window or Google's whoever it is especially if you're thinking at Google with Google Sheets and all that >> and Microsoft doesn't have their own model and I think that's what they're missing from a lot of the other big players right now. Okay. So C Limited.

[00:11:19]>> Yeah. So C Limited has three businesses in short. They're an e-commerce company.

[00:11:24]They have a financials uh businesses lending money payments and then they also have a gaming business. Freef Fire is the name of their very very popular uh mobile app which is like a first-person shooter and then they also have a a gaming publishing business.

[00:11:38]There are primarily operating in Southeast Asia, seven countries there and then also Brazil. Uh in terms of e-commerce though, they've already gotten more into the logistics, the fulfillment layer. In terms of like lending money, they have actual money they're lending out and all that.

[00:11:52]They're in terms of gaming, you know, it's their own IP. So, I'm going to put this in S tier. I don't see AI as actually really a threat here.

[00:11:59]interesting.

[00:11:59]>> Um, if anything, it's an enabler in, you know, helping them price credit risk better. Also can help, you know, better surface search ads and stuff like that.

[00:12:08]The only AI risk people would bring up on e-commerce is AI agents. But as long as you own, you know, the logistics network and you have all the sellers onboarded and you have all the consumer payment info, it's going to be the AI agent using Shopee, which is the name of their platform. Now, there's a little bit of a risk. Now, if you're starting your commerce journey on an AI agent instead of shopppee, do you get more price comparison? Yeah, that's possible.

[00:12:31]And so then it's a little competitive between these other apps.

[00:12:34]>> Okay. So, next next up is Marcato Libre.

[00:12:37]>> Pretty similar deal. E-commerce company, fintech arm, one of the biggest um fintech providers in uh Latin AM, not only lending money but also doing payments. E-Wallets, their e-commerce arm, they have their own uh fulfillment logistics kind of layered on top of that. uh also their own uh membership program as well. And so I think it's similar to C. I think it's going to be S tier. I don't see AI as being a direct threat.

[00:13:01]>> Do you see uh Marcato Libre and C having the same tailwinds and headwinds? And why would that be?

[00:13:09]>> C has the the gaming operation which has a little bit more of a benefit for AI because it could help generate content quicker. It could also put players or computers in the game that are you're not able to tell if it's a human or not.

[00:13:22]So that could help kind of increase the play and then you could generate worlds or maps kind of live and so that AI aspect helps the gaming within their other existing operations. Yeah, I I see it as being pretty similar.

[00:13:34]>> Okay. So before we go to AMD, everything here is SA and B tier. What would be D tier? Like what SAS is D tiered to you or do you think the entire thing's overblown? I haven't yet actually researched a company that I found no reason to exist. That doesn't mean it's not out there. It's just I've only looked at, you know, 10 SAS names in depth. So, what would a D tier look like? It would be someone that's a point solution that does a single function and is probably going to get just become someone else's feature. It's going to become a service now or a Salesforce now's future, something like that. Last is not SAS, but you did a deep dive and I thought it was worth talking about.

[00:14:18]Why is AMD a clear D tier here? Because it's obviously a a horrible business and AI is going to kill it.

[00:14:25]>> Yeah. So AMD was for a very long time belaggered to Intel in the CPU arena and then in part because uh Foundry and then design got split off and TSMC became the best in Foundry and then AMD just focused on design. They're able to basically kind of take uh leadership in CPU in data center. If you're looking at Epic, which is EPYC, the name of their uh CPU, it's able to do um more virtual machines in containers uh per server.

[00:14:57]And so that's another way of kind of saying efficiency. And that's another way of saying your total cost of ownership is going to be less. And so for all those reasons, their chip has become very popular. You have Meta and Open AAI striking deals just with AMD uh for their chips to be in their data centers. And in some cases, they're also customizing certain aspects of their data center as well. And so they kind of were there in the right place at the right time by having a better CPU than Intel. Intel right now, of course, they're also having their moment because a lot of Silicon Valley and the US government is trying to prop them up and and make sure they continue to to be a leader um in you know, CPUs as well. But that's kind of a lot of effort to get them to that point. Whereas uh AMD's uh been working hard at it for a long period of time. they've had just basically the same tailwind as a lot of you know the AI money there's just a huge need for compute and anything that can touch that and provide that and can do so you know that more efficient than someone else yeah the demand's going to be there but we'll you know we'll have to see ultimately whether or not this is is going to be cyclical which is uh something some people don't believe.

[00:16:03]>> Do you think it's going to be cyclical?

[00:16:05]>> I think it's very hard to um outpace uh the bends of history. There's been times before where people have thought that cyclical things wouldn't be cyclical and then it turned out that that was not true. Uh and as I said uh you know before I I think almost by design though all of these companies are spending so much money on AI that they don't undersshoot demand and they're so concerned about undershooting demand that there's no chance if you play this game long enough they don't eventually overshoot it. That almost seems by design what they're trying to do. And so what happens when you overshoot demand by uh having a lot of money flow into supply? This is like you know the capital returns book kind of 101.

[00:16:43]There's a lot of money being drawn to industries with high returns. That money is basically pushing out supply. Uh that supply eventually gets overshot. When supply overshoots demand that's when you get pricing pressure and margin pressure. And so that's not to say it it happens anytime soon. And that's not to say you know their new trough can't be a lot higher than it was before. But I think it is a little silly to believe that there's going to be no cyclicality in these businesses ever again.

[00:17:08]>> Okay. So AMD is S tier >> for an AI risk. Yeah, it's S tier.

[00:17:13]Everything in AI is a beneficiary to them including um the fact that their language if you think of the CUDA framework that people use for Nvidia.

[00:17:22]They have their own version RockM and people don't talk about this much but I kind of wonder out loud whether or not AI will help people convert and translate CUDA better to ROM which could potentially reduce the grip that Nvidia GPUs have because in in AMD has their own GPU called Instinct. Awesome, Drew.

[00:17:41]Thanks for your time