Roman Dubczak, Deputy Chair, CIBC Capital Markets sits down with colleague Daniel Lee, Managing Director, Technology and Innovation, Global Investment Banking, CIBC Capital Markets to explore how generative AI is reshaping the software landscape. The discussion covers the divergence in performance between AI beneficiaries and traditional SaaS companies, the rise of AI-native business models, and the economic implications of autonomous “agentic” AI. Daniel shares insights on how AI is driving new pricing strategies and the complexities investors face in valuing AI-driven companies.
Roman Dubczak
Deputy Chair, CIBC Capital Markets
Hello, everyone. I'm Roman Dubczak, Deputy Chair of CIBC Capital Markets. Welcome to another edition of CIBC Perspectives. Today's discussion will centre on something that we all know quite a bit about, Generative AI, and we were all talking about it. However, it's actually heavily influenced capital flows in both the public and private markets. Over the last few years, we've seen many, many shifts in technology, but probably nothing as pronounced as what's going on in the world of artificial intelligence and its impact on society. With me here today is my partner, Daniel Lee, and Managing Director in our technology practice here in Toronto. And we're going to talk about the impact that AI has had on our sector. And, Daniel, you know, one might question about, you know, doing a webcast today on AI and like, you know, what could possibly be new? But I think you've got some insights on what the impact this has had on the financial sector and capital markets, in particular in North America. So just off the top, what are your views on how it's had influence?
Daniel Lee
Managing Director, Technology and Innovation, Global Investment Banking, CIBC Capital Markets
Yeah, absolutely. So I mean, a couple of years ago there was a lot of hype, right? There's a lot of chatter around it, but we hadn't seen much in the way of tangible results. And we're starting to see that now. And we're starting to see how it's actually playing out in both the public markets and the private markets. And happy to get into where we're seeing some of that play out. So, we all know it's a big paradigm shift, big fundamental shift. And we all know about the Mag (Magnificent) 7, right? The Mag 7’s all done very well. And they were companies that were seen to be the biggest AI beneficiaries. But we're past the Mag 7 now. And what we're seeing is that software companies, and we've just been talking in software for now, because I really want to focus that are seen to be AI beneficiaries are actually the top performing stocks by a long shot.
Roman Dubczak
Deputy Chair, CIBC Capital Markets
Yeah.
Daniel Lee
Managing Director, Technology and Innovation, Global Investment Banking, CIBC Capital Markets
I'll just give you a couple of quick examples here. Cloudflare is up about 82% year-to-date and is trading at a 33X forward revenue multiple. Palantir is up 80%. It's trading at about 100X. I had to double check that on my own. But yeah, but a 100X forward multiple. You know, Snowflake's up 45%, CrowdStrike up 50%, Rubrik, it's a company that we participated in the IPO on, that's up close to 40% year-to-date. And so, you're seeing some real capital flows into companies that are considered AI beneficiaries. In contrast, if you take a look at the broader software sector, the broader SaaS (Software as a Service) sector, it's down about 15% year-to-date. And in terms of multiples are trading below pre-COVID long-term averages. Pre-COVID, software companies tended to trade in around 7.8 times, 7-8 times forward revenues. Today, they're about just over five.
Roman Dubczak
Deputy Chair, CIBC Capital Markets
Yeah, let me just break that down a little bit. Is that because they're not as exciting as AI, or they're actually facing headwinds in their business model, because of AI?
Daniel Lee
Managing Director, Technology and Innovation, Global Investment Banking, CIBC Capital Markets
They're facing some pretty big headwinds in their business models because of AI. I'll start with AI beneficiaries first.
Roman Dubczak
Deputy Chair, CIBC Capital Markets
Yeah sure.
Daniel Lee
Managing Director, Technology and Innovation, Global Investment Banking, CIBC Capital Markets
And what that actually means for those who are not considered AI beneficiaries. And so, with respect to AI beneficiaries, there's really three levels to it. The first is, you know, to what extent is the company leveraging AI to transform itself into an AI first operating business? A great example of this is actually Shopify. I think they've got this new rule where, before you get approval to bring on a new hire, you need to prove that you can't leverage AI to have those same tasks done. And so, companies have started reporting some vanity metrics, like how much code is being created by AI. But ultimately, I think, what you're gonna start seeing and you already are, you're seeing it show up in revenue per FTE. So these are, real, tangible metrics. And AppLovin is actually a really good example of this. They went from about $3.6 million in revenue per employee to just over $7.5 million in 3 years. At the same time, they cut their employee count down from 1000 to just under 800, while revenues increased from $3 billion, to $5 billion. And so, you're seeing some real operating leverage in that business because of the way they transformed how they operate. The second level to it is, is AI a tailwind or a headwind to the category of software that you are in yourself?
Roman Dubczak
Deputy Chair, CIBC Capital Markets
Okay.
Daniel Lee
Managing Director, Technology and Innovation, Global Investment Banking, CIBC Capital Markets
So, a really good example of some of these great tailwinds, some of the top performing companies, in fact, they're all cybersecurity. Why? Because you've got all these coding tools out there.
Roman Dubczak
Deputy Chair, CIBC Capital Markets
Right.
Daniel Lee
Managing Director, Technology and Innovation, Global Investment Banking, CIBC Capital Markets
That could create an increase in the level of threats for hacking and all that other good stuff. Conversely, the category of robotic process automation has been hit pretty hard. These are companies who create software that drive rules based automations that are now just being replaced by AI, who can run these automations dynamically. The third level to it is, companies who have been able to leverage existing data sets or capabilities to create entirely new, moonshot AI native product categories. A good example of this is actually, Palantir, who actually started off as more of a services business.
Roman Dubczak
Deputy Chair, CIBC Capital Markets
True enough, yeah.
Daniel Lee
Managing Director, Technology and Innovation, Global Investment Banking, CIBC Capital Markets
Yeah.
Roman Dubczak
Deputy Chair, CIBC Capital Markets
Yeah.
Daniel Lee
Managing Director, Technology and Innovation, Global Investment Banking, CIBC Capital Markets
But they've actually morphed into an AI platform. They've created this, new capability, which they actually call ‘The Artifical Intelligence Platform’, that provides customers in government and defence with a secure platform on which to build and deploy agents using their own internal data. It turns out there's huge demand for this. And AIP has been a huge driver of, of their stock price as well as their rerating in terms of valuation.
Roman Dubczak
Deputy Chair, CIBC Capital Markets
Yeah, and many other companies as well, with massive databases like you said, it created a whole new category. But just to go back to SaaS, SaaS companies, it's not like they're not doing any business. But what sort of headwinds are they facing in, you know, what would you say the prognosis is for, a lot of these are, let's just say public stocks for the time being?
Daniel Lee
Managing Director, Technology and Innovation, Global Investment Banking, CIBC Capital Markets
Yeah. So, the number one place that it shows up is actually growth rates.
Roman Dubczak
Deputy Chair, CIBC Capital Markets
Yeah.
Daniel Lee
Managing Director, Technology and Innovation, Global Investment Banking, CIBC Capital Markets
So, I think as recently as this quarter, the number of SaaS companies, take out sort of the AI beneficiaries, who, are achieving over 30% year-over-year growth might be three or four. It's not a big number.
Roman Dubczak
Deputy Chair, CIBC Capital Markets
Yeah.
Daniel Lee
Managing Director, Technology and Innovation, Global Investment Banking, CIBC Capital Markets
Whereas back in 2020, 2021, that would have been a much bigger bucket of companies. And the reason, one of the reasons we're seeing slowing growth rates is because customers are often seeing what new AI tools might be out before committing to multi-year contracts for some of these companies. On top of that, some customers are actually looking to leverage, coding tools internally to be able to build their own homegrown applications instead of buying and I know at CIBC we're exploring some of that ourselves. And so, that is slowing down sales cycles. The other reason is, and it's related to that, it's just an increase in competition. You're seeing a huge rise in competition across multiple categories in software because of the erosion of technical notes, AI makes it way easier to code and ship product. The other factor is just the go to market motion. So, one of the things that SaaS companies had done historically really well, was created top of funnel through content strategy. So, blog posts, you know, writing white papers, getting themselves higher ranked on search engine optimization, so on Google, and so when you've got good top of the funnel, you know, you can start moving prospects down and eventually close. Generative AI has changed all of that. A huge proportion of searches now don't actually go into Google. They go to tools like ChatGPT or Perplexity. Yeah. And whereas with Google, you know, for every 15 to 20 pages it scrapes, it'll send you one visitor with a, ChatGPT and Anthropic, even, like these numbers go up to 20,000. So, for every 20,000 pages they scrape, they'll send you one visitor. And so, the net impact of that is that, it constrains your ability to create top of funnel, the metric that you can use to monitor all of this is the increase in, CAC payback. So, software companies measure themselves, their go to market efficiency through this metric called CAC payback, which is how long it takes to pay back your customer acquisition costs. Historically, this was about 18 to 24 months. That was considered, okay.
Roman Dubczak
Deputy Chair, CIBC Capital Markets
Yeah.
Daniel Lee
Managing Director, Technology and Innovation, Global Investment Banking, CIBC Capital Markets
Not bad. If you look at Q1, we're up approaching sort of 4 to 5 years, and that's the median for for SaaS companies that are publicly traded.
Roman Dubczak
Deputy Chair, CIBC Capital Markets
Yeah, yeah. I think you touched on your like talking around this topic of, Agentic AI, I maybe touch on that a little bit as to, you know, what that actually is and how that specifically is having an impact on the use of AI tools.
Daniel Lee
Managing Director, Technology and Innovation, Global Investment Banking, CIBC Capital Markets
Yeah, absolutely. So, there has been a lot of buzz, you know, Agentic AI. And really, I think how I would define it is that it's an autonomous program... software, powered by AI, where by you can give the agent an objective, it can create subtasks for themselves, but then use tools to actually complete the tasks and then reprioritize those tasks and loop them until the objective is reached. And so, I think where we're heading to is we're moving from a world where software just helps us get work done to software actually just doing the work for us.
Roman Dubczak
Deputy Chair, CIBC Capital Markets
Right.
Daniel Lee
Managing Director, Technology and Innovation, Global Investment Banking, CIBC Capital Markets
The term de jour, I think for this transition is going from ‘Software as a Service’ to, ‘Services as Software’. The economic implication of all of this is that now you're moving from tapping software budgets, which globally is somewhere around, depending on what source you look at, $400 to $500 billion a year. Yeah. - To being able to go after, the labour market TAM, which depending on which side you look at, is somewhere between 4 to 5 trillion dollars.
Roman Dubczak
Deputy Chair, CIBC Capital Markets
Right. Let alone all the other operational budgets that exist in the world. So, traditional just to get on valuation, SaaS models were basically a predictable, seed based, recurring revenue model. Traditional valuation. Financial institutions can lend against that. It's a predictable cash flow, the emergence of Agentic AI and their ability to deliver work transforms that model. What model does it actually transform it into? And that was how is that impacting financing, for example.
Daniel Lee
Managing Director, Technology and Innovation, Global Investment Banking, CIBC Capital Markets
Yeah. Yeah. I mean it's interesting, right? Because, you're not selling to seats, you actually replacing seats you can't sell on a seat-based model anymore. And so, what we're shifting towards is we're going from a world where you're paying for access to software to paying for work that's delivered. And so, what that means is moving away from the seat-based subscriptions to charging for, a successful resolution or per task completed or, you know, per conversation. And so, some of the new pricing models, that arise that, a lot of AI native and even the incumbents are experimenting with are things like price for agent action. So, it's more like a consumption model of tokens. It's how, a lot of these large language models get priced. There's, a price per agent workflow. So, you know, name your number of workflows and we'll, we'll sell you a bundle of workflows and price against that. There's outcome-based pricing, right? So, you know, if on average it costs you from a customer service perspective, $20 to resolve a customer issue, maybe now you pay, you, your AI software vendor ten bucks per resolution and then ultimately there's this price per agent. So, this is like an FTE replacement model, and there's trade-offs with, with these models, with these pricing models. So, on the one hand, the variable pricing model and being able to price, for things like outcomes gets you to be able to actually capture more of the ROI delivered to the customer. So, you know, in economics, I think it gives you the ability to price, discriminate and reduce the amount of consumer surplus. On the flip side, you end up with volatility in terms of, how well you're gonna be able to predict your revenue from month-to-month or, quarter-to-quarter. The other issue in and around this is actually COGS volatility. Because unlike...
Roman Dubczak
Deputy Chair, CIBC Capital Markets
Cost of goods sold.
Daniel Lee
Managing Director, Technology and Innovation, Global Investment Banking, CIBC Capital Markets
Yeah.
Roman Dubczak
Deputy Chair, CIBC Capital Markets
Yeah.
Daniel Lee
Managing Director, Technology and Innovation, Global Investment Banking, CIBC Capital Markets
Because unlike traditional software, you actually have to pay for these tokens, that you're purchasing from the foundational models like an OpenAI or an Anthropic. And so, if you're not optimizing those two things very well, you could run into a situation of, selling negative gross margin services.
Roman Dubczak
Deputy Chair, CIBC Capital Markets
Yeah. Yeah. Interesting. So, by extension, how would an investor value these different business models? Because as I mentioned earlier, like you had, SaaS was kind of getting easy in terms of valuation, per cash flow, so to speak. This, as you just outlined, could take a variety of different forms, like how do you look at it valuation wise?
Daniel Lee
Managing Director, Technology and Innovation, Global Investment Banking, CIBC Capital Markets
Yeah. It becomes a lot more complex.
Yeah.
Daniel Lee
Managing Director, Technology and Innovation, Global Investment Banking, CIBC Capital Markets
Because now you're in a world where, trying to predict revenue and profitability is going to require a much deeper understanding of actual product adoption, usage ramps of customers, seasonality, how product improvements could alter all of these things. And this, of course, to your point, strains traditional SaaS metrics, which rely on predictable revenue streams and, and subscription revenue. And so, the upshot here for valuation is if you think about what the intrinsic value of a business is, it's a DCF (Discounted Cash Flow) ultimately.
Roman Dubczak
Deputy Chair, CIBC Capital Markets
Yeah.
Daniel Lee
Managing Director, Technology and Innovation, Global Investment Banking, CIBC Capital Markets
And the biggest component of value in a DCF is a terminal value. Terminal value is a few years out. We're still experimenting with all of these pricing models, and we don't know where COGS going to land.
Roman Dubczak
Deputy Chair, CIBC Capital Markets
Right.
Daniel Lee
Managing Director, Technology and Innovation, Global Investment Banking, CIBC Capital Markets
How do you place a value on the terminal value. So, it creates a lot of uncertainty, in terms of valuation. But then, I'll add another one that's actually fairly unique to where we are today in the market. And that's this idea of a zero-value threshold that some investors are starting to look at. And, this is a situation where, an AI application company in particular, could face a critical point where the enterprise value of the business effectively drops to zero if they end up being displaced, and their core technical advantage ends up being matched by either some sort of open source alternative, or, you know, the foundational models themselves come and create these applications. We've seen this happen, with OpenAI extending themselves into the application layer, from just a model layer, right? They've gotten into image generation, right? They looked at getting into code generation... By offering this up, by offering these applications up to their broad base of users, they effectively render a lot of these application-level start-ups, obsolete. Unless there's a massive technical advantage or performance advantage. And so, because traditional valuation methods don't account for this displacement risk, what investors are starting to look at is, “Is there a way to quantify what this displacement risk looks like?” So, you know, you could have great ARR (Annual Recurring Revenue) growth and a valuation multiple that's associated with what that looks like. But then ultimately you're going to have to adjust that with some displacement risk factor, that I know investors are working towards sort of quantifying.
Roman Dubczak
Deputy Chair, CIBC Capital Markets
And I guess the fall back to all that is, for lack of a better guess, multiple of revenue, right? That kind of seems to be where we're, where we're landing on a lot of these, yeah. Okay Daniel, sort of the last question, I guess, in theory, this, podcast, if we... this webcast, rather if we, if it happens in 12 months from now, it may not be you and I, it may be AI generated, you never know how these things go. I’d like to believe we, we survive as a species, doing what we do in the financial sector... But, I’d just like to get your view. Like, if you look at, you know, the pro cases as it relates to AI, productivity improvements, like, huge productivity improvements, society benefits, the bear case is as the aforementioned job losses and, you know, it's a deeper topic. But, you know, AI could get dangerous in certain aspects of its, utility... You know, where do you sit on this? Like, what's your what's your view today? You know, things may change, but where are you today on that?
Daniel Lee
Managing Director, Technology and Innovation, Global Investment Banking, CIBC Capital Markets
Yeah. I'm cautiously optimistic, you know, on the pro side, I think the history of economic growth would, suggest that breakthrough technologies, whether it was the steam engine or electricity that's always unlock new frontiers of productivity and led to fairly sizable improvements in, standards of living. And so, I think if AI can automate knowledge work and accelerate discovery of new ideas, we might seen GDP growth rates rising to levels we've not seen before. And so, in that scenario, you know, I think you end up with potentially runaway innovation that can continue to compound on itself, which will then make goods and services vastly more abundant and affordable than ever before. At the same time, we know that these transitions are almost never smooth. And so, the downside risk isn't just a major dislocation in labour markets, but also, wealth inequality and social instability. And so, I think, you know, it will be incumbent upon us as a society to find new ways of distributing the value that's going to be created by AI.
Roman Dubczak
Deputy Chair, CIBC Capital Markets
Great. Well, that's a very thoughtful close. Thanks, Daniel. And it's that thoughtful close that leads me, as an optimist, to think that you and I will be here as financiers, to talk about this in the future as to the impact that AI has had. But it's certainly a very exciting time. A lot of change in the last year or two, and I suspect it's going to be yet another crazy year in terms of progress, ahead of us. And I want to thank all of you for joining us today on a very exciting topic. Please feel free to reach out to any of your partners here at CIBC. We're happy to walk you through the evolution of what's happening in the technology markets, AI, and the economy in general. Thanks for joining us. Hope to see you again soon.
CIBC PERSPECTIVES
Navigating the AI Revolution in Capital Markets
Running time: 20 minutes, 17 seconds
Host
Roman Dubczak, Deputy Chair, CIBC Capital Markets
With
Daniel Lee, Managing Director, Technology and Innovation, Global Investment Banking, CIBC Capital Markets