Do Gross Margins Matter Anymore?
Yes, Duh
When it comes to venture capital investing, does it really matter what a company’s margin profile is anymore? The definitive answer to this question is YES, in ALL CAPS, of course margins matter.
However, in the old world (see: before transformer models), the beautiful thing (and somewhat scary thing) about venture capital was that the ASSUMPTION for the past 25+ years has been that venture-backable was synonymous with really strong gross margins. If you didn’t have at least 75% GM rate, you probably weren’t a scalable enough business to warrant venture capital. And that’s what matters to venture investors: scalability. If you have gross margins north of 80%, there is a good chance that your fixed cost base will remain relatively static, regardless of the size of your company. That means that as you grow revenue, your profits can scale at an even faster rate than your revenue. Thus, you can reinvest that cash aggressively in go-to-market, R&D, partnerships, or other potential growth levers. Which, in turn, means that you could grow even more rapidly. This would create quadratic growth in the best scenarios, and generate ludicrous returns.
[Of course, there are some notable exceptions to this, such as companies who used blitzscaling to pour capital into a specific problem area and grow rapidly, margins be damned. There are only a handful of examples of that actually working, however, and most of them feel like outliers (Twilio, Spotify).]
As a result of this need for scalability, the venture industry found some business models that allow for ultra-scalability, like SaaS and heavily network-effected businesses. And, as you might imagine, those models allowed for great margins. Which allowed for those companies to spend aggressively. And in turn, even if those companies needed to raise more capital, it was okay, because gross margins were so strong. When those companies needed to turn from FCF-negative to FCF-positive, they could do it effectively, because, again, they were high gross margin businesses, and that allowed for a more flexible business model.
There is a really recent case study of this FCF-elasticity happening to impressive effect. Post ZIRP-era tech stocks had to quickly turn on FCF to regain investor trust. Many of them did:
There is a reason SaaS companies use revenue multiples, and this is probably it.
On top of generating pliable gross margins, software is cool because you build it once, then can sell it a million times. That’s the definition of scalable.
Then AI rolled along and started making people ask some interesting questions. Like, what does this mean for software in general?
[For the record, I am not a software-is-over pessimist. I think the ability for tools like Cursor to help write software means that there will be more software in the world, and more software means there will be more problems to solve, which will in turn require more people and other software tools to help solve them. It’s just going to get more complicated and less complicated all at the same time.]
But one of the funky things about AI is that the margins aren’t necessarily that great. LLMs are expensive, and take up a lot of compute. And it is definitely not a build once, sell a million times scenario.
But, that cost of compute is going down. And we probably have another Moore’s Law in effect. The cost will continue to go down, depending on who you listen to, by as much as 10x per year. A big reason Moore’s Law took effect is because there was a massive economic incentive for innovation in the space. That same thing is true for the cost of LLM compute.
But even still, the nature of AI is that each output is a new output (sorta). And there is a good chance that the margin never quite reaches SaaS level margins (although it’s not impossible that it does!).
What does this mean for venture? I have been rereading the Innovator’s Dilemma (Clayton Christensen, what a guy!) and I was reminded of the following charts:
Disruptive innovations generally take on different margin profiles than sustaining innovations. The value chain that matters to AI companies might be pretty different than enterprise SaaS, or other traditional VC-backed companies.
There’s a chance that the margin profile for AI tools is just different from that of SaaS companies (or other high margin tech business models). New players can afford to break into the market using margin as their opportunity. Additionally, AI should, implicitly just make a business more scalable because it is a deflationary tool – meaning that it should reduce the labor required to build a business and keep a business operating. Humans are frequently the most expensive part of a corporate organizational structure. When it comes to operating margin, human capital is generally the biggest line item. So even if you could reduce that marginally, that’s a pretty big impact on bottom line cash flow. Admittedly, this is somewhat of a logical leap. But, while gross margin might be degraded or impaired for AI-forward companies compared to their SaaS counterparts, their operating margin might be greatly improved.
So while gross margin might degrade (MIGHT!), operating margin might improve. The risk here, of course, is that operating margin is a lot less flexible than gross margin, and so it might be trickier to turn on the cash flow spigot during tough times with this new proposed expense structure of AI companies, but that’s a bridge to cross at a later date. They also might have stronger network effects, which could drive up marginal value over time as well – as more people use the product, the product gets better, etc.
Gross margin may no longer be the sole benchmark for scalability in the age of AI. But for founders and investors alike, understanding why your margin profile looks the way it does—and how it might change—is more important than ever.