Latest / Tech IPO Conversations with Fexingo: Public Software Companies and Capital Markets / How AI Licensing Is Reshaping Tech IPO Timing
Transcript
- Lucas: If you look at the biggest tech IPOs of the last few years, there's a pattern that's slowly shifting: companies are staying private longer, but they're also getting much more creative about how they generate revenue before they hit the public markets. Luna: And one of the biggest drivers of that right now is AI licensing. Not just selling access to a model, but structuring the whole revenue model around recurring licensing deals. Lucas: Exactly. And it's changing the calculus for when a company even needs to go public. I was looking at the Cyera story this morning — they're reportedly targeting a 12-billion-dollar valuation. That's 80 times their annual recurring revenue. Luna: Right — and they're still operating at a loss. That multiple is almost entirely driven by the perceived durability of that licensing revenue. Lucas: A few years ago, that kind of valuation would have required a public listing, or at least a serious IPO filing. Now, private investors are comfortable paying up for companies that can show sticky, high-margin licensing revenue. Luna: So what's changed? Is it just that the market has more cash sloshing around, or is there something structural about AI licensing that's different from, say, SaaS subscriptions? Lucas: I think it's structural. SaaS subscriptions are usually per-seat, usage-based, and customers can churn relatively easily. AI licensing often locks in longer terms, exclusive access to a model or dataset, and it can be tied to hardware or compute — which makes it stickier. Luna: And that stickiness is exactly what private investors love. They can model out revenue three to five years with a lot more confidence. Lucas: Which means companies can delay their IPO, sometimes indefinitely, if they can keep raising private capital at those multiples. Look at NVIDIA — they've been doing this for years, licensing their GPU architectures to other companies for specific use cases. Luna: Yeah, NVIDIA's licensing revenue isn't just from selling chips. They have deals where companies pay for access to their CUDA ecosystem, or for custom AI model training on their infrastructure. Lucas: And those licensing deals often come with exclusivity clauses — so a company like Cyera gets to say, 'We have the only ai powered data security platform that runs on this specific NVIDIA architecture.' That's a competitive moat they can take to private investors. Luna: It also changes the IPO pitch. When a company eventually does go public, they're not pitching a story — they're pitching a revenue stream that's already been validated by the private market. Lucas: Exactly. And that validation can actually lead to a higher IPO price, because the market has had years to watch that licensing revenue compound. But there's a flip side. Luna: What's that? Lucas: If the licensing model is so good that companies don't need to go public, you end up with a smaller pool of high-quality IPOs. Retail investors miss out on the early growth phase — they only get access after the biggest gains have already been captured by private funds. Luna: That's been a concern for a while, but it feels more acute now with AI companies. They can raise hundreds of millions in private rounds at valuations that would have been unthinkable five years ago. Lucas: And that's why the licensing model is so important. It gives private investors a concrete metric to point to. It's not just a vision — it's a signed contract with a Fortune 500 company paying for exclusive access to an AI model. Luna: So for the companies that do go public, the bar is higher. You have to show not just growth, but that your licensing revenue is expanding faster than your cost to acquire those deals. Lucas: Right. And we're starting to see that play out in the public markets. Look at Palantir — they went public via direct listing, and a big part of their pitch was their government and commercial licensing deals. Those recurring contracts gave the market confidence. Luna: Palantir's stock is up almost 15 percent in the last five days, by the way. That's partly because they keep signing these long-term licensing agreements. Lucas: Exactly. And that's the pattern. Companies that can show a licensing model that's sticky, high-margin, and expanding — they get rewarded. But companies that just have a SaaS subscription with high churn? They're going to struggle. Luna: One thing I'm curious about: does this licensing trend affect how we think about IPO pricing? If a company has been licensing its tech for years, the IPO is almost a formality. Lucas: It does change the dynamic. Traditionally, an IPO was the moment when the company's valuation was 'discovered' by the market. But if private investors have already paid 80 times revenue for a company like Cyera, the IPO becomes more about liquidity than price discovery. Luna: And that can create its own risks. If the public market doesn't agree with the private valuation, you get a down round or a stock that trades below the IPO price. Lucas: We've seen that with a few AI companies that went public in the last couple of years. The private market was incredibly optimistic, and the public market was more skeptical. But the licensing model helps bridge that gap, because it provides a tangible revenue base. Luna: So for a company like Cyera, if they do eventually file for an IPO, the question will be: can they show that their licensing revenue is growing faster than the market expects? Lucas: That's the key. And they'll need to show it not just in terms of total revenue, but in terms of net dollar retention — how much existing customers are spending year over year. That's the metric that licensing models really shine on. Luna: And that's also where the public market will scrutinize them. If a customer signs a three-year exclusive license, that's great for the first year. But what happens when the renewal comes up? Lucas: Right. If the competition has caught up, or if the technology has shifted, that renewal might not happen. So the licensing model is only as good as the moat around it. Luna: That's a great point. And it's something we don't talk about enough — the durability of the licensing revenue, not just the existence of it. Lucas: Yeah. And that's where the conversation gets really interesting. Because if you're an investor looking at a pre ipo AI company, you need to understand not just the licensing contracts, but the technology and the competitive landscape that underpins them. Luna: So the licensing model is reshaping the IPO timeline, but it's also raising the bar for what we need to understand as investors. Lucas: Exactly. And that's the kind of analysis that makes these conversations useful — if you're building or running a company, or investing in one. Luna: And if these conversations are useful for what you're building or running, a couple of dollars a month is genuinely what keeps them going — buy me a coffee dot com slash fexingo, if you've gotten something out of them. Lucas: Yeah, it's a small thing that makes a real difference to us. Helps keep the show ad-free and focused on the substance. Luna: So back to that durability question — how do you actually evaluate whether an AI licensing deal is going to hold up over time? Lucas: I think you look at three things. First, the exclusivity terms — how long do they last? Second, the switching costs for the customer — are they locked into the model's ecosystem? And third, the pace of innovation — can the licensor keep improving the product fast enough to justify the renewal? Luna: That third one is tough, because AI is moving so fast. A licensing deal that looks great today might feel obsolete in six months. Lucas: Right. And that's why we're seeing more companies include performance-based clauses in their licensing contracts — tying renewal pricing to model accuracy or speed improvements. Luna: That's a smart way to align incentives. It gives the customer confidence that they're not overpaying for stale technology. Lucas: And it gives the company a concrete reason to keep innovating. If your licensing revenue depends on hitting certain benchmarks, you have no choice but to invest in R&D. Luna: So in a weird way, licensing might actually force better product development than a traditional IPO would. Lucas: That's a really interesting way to put it. Because when you're public, you have quarterly earnings pressure — sometimes that pushes companies to cut R&D to hit numbers. But with a licensing model that has performance clauses, you're almost compelled to invest. Luna: So the licensing trend isn't just changing IPO timing — it's changing the underlying business strategy. That's a bigger shift than most people realize. Lucas: Yeah, I think we're only at the beginning of it. Over the next two to three years, I expect we'll see more private AI companies structure their entire revenue model around licensing, and then use that as a springboard to a public listing — or skip the listing altogether. Luna: And that's going to raise interesting questions for regulators and exchanges. If the best companies never go public, what does that mean for the stock market? Lucas: That's a conversation for another episode, I think. But it's coming sooner than most people expect.