He sold his company with zero revenue on the books. That's the story of angel investor Alexander Kardos-Nyheim. He founded an AI startup himself and closed the sale before a single dollar of revenue came in. Since then, he's sat on the other side of the table, evaluating other AI startups, and has written up the valuation criteria he distilled from that exit.
How does a company sell with no revenue at all? His answer: the acquirer wasn't buying a product — it was buying the technology's defensibility. That distinction leaves Korean founders with an uncomfortable question. Am I building an app on top of AI, or am I building AI itself?
The Line Investors Draw: App or Foundation?
When Kardos-Nyheim evaluates an AI startup, the first question he asks is this: "Is your product a layer built on top of an existing AI platform, or does it work with AI itself?" The logic behind the question is straightforward. If you build a service by calling an API from a platform like OpenAI, Google, or Anthropic, you get replaced the moment that platform extends its own feature set by one notch.
The AI startups he rates highly for long-term value sit at a different layer. They're teams solving problems deep in the technical stack — model training, infrastructure design, novel inference methods. These companies don't wobble when the underlying platform changes. If anything, the platforms end up needing their technology.
This distinction has only become a major factor in investment screening over the past two or three years. As AI commoditized rapidly, the barrier to entry for "using AI" all but disappeared — anyone could call an API and spin up a service. Paradoxically, that made the services built on top of those APIs less defensible, since any competitor could build the same thing the same way. The gap between the value platforms capture from below and the value services manage to hold onto above keeps widening.
The Counterargument: This Only Applies to Big Teams
There's a hidden premise in Kardos-Nyheim's framework: enough technical talent and capital, plus a runway long enough to survive without revenue. Realistically, a Korean solo founder or a small team can't build a company that trains models or builds inference infrastructure. Read literally as "if it's not foundational AI, it's worthless," the framework leaves most small founders concluding there's nothing they can do.
The piece is really aimed at Silicon Valley-backed teams or startups fielding angel rounds. A solo founder or small-team operator can't simply follow the prescription to "build foundational technology" — trying to would likely pull them away from their actual strengths. Read as a universal mandate that every founder must build foundational AI, the article has little to offer small Korean businesses.
But it would be a shame to stop there. Buried in Kardos-Nyheim's evaluation question is a layer that applies regardless of scale: what, in my business, is irreplaceable?
Translating Investor-Speak Into Founder-Speak
Track the investment decisions of a dozen venture capitalists and one pattern shows up again and again: investors buy defensibility, not scale. Before team size or revenue figures, they ask, "Could this service exist without this specific team?" Flip it around — "if this team vanished, could another team clone it in three months?" — and if the answer is yes, that business is weakly defensible no matter its size.
Translate that question into the context of a Korean solo founder, and a few self-check items fall out.
Would my service disappear if ChatGPT or Claude started offering the same feature natively? If it's a simple summarizing, translating, or text-generation tool, a single platform update could make it obsolete. If the answer is yes, it's time to take a hard look at what the service is actually accumulating.
Do the data, relationships, and community I've built generate value independently of AI? If so, AI is just an efficiency layer stacked on top — and that kind of business doesn't shake much when the underlying platform changes. In a business customers seek out for reasons that have nothing to do with AI, AI's job is simply to cut costs or speed things up.
Is the domain I work in one AI still hasn't cracked? This covers things AI can't turn into training data — hands-on knowledge specific to an industry, context pulled from customer interviews, relationships built up over years in a trade. Anyone generating that data or maintaining those relationships directly holds a defensible position, no foundational technology required.
I'd compress those three checks into a single question: could this business exist without AI, or did AI make it possible in the first place? In the first case, AI is a tool for boosting efficiency. In the second, AI is the business's skeleton. As an investor, Kardos-Nyheim is looking for teams that fall into the second category — and that same yardstick works perfectly well for a solo founder's self-diagnosis.
Kardos-Nyheim could sell his company without revenue because he convinced the acquirer his technology was irreplaceable. Defensibility sold before revenue ever had to. Most Korean founders and planners aren't positioned to build foundational AI technology themselves. But the question "what's irreplaceable about my business" is one anyone can put to use right now, regardless of scale. At a moment when AI services are flooding the market, the ones who last a little longer are the ones who first figure out which layer their app is actually standing on.
