How AI startups sell the same shares at two different prices

The unicorn status in the world of startups has long ceased to be a rarity, but in the artificial intelligence segment, it has become something like a mandatory entry ticket. Without a valuation of a billion dollars, it's difficult to attract press attention, lure top engineers, and convince corporate clients of serious intentions. Unsurprisingly, some founders of AI companies have found a way to attain the coveted status without waiting for their business to grow to it organically. As TechCrunch reports, more and more AI startups are resorting to a non-standard valuation mechanism by which the same shares are effectively sold at two different prices.

The mechanics of this scheme are not new to the venture world, but it has achieved mass distribution precisely in the AI industry. The essence is as follows: when conducting a funding round, a company structures the deal so that some investors receive shares with built-in protection—minimum return guarantees, liquidation preferences, or the right to convert on special terms. These privileges significantly reduce the actual risk for the investor, and thus the actual price they pay for a stake.

However, in the press release and on paper, the nominal share price appears—without accounting for the cost of all these "safety nets." It is on the basis of this inflated nominal price that the company's valuation is calculated, which ends up in the headlines.

Imagine an apartment being sold for ten million rubles, but the buyer is also given a guarantee of buyback at the same price after a year. The real value of the deal for the buyer is significantly lower than ten million because they are taking virtually no risk. But the cadastre will still record the figure of ten million. This is roughly how this scheme works in the venture world—only instead of an apartment, it's stakes in companies, and instead of a cadastre, it's databases like PitchBook and Crunchbase, which the entire market looks to.

The practice raises serious questions on several levels. First, it blurs the very concept of startup valuation. When two investors in the same round are effectively paying different prices for the same stake, which of these prices reflects the real value of the company? Second, it creates information asymmetry: employees receiving options rely on the public valuation and may overestimate the value of their compensation. Potential clients and partners make decisions based on figures that don't reflect economic reality. Third, if the practice becomes widespread, it will undermine confidence in the entire AI startup ecosystem—and this is a market that already raises concerns about overheating.

Context here is critically important. Over the last two years, the artificial intelligence sector has experienced unprecedented capital inflow. Dozens of companies announced billion-dollar valuations, often at early stages with minimal revenue. Investors, seized by fear of missing the "next OpenAI," were willing to agree to terms that in any other industry would seem absurd. It was this overheated market that created the perfect environment for the spread of creative valuation schemes. When everyone wants to be a unicorn and investors are willing to turn a blind eye to details, the boundary between ambitious positioning and manipulation becomes blurred.

It is important to note that formally, nothing illegal about these schemes. Venture deals are private transactions between sophisticated investors, and each party is free to structure them as they see fit. The problem is rather ethical and systemic. When a significant portion of "unicorns" in the AI sector have achieved their status through financial engineering rather than genuine business growth, it creates a bubble of expectations that will sooner or later collide with reality. This collision can be particularly painful for startup employees whose options will turn out to be significantly less valuable than they had calculated.

The venture industry is cyclical, and each market overheating generates its own characteristic abuses. In the dotcom era, it was IPOs of companies without revenue; in the crypto era, it was tokens with phantom utility. Now the AI sector has gotten its own mechanism for distorting reality. The question isn't whether this bubble of inflated valuations will burst, but how painful the correction will be—and how many truly promising companies will suffer reputation damage from those who preferred financial engineering to genuine business building.