The AI premium on EquityZen's Top 20 is mostly a price on somebody else's mark

A premium on a secondary marketplace is not the same object as a premium on an exchange. Haslett's own account of the mechanics is that secondary investors are "making the same calculus as primary venture and growth investors," and that much of the discount for capital intensity is already baked into the headline valuations set by primary raises ^[8]. Read literally, that makes a Top 20 list a ranking of which primary marks buyers are still willing to transact near, rather than an independent verdict on value.

The discount mechanism he describes is time. Building a factory or waiting on regulatory approval pushes out the date at which a company can raise its valuation or reach an exit, and buyers price that delay into what they will pay ^[7]. That is manageable arithmetic when delay is the only thing standing between the buyer and the outcome. Capital structure is the other thing. Five or six years ago, Haslett says, a battery company or a new chip manufacturer had little practical alternative to selling equity; in 2026 credit and asset-based financing are available ^[9]. He is direct about the consequence: if one of these companies underperforms or ends in a distressed asset sale, creditors and lenders are paid first, and secondary investors have to factor that in ^[10].

That is the part a price per share cannot show. A marketplace quote tells a buyer what the last transaction cleared at; it does not tell them how much senior paper has been layered on since the round that set the reference mark. The published exchange names no company as having taken on that financing, and puts no number on either the premiums or the discounts it opens by describing ^[11].

Which is worth being plain about, because the framing came from the publisher. Crunchbase's introduction sets up the interview around why AI companies command premiums while many older startups trade at discounts ^[1]. What the transcript delivers is a list of new entrants including Figure AI, Project Prometheus, Redwood Materials and Scale AI ^[3], Haslett's read that this is a thematic move into AI infrastructure, space tech and robotics rather than a scramble for scarce supply ^[4], and a practitioner's description of how the discounting works. That is useful. It is not price discovery, and operators reading these lists as a signal should not treat it as one.

The exit that would settle the argument is half open. Q2 was one of the strongest venture-backed IPO quarters since 2021, though SpaceX drove much of it ^[5], and Haslett says conditions are better than three or six months ago now that the SpaceX hurdle has cleared, with the stock market at an all-time high ^[6]. His verdict on what happens after the listing is one word, "meh," with Cerebras having come down [c7b]. Buyers paying near primary marks on the strength of a theme are underwriting an exit whose most recent examples faded after the print.