Carta’s third-quarter report says median startup valuations now sit above their 2021 peak at most fundraising stages. Forge’s trading record for September says the median private share changed hands nine percent below its company’s last round price. Those are two prices for one share: the one a round agrees to, which every offer letter quotes, and the one a holder who can sell actually gets. They parted because buyers left as sellers arrived.

At a glance

  • The shift: The price a round sets and the price a holder clears have parted.
  • The number: nine percent below the last round, the median trade in September.
  • The risk: record marks set by a fifth fewer deals.

When the mark stopped clearing

The order of events matters. Through September, companies on Carta raised $49.3 billion across 1,179 rounds, and median valuations at most stages climbed past their 2021 highs [1]. In August, the median trade on Forge’s marketplace cleared three percent below the seller’s last round. In September it cleared nine percent below [2]. I read both reports twice before the shape of it settled, because neither one is wrong and together they say what neither says alone.

The top of the market moved harder than the middle. The 90th percentile went from a 58% premium to a 15% premium in a month, and the 75th from 11% to 6% [2]. The premium on the best-known names did not fade. It lost forty-three points.

What changed was who was bidding. Buy-side interest fell to 43.5% of new and updated bids in September, the fourth straight monthly decline and the lowest share in about two years, from 71.4% in May [2]. Sellers arrived as buyers thinned.

The record marks carry a second weakness that has nothing to do with second-hand trading: fewer deals set them. Total deal count fell 21% year over year, and since early 2023 the cash raised each quarter is up 157% while the round count is down eight percent, so the average round has more than doubled [1]. A record set by 1,179 deals is a real price. It is also a thinner reading of the market than one set by 1,500.

The exit side explains why holders sell. Just 18 companies went public in the third quarter, and 992 active unicorns carried $5.7 trillion of paper with no clear path to liquidity [3]. When the door is that narrow, the second-hand market is where a vested share becomes money. Its price is also the first market quote most holders ever see.

By the numbers

  • $49.3 billion: Third-quarter capital across 1,179 rounds, roughly a fifth fewer deals than a year earlier [1].
  • Nine points: The median trade’s discount to the last round in September, up from three in August [2].
  • $91,000: What a $100,000 block quoted at the round’s price would clear at that median discount [2].

The number to ask for

Anyone holding a grant, or weighing an offer with a large paper component, now has two prices to compare, and usually only one of them is written down. Ask for the round price per share by name: what the last financing paid, when it closed, how long ago. A grant described as worth a certain amount at that price is worth nine percent less at the September median, and far less at the bottom, where the 25th percentile still cleared 30% below its round [2].

Tenders are where this gets practical. A company-run buyback prices its offer against the most recent round, and the arithmetic of taking one now against waiting for a listing is what the tender-versus-listing model is for.

My read is that October’s print is the number to watch: whether it holds at nine percent or widens. The one conversion to watch this quarter is Anthropic’s. Its amended registration statement, filed on 10 November, sets an estimated range of $70 to $78 for 825 million Class A shares on Nasdaq [4]. Press reports put pricing on 18 November and the first trade a day later; those dates are reported, not filed. That is the rare case where a very large private mark becomes a price anyone can check, and it will say more about the top of this market than a quarterly median can.

The catch

Forge’s nine percent is a median trade on one marketplace in one month, measured against each company’s own last round. Forge warns that moves at the tails can come from the mix of companies that traded rather than from a repricing [2]. The bottom of the distribution also improved: the 10th percentile’s discount narrowed from 87% to 51% [2], which is not what a uniform sell-off looks like.

I could be wrong that any of it matters for a company that raises again at a higher price next year, because the next round resets the mark and the discount becomes a footnote. What the record supports is narrower and more useful. In the quarter the marks set records, the median holder who cleared a share took nine percent less than the mark, and the best-known names lost their premium first. A grant is a claim on that clearing price, not on the round’s headline.

Related reading

Sources