Anthropic's Founders Sell Last, Not First

Everybody searching for this expects the same story. Founders ring the bell, founders cash out, retail holds the bag.

Read pages 307 through 323 of the filing and the opposite happens.

Anthropic split its insiders into four groups with four different release schedules, and the seven co-founders sit in the slowest one. Ordinary employees beat them to the exit by roughly six months.

1. What a Lock-Up Does

When a company sells stock to the public, the people who already own shares agree not to sell for a while. That agreement is the lock-up.

The reason is supply. A newly listed stock has a small public float, so if every existing holder sold into the first week, the price would collapse under the weight of it. The banks running the sale require the agreement to protect the offering, and they are the ones who hold the right to cancel it.

So "when does the lock-up expire" really asks "when does new supply arrive." For Anthropic that question has four answers, not one.

2. The Four Groups

Page 308 lists them. Paraphrased, with the filing's own two anchor points kept intact, where each period ends at whichever comes earlier:

Table 1. Lock-up cohorts and terminal release dates, Anthropic, PBC. Day counts run from the date of the prospectus. Source: Anthropic, PBC Form S-1, 26 Oct. 2026, pp. 307-308, 318, 322.

Note which line the company drew. Owning more than half a percent of the company puts a stockholder in the slow lane with the founders, on a 365-day clock. Everyone below that line sits with the employees on the 180-day clock. The threshold appears three times in the filing, on pages 117, 318 and 322.

One warning if you go reading this yourself. The summary table on pages 307 and 308 shows these cohorts as "approximately ___%", blank, because that table reports how much of the company each cohort holds, which the preliminary prospectus has not filled in. The cohorts themselves are defined precisely. The Underwriting section is where the real terms live.

3. Now the Early Exits, Which Are the Real Story

Those terminal dates are ceilings. Each group also carries earlier partial releases, and this is where the ordering inverts.

Employees, officers and non-founder directors go first. Page 321: on or after the start of trading on the first trading day following the prospectus, up to 10% of their holdings may be transferred. After the company publishes its full-year 2026 results, up to a further 35%.

The smaller stockholders go second. Page 321, for holders of 0.5% or less: 6% on day 27, another 6% on day 48, then 7% on day 62 and 7% on day 76. Four tranches inside the first eleven weeks, then 12.5% after the full-year 2026 results and 9.5% on each of five further dates after that.

The larger stockholders go third, on the same calendar but a thinner slice. Page 322, for holders of more than 0.5%: the same days 27, 48, 62 and 76, but 1.8%, 1.8%, 2.1% and 2.1%. Roughly a third of what the smaller holders get, on identical dates. Size your position above half a percent and the company slows you down.

The founders go last. Page 323 gives them one early carve-out: up to 29% of their holdings, beginning at the earlier of two trading days after Q1 2027 earnings or day 180. Nothing before that. Because 29% is a cap rather than an allocation, at least 71% of their holdings stays locked until Q3 2027 earnings or day 365.

So on the first day Anthropic trades, an engineer with vested shares can sell and Dario Amodei cannot.

Figure 1. Insider release sequence by cohort, Anthropic, PBC, ordered by earliest date available for sale. Percentages are of each cohort's own locked holdings, not of total shares outstanding, and are stated in the Underwriting section; the filing states no share counts anywhere. Source: Anthropic, PBC Form S-1, 26 Oct. 2026, pp. 307-308, 321, 323.

4. You Get the Percentages. You Do Not Get the Shares.

The disclosure here runs unusually deep. The Underwriting section hands you a release percentage for every cohort on every date, walking from the first trading day through day 365, keyed to four separate earnings releases, with steps at 13, 27, 34, 41, 48, 55, 62, 69, 76 and 83 days after one earnings date or another.

What it never gives you is a share count. The parallel summary table on pages 307 and 308 keeps that column blank on every row:

Up to _____ shares of Class A common stock, representing _____% of the shares subject to the extended 365-day lock-up period.

The gap matters because lock-up analysis is arithmetic. Shares released, divided by average daily volume, tells you how many days of selling pressure a tranche represents. The percentages give you the shape of the wave. Only the share counts give you the size, and those arrive with the pricing amendment.

So treat any absolute number you see this week with suspicion. "Six percent of the sub-0.5% cohort's locked holdings on day 27" comes straight from page 321. "Eleven million shares on day 27" does not come from this document at all.

5. The Clause That Can Void All of It

One sentence on page 323 outranks every date above:

The representatives, in their joint discretion, may release the lock-up securities described above in whole or in part at any time.

The representatives are Goldman Sachs & Co. LLC and Morgan Stanley & Co. LLC (p. 316). They wrote the schedule and they can suspend it, for any holder, in any amount, whenever they choose.

The same passage adds a second mechanism worth knowing: in certain circumstances, releasing one holder early triggers a pro rata release for certain other holders. So an early waiver does not necessarily stay contained to the party who asked for it.

Treat the published calendar as a floor on how long insiders are restricted, not a ceiling. It describes what happens if nobody intervenes.

6. Why the Inversion Is the Interesting Part

A founder lock-up twice as long as the employee lock-up is a governance signal, and it points the same direction as the rest of this filing.

Anthropic is not structured for its founders to exit. It is structured for them to keep control. A single share of Class F common stock, held by a limited liability company organized by the co-founders, carries votes equal to 50.1% of total voting power, and the filing states plainly that future issuances of Class A or Class N stock will not dilute it, because the 50.1% is fixed and tops itself up automatically as new shares are issued (pp. 1, 117). A separate class of stock, held by a long-term benefit trust, elects a majority of the board (pp. 1, 10).

Read the lock-up next to that and it stops looking like a concession. People planning to sell negotiate for liquidity. People planning to stay negotiate for control, and accept the longest lock-up in the document because selling was never the point.

Whether that is reassuring or alarming depends on what you think of the seven people holding the votes. It is not ambiguous, though, and it is not what the headline version of this story says.

7. The Seagull's Verdict

The order is backwards from the cliche, and the sizes are missing.

If you are trading this listing, the dates that matter first are not day 180 or day 365. They are the first trading day, when up to a tenth of employee holdings becomes sellable, and days 27 through 76, when both stockholder cohorts release four tranches each. The founder releases sit a long way out, and by the time they arrive the stock will have had two earnings reports to establish what it is worth.

And keep page 323 in view. Two banks can rewrite the whole calendar at will, and one early release can pull others along with it.

For the accounting story behind the headline loss, read The $80 Billion Loss That Never Happened. For how lock-up expirations behave across listings generally, read The Retail Allocation Trap. For every other 2026 offering with its own lock-up dates, see the 2026 IPO Radar.

Terms in this piece appear in the Dictionary and the Glossary.