Build1 publisher2 min readPublished
Anthropic's IPO plan puts 50.1% of key votes in one share its seven cofounders control
Anthropic's IPO prospectus gives a single Class F share, voted by its seven cofounders, 50.1% of votes on key corporate matters. Teams on Claude get a vendor public shareholders cannot outvote on those matters while the founder group holds together.
The Engineer · Build desk
What happened
- The prospectus warns the structure could produce decisions against Anthropic's short, medium or long-term financial interests and could lower the value of Class A shares.
- After the listing, Class F and Class A holders will elect three directors: Dario Amodei, board chair Daniela Amodei, and one director not yet named.
- Anthropic's Long-Term Benefit Trust, whose trustees include Ben Bernanke and Richard Fontaine, will elect the other four board members.
- A founder can be dropped from the group for leaving the company, dying, selling too many shares, or being removed for cause.
Compiled by The EngineerSomething wrong?How this is made
Why it matters
- exposure Teams building on Claude depend on a vendor whose public holders, capped at 49.9% on key votes, cannot reverse the founders' direction at a shareholder meeting.
- constraint Investor pressure cannot reopen product lines Anthropic says it skipped, such as image and video generation, while the founders' ballot outweighs all other shares.
- decision Vendor-risk reviews of Anthropic now have to track founder departures and share sales, because each one moves Class F closer to losing its extra weight.
Four of the seven founders in the Founders LLC have to agree before the Class F share is cast [1][1]. In effect, four people decide a single ballot that outweighs every other share combined on the matters it covers. Class A common stock, the class sold to the wider market, carries one vote per share [4]. Together with the other classes, it splits at most 49.9% of the vote on those matters [3][2]. Strategic partners get minimal voting rights [5].
Board seats are where that reach stops. The Long-Term Benefit Trust fills four of the seven seats the report describes, a majority [3]. So the founders' share wins any shareholder vote on key matters but cannot seat a board majority on its own. I think that split is the best-designed part of the plan. According to mezha.net, it spreads influence between the chief executive and a group of working cofounders, unlike the SpaceX model that concentrates power in Elon Musk [17].
The arrangement also has a sunset. Class F loses its extra voting weight gradually once the founder group shrinks past a threshold, and the available text of the report cuts off before the number [15]. For now the group is intact. Its members left OpenAI in 2020 and remain a close group five years after founding Anthropic [16].
For an API customer, the prospectus warning about decisions that cut against financial interests [6] also covers access decisions. This risk factor describes the governance working as intended. Anthropic has already restricted or delayed capabilities on safety grounds, and it gave Mythos Preview, a model especially strong in cybersecurity, to a limited set of users [12]. The prospectus says it passed on image and video generation models to keep compute on research and safety [13]. It will also remain a Delaware public benefit corporation, a status that lets its leaders weigh society's interests alongside investors' [8].
The details come from mezha.net's report on the prospectus, which cites Reuters [18]. An Anthropic representative declined to comment [7].
What to watch
- The full prospectus terms setting the membership level at which Class F starts losing its extra voting weight.
- Whether limited releases like Mythos Preview become the default for new Claude capabilities after the listing.
- Any change to strategic partners' minimal voting rights before the shares price.