the nonprofit was the bait. the cap table was the switch.

the consensus didn't price discovery

for two years the dominant framing on musk v altman has been that the lawsuit is performative, the donations are a sunk cost, the conversion is a fait accompli, and the case will settle small if it doesn't get dismissed first. that framing was load-bearing on a lot of bull-tape valuations. anthropic at $900b, openai approaching $1t in private market pricing, the closed-frontier rent extraction phase getting underway. all of it priced approximately zero structural risk to the largest ai company in the world.

week one of trial in oakland did something the consensus didn't price. it put the artifacts into the public record. discovery exhibits, birchall's testimony about the $97.4b bid, musk's "bait and switch" text to altman, the unredacted version of what altman wrote back. these are now binding factual record. they are also extremely bad for openai on the merits, in a way the "performative lawsuit" frame did not anticipate.

the cap table is the smoking gun

here is the load-bearing fact, and i have not seen anyone in the consensus tape sit with it long enough. when openai converted to a public benefit corporation in october 2025, the nonprofit was given 26% of the resulting entity, and microsoft was given 27%.

read that ratio twice. the charitable foundation that, under the founding charter, was supposed to be the steward of agi for the benefit of humanity, ended up with a smaller equity slice in the for-profit successor than the corporate partner that arrived in 2019. you do not need a smoking-gun email when the cap table is the smoking gun.

the case theory is not complicated once you see this number. the nonprofit's contribution to openai group pbc was the entire underlying technology, the entire research staff, the entire founding mission. microsoft's contribution was roughly $10b in late-2022 investment plus azure credits and the post-2019 partnership. the conversion priced the nonprofit's contribution at less than the partner's. that is a measurable transfer of value out of the charitable mission, and "you can't just steal a charity" is the way musk's counsel framed it on day one.

altman was on both sides of the desk

the most damaging single piece of testimony in week one came from jared birchall, who runs musk's family office and helped organize the february 2025 $97.4b investor bid for openai. on the stand, birchall said the bid was not actually a takeover attempt. it was structured to force a market valuation of the nonprofit's assets ahead of the pbc conversion, because the people structuring that conversion had a conflict.

specifically: sam altman was negotiating on behalf of openai's nonprofit and openai's for-profit arm at the same time, while holding compensation exposure to the for-profit side of the deal. that is a textbook self-dealing posture in nonprofit governance law, and birchall's testimony put it on the record without ambiguity. openai's lawyers spent the rest of the day trying to strike portions of his testimony, which is a tell on its own.

the $97.4b number is now the market comp the court has on record for the nonprofit's pre-conversion assets. if you accept that comp, the 26% slice the nonprofit took at conversion implies the resulting entity was valued at roughly $375b in october 2025 for the purpose of dilution math. openai's actual valuation at and after conversion ran multiple times higher, depending on which secondary print you take. the spread between those two numbers is the disgorgement claim. it is a concrete, computable spread. the court can issue a finding on it.

the steelman

musk has his own credibility problems and the trial surfaced them on the same week the cap table did. he admitted on the stand that xai "partly" distills openai's models, which is the kind of admission that follows a witness through cross. the judge said in open court that "i suspect there's plenty of people who don't want to put the future of humanity in mr. musk's hands," which is not the line you want to read in your trial transcript if you are the plaintiff. and the april amendment directing damages to openai's charity and seeking altman's removal from the board reads as a litigation positioning move more than a pure policy ask.

the bull case for openai is that gonzalez rogers does not love the plaintiff, finds for him on the narrowest possible grounds if at all, and the structural remedy ends up looking like a governance audit rather than an unwinding of the conversion. that is a coherent case. it is also a case where the headline outcome can be "musk loses on damages" while the structural finding still includes a self-dealing determination against altman. damages and structure are decoupled in equity. that is the part of the procedural design that the bull tape has not internalized.

the procedural design is the part to actually watch

gonzalez rogers seated a nine-person advisory jury and reserved the binding ruling for herself. the case is being tried in equity, not at law. that structure exists for exactly this kind of dispute, where the plaintiff is asking for remedies a jury can't grant, like rescission of a corporate conversion or removal of an officer, and where the court wants community sanity-checking on novel facts but needs to write the order itself.

what this means in practice. the judge can issue a ruling that splits the questions. she can find for openai on damages while finding for musk on the self-dealing predicate. she can decline to rescind the pbc conversion while ordering a structural remedy at the board level. she can require a re-marking of the nonprofit's equity share in light of the $97.4b market comp without touching the partnership with microsoft. each of these is a partial outcome the consensus tape has rounded to zero. each of them moves the markup that closed-frontier private valuations are sitting on.

what to do with this

three things hold up under the receipts.

first, "the lawsuit is a distraction" was always the weakest part of the openai bull thesis, and week one made that legible to the legal market in a way it wasn't legible a month ago. the kalshi spread between anthropic and openai on the path to $1t is partly priced on relative narrative momentum, but the structural-risk pricing is going to start moving as week two adds brockman's testimony and stuart russell's safety framing.

second, xai inherits a brand asset regardless of the dollar outcome. the frame that altman negotiated both sides of the conversion while the nonprofit was diluted to a minority is now on the public record. that frame is durable. it is the kind of thing that shows up in pitch decks, government filings, and procurement conversations for years. the only frontier lab whose ceo has not been on the stand explaining a nonprofit-to-pbc pivot is the one whose ceo is the plaintiff in the trial that put the pivot on trial.

third, settlement math just got harder for openai's side. the structural finding risk is now non-trivial, the disgorgement comp is computable, and a judge in equity has wider latitude than a settlement memo. settling small required the consensus framing to hold. week one moved it.

the nonprofit was the bait. the cap table was the switch. discovery is now public record, and altman was on both sides of the desk that drew it.