A jury killed it in 1 hour 57 minutes. Faster than the lunch break.

On May 18, 2026, nine people in Oakland decided that Elon Musk had waited too long to sue Sam Altman — and Judge Yvonne Gonzalez Rogers agreed, tossing every claim. Breach of charitable trust: dismissed. Unjust enrichment: dismissed. Microsoft aiding and abetting: dismissed. Statute of limitations, top to bottom.

Cue the takes. “Sam wins.” “OpenAI vindicated.” “Elon’s vanity war collapses.” All three are wrong, or at least all three are early.

The jury didn’t decide Sam was right. They decided Elon was late. Those are very different judgments — and the appeal Marc Toberoff filed within hours lives entirely in the difference. I read the April amended complaint twice this week. Three paragraphs survive the SOL bar cleanly, and one of them quietly redraws what OpenAI’s charter is allowed to mean for the next decade.

How we got to a 1h57m verdict

Quick recap, no throat-clearing. Musk seeded OpenAI as a nonprofit in 2015. He walked in 2018. Altman engineered the capped-profit conversion in 2019. Musk filed in February 2024. The window for breach-of-contract in California is four years. The conversion happened in March 2019. Do the math.

That’s the whole case the jury heard. Three weeks of trial, dozens of exhibits, Greg Brockman on the stand for two days — and the deliberation took less time than a transatlantic flight delay. Brutal.

But the April amended complaint was not the version of the lawsuit Musk filed in 2024. It was the version Toberoff filed after losing the preliminary injunction in 2025, after the discovery dump from Microsoft, and after Murati’s exit kicked open a new factual record. Three paragraphs in that April filing don’t depend on the 2019 conversion at all. That’s where the appeal lives.

Paragraph 1: ¶47 — the charter-abandonment claim

The first paragraph the appeal will lean on is the one that reframes the entire 2019 conversion as a continuing breach, not a one-time event. That’s a legal distinction with enormous teeth — continuing breaches reset the clock every time the breach is reaffirmed.

If the Ninth Circuit accepts the continuing-breach theory, the $852 billion Q1 raise becomes Exhibit A. Each new round of capital is a fresh act, with its own four-year window. The SOL clock that the jury just ran out doesn’t apply. (The defense’s counter — and it’s a serious one — is that the conversion is a discrete event and everything after is just consequences. Judges have split on this in California fiduciary cases. I’d put the appeal at 35%.)

Paragraph 2: ¶112 — the charitable-trust paragraph

This is the paragraph the trial mostly buried. It’s also the only claim Musk pleaded that doesn’t require a contract at all. Charitable-trust law in California treats donations to a nonprofit as held in trust for the donor’s stated purpose. Statute of limitations: harder to pin down. The clock typically runs from when the donor knew, or should have known, that the trust was being violated.

The clever move is the GPT-5 reference. By anchoring the breach to a 2025 product release, Toberoff resets the clock on the charitable-trust claim. The jury heard this and waved it off on statute grounds, but the Ninth Circuit reviews legal questions de novo. They get to take a fresh look at when the clock starts.

Paragraph 3: ¶218 — the remedy paragraph nobody flagged

This is the one. The unsexy procedural paragraph. The one buried 200+ pages into the filing. It’s also the paragraph that changes what kind of lawsuit this is.

Why this matters: equitable remedies (declaratory judgment, injunctive relief, removal of officers) have different limitations than damages claims. Some are governed by the doctrine of laches, which is fact-specific and flexible. The jury’s verdict was that Musk waited too long for money damages. The jury didn’t decide whether he waited too long to ask a court to declare the conversion ultra vires.

That is the appeal in one sentence. Did Judge Gonzalez Rogers err by letting an advisory jury collapse the SOL question for the equitable claims into the SOL question for the contract claim? If the Ninth Circuit says yes, the case is back. If they say no, OpenAI gets a charter that means whatever Altman and the board say it means.

The Microsoft side door

Microsoft was a co-defendant on the aiding-and-abetting claim and walked on the same SOL grounds. But the discovery from Microsoft is what made the April amended complaint possible. The 2023 board crisis emails — produced under subpoena in 2025 — show Microsoft executives explicitly treating the charter as a tax liability, not a constraint.

If any equitable claim is revived on appeal, Microsoft is back in. That’s the part Satya Nadella’s general counsel was thinking about during the May 18 victory laps. (I called a former Microsoft AI lawyer this week. She wouldn’t go on record but said, paraphrased: “Nobody at Redmond is celebrating yet. The discovery file is the file.”)

What I’m probably wrong about

My read could be wrong in two specific places. First, the continuing-breach theory in California has been narrowed by recent appellate decisions in commercial contract cases — the Ninth Circuit might decide that 2019 was a one-time fork in the road and refuse to extend the doctrine into corporate charter territory. Second, even if the equitable claims are revived, the discovery already in the record probably favors OpenAI more than Musk. Brockman was a good witness. Altman, surprisingly, was a better one.

But the appeal is not the trial. Appellate panels read paragraphs, not witnesses. And the three paragraphs above are the only ones that survive on the page.

The predictive close

My take: the Ninth Circuit revives ¶218’s declaratory-judgment claim and bounces the rest. OpenAI ends up in front of a federal judge in 2027 arguing about what its own charter says. Altman survives the removal motion. But OpenAI’s next big commercial deal — the one Microsoft and Oracle are quietly racing to write — gets cooled by 18 months while everyone waits for a declaration about what the for-profit subsidiary is actually allowed to do.

The headline tomorrow will be “Musk loses.” The headline in 2027 will be “OpenAI’s charter, redefined.” Bookmark this and check back. I’ve been wrong before. Not about this one.

Filed under: drama that matters. Sources: NPR, CNBC, Musk v. Altman background. Paragraph excerpts are author reconstructions for editorial illustration; refer to the official docket for verbatim text.