• On 25 March 2026, Judge Yvonne Gonzalez Rogers denied efforts to exclude C. Paul Wazzan's opinions. That was an evidence ruling: it permitted the methodology to be presented if a remedies phase became necessary; it did not validate a $134 billion award.
  • The remedies phase never arrived. On 18 May, a nine-person advisory jury unanimously found Musk's remaining claims untimely, the judge adopted that finding and dismissed them. The merits of the alleged charitable-trust breach and the damages model were not decided.

$134 billion was the top of an expert range

Musk's damages expert, financial economist C. Paul Wazzan, assumed liability and estimated the share of alleged wrongful gains attributable to Musk's early financial and non-financial contributions. The defendants' filed motion records approximately $38 million of financial contributions and an attribution range of 50% to 75% of the OpenAI nonprofit's value.

Wazzan put the alleged OpenAI wrongful gains attributable to Musk at $65.50 billion to $109.43 billion, and Microsoft's at $13.30 billion to $25.06 billion. Adding the two ranges gives $78.80 billion to $134.49 billion. The familiar $134 billion headline was therefore a rounded upper bound, not a judicial valuation, agreed liability or award.

The March order decided admissibility, not value

The OpenAI defendants asked the court to exclude Wazzan's opinions three through six under Federal Rule of Evidence 702 and Daubert; Microsoft separately challenged the attribution of its gains. The court denied both challenges in its written 25 March order.

The judge found that Wazzan had explained his 50% to 75% estimate through Musk's financial share, a proposed 2017 equity stake, his xAI stake and claimed non-monetary contributions, supported by cited literature and experience. Objections about cherry-picking and comparisons with other contributors went to weight and credibility rather than the threshold for admission. The court also said any double recovery could be addressed later.

Admission did not turn the model into a court finding

A damages expert may assume liability in order to model a remedy. That does not prove liability, causation or the amount ultimately recoverable. The order expressly framed the question as whether the method was too unreliable or fact-poor to be heard, not whether the court agreed with it.

This distinction corrects the old article's central ambiguity. Reports of skeptical comments at a hearing can describe courtroom pressure, but the signed order is the operative result: the testimony survived an admissibility challenge. It remained contestable through cross-examination, competing evidence and the court's control of equitable remedies.

The proposed beneficiary changed before trial

By 17 April, Musk said he would not seek a remedy benefiting himself personally and argued that any disgorged assets should return to OpenAI's charitable arm. The pretrial order recorded the judge's significant reservations and required a verified waiver covering Musk, xAI and any fund he controlled if he maintained that position.

The same order bifurcated liability from remedies, kept Wazzan away from the advisory jury because his evidence concerned remedies only, and warned that an adverse limitations finding would probably end the case. Musk then voluntarily dismissed his fraud and constructive-fraud claims before trial; the surviving theories proceeded without a damages presentation to the jury.

Timing ended the case before merits or remedies

Trial began in Oakland on 27 April. On 18 May, after less than two hours of deliberation, the nine-person advisory jury unanimously found that Musk had brought the remaining claims too late. Associated Press reported that Judge Yvonne Gonzalez Rogers accepted the verdict as the court's own and dismissed the claims.

That sequence matters legally and editorially. The jury did not reach whether OpenAI, Sam Altman or Greg Brockman breached a charitable trust, whether Microsoft aided any breach, or whether anyone received wrongful gains. Wazzan's figures were never tested in a remedies phase and no part of the $78.80 billion to $134.49 billion range became an award.

An announced appeal is not a merits judgment

Musk said he would appeal to the Ninth Circuit. This review records that announcement, not an assumption about a later appellate filing or outcome. A future notice of appeal, briefing schedule, ruling or remand should be reported from the docket when it exists.

For readers assessing AI governance, the durable lesson is narrower than either side's rhetoric: a large expert number can pass an evidence gate and still never control the case. Procedural timing, liability, causation and remedy are separate questions and must stay separate in headlines, metadata and analysis.

What to watch

  • A docketed notice of appeal and the issues actually presented to the Ninth Circuit.
  • Any post-trial order clarifying the judgment, costs or remaining counterclaims.
  • Whether appellate review addresses limitations only or reaches any underlying charitable-law question.
  • OpenAI governance and nonprofit-control disclosures that are independently verifiable outside the litigation rhetoric.
  • Future damages reporting that distinguishes an expert range, an admitted opinion, a requested remedy and an awarded amount.

Sources