Shivon Zilis, Neuralink's chief operating officer and Elon Musk's longtime confidant, testified in federal court in San Francisco late last week that Musk offered Sam Altman a seat on the Tesla board in late 2017 and enlisted OpenAI staff for what she described as "months of free work" on Tesla's Autopilot team during the period when OpenAI was still a nonprofit. [1] Her three-day testimony, which closed Friday, was the prosecution-of-motive segment in Musk v. Altman, the Brockman fiduciary-trial that wraps this week with closing arguments and jury instruction. [2] The trial wrap lands two trading days before Cerebras Systems prices its initial public offering at a range Reuters reported Sunday afternoon will lift to $150 to $160 a share, up from the $125 to $135 the May 10 paper carried. [3]

The paper's Saturday account of the Zilis-week midpoint named the Brockman five-stake roll-call — OpenAI's roughly $30 billion stake, Stripe, Cerebras, CoreWeave, and Helion — as the fiduciary-disclosure document the Cerebras prospectus would have to absorb. The companion brief on the order book at twenty times oversubscribed at the prior range now reads against a range that has lifted twenty percent inside a weekend and against Wednesday's pricing window. The Burry-tracker handle on X named the execution-week landscape Friday evening: "Keep an eye on May 15."

What Zilis testified

The Altman board-seat offer is the load-bearing testimony for the prosecution's motive theory. Zilis testified that Musk in late 2017, in the months before the Tesla board declined Musk's own proposal to become Tesla's chief executive on top of the role he already held, offered Altman a board seat conditional on Altman taking on a senior advisory role at Tesla. [1] Altman declined. The offer is the prosecution's evidence that Musk's relationship with OpenAI was structurally entangled with his Tesla obligations and that his subsequent decision to leave OpenAI's board in February 2018 was, in the prosecution's framing, an act taken with fiduciary asymmetry between the two organizations.

The "months of free work" Zilis described — OpenAI engineers and researchers diverted to Tesla's Autopilot team in 2017 — is the operational counterpart to the board-seat offer. Zilis testified that the work was, in her view, "uncompensated and unattributed" between the two organizations. The Tesla side of the ledger did not, in her recollection, book a transfer-pricing entry. The OpenAI side did not book a research deduction. The two organizations were, on the labor-flow data, operating as a single corporate body during a period in which OpenAI was a nonprofit and Tesla was a public company. [4]

The defense's cross-examination focused on whether the labor flow was, in Zilis's contemporaneous understanding, authorized by both boards. Zilis confirmed she had not been at OpenAI in a board capacity during the period in question; her testimony was based on her conversations with Musk and her observation of OpenAI engineers in Tesla offices. The defense's framing — that Zilis's testimony was perception, not documentation — was the central exchange of the three-day window.

The prosecutorial arc

Greg Brockman, OpenAI's co-founder and president emeritus, testified Tuesday and Wednesday before Zilis took the stand. His testimony covered the five-stake roll-call: OpenAI (Brockman's roughly $30 billion equity-equivalent stake inside the for-profit restructuring), Stripe (early angel investment), Cerebras Systems (early-investor stake), CoreWeave (early-investor stake), and Helion Energy (board-level investment). The roll-call is the prosecution's documentary scaffold for the disgorgement question — the relief the plaintiffs are asking the court to award if the jury finds Musk's 2018 OpenAI departure breached his duty of loyalty.

The Cerebras stake is the cross-cut. Brockman's investor position in Cerebras — disclosed on the courtroom record last week — places him simultaneously inside two operative trading windows this week. Wednesday's IPO pricing will set the public-market valuation of a chip company whose largest customer is OpenAI, whose largest equity holder is OpenAI by the customer-revenue concentration measure, and whose early-stage investor is a co-defendant in a federal trial that ends three days before pricing.

The Wednesday window

Cerebras priced its initial public offering at $125 to $135 a share when the roadshow opened. Sunday afternoon's Reuters break said bookrunners are lifting the range to $150 to $160 a share, taking the offering count from twenty-eight to thirty million shares and the maximum aggregate raise to $4.8 billion from $3.5 billion. [3] At the top of the new range, the company would be valued at roughly $22 billion before the underwriters' overallotment exercises. The book is past twenty times the offered shares at the prior range.

The structural feature of the lift is that it happens inside the trial wrap. The bookrunners — Goldman Sachs, Morgan Stanley, JPMorgan, and Citigroup — have a prospectus on file with the Securities and Exchange Commission that does not, in its risk-factor section, list "co-founder of customer is co-defendant in federal trial concluding three days before pricing" as a risk. The prospectus does list customer-concentration risk: OpenAI is more than eighty percent of Cerebras's revenue. The trial's documentary output — Zilis's "months of free work," Brockman's five-stake roll-call, the Altman board-seat offer — is now part of the public record the prospectus's customer-concentration risk runs against.

The two trading days between Friday's testimony close and Wednesday's pricing are the underwriting window inside which the bookrunners decide whether the lifted range holds. The structural read, on Monday morning, is that the order book is at twenty times the offered shares at the higher range, which means the demand has, so far, priced through the trial's documentary risk. The fiduciary-disclosure question is whether the bookrunners' final pricing language Wednesday acknowledges the trial output or absorbs it into the customer-concentration risk factor as already-disclosed.

What the wrap is, in disciplined-cohort terms

The Brockman trial is the disciplined-cohort thread's undisciplined-cohort mirror. The disciplined cohort — Apple at $100 billion in buyback authorization, Berkshire at $325 billion in cash, Pfizer at zero fiscal-year buybacks — runs its execution week this week against a cohort whose load-bearing capital-return events are not buybacks but warrant grants, related-party transactions, and prospectus revisions inside open court records. The Brockman five-stake roll-call is the largest documented private-market stake set in a single trial in American history, and the Cerebras pricing window puts the most contested of those five stakes onto the public market while the trial is still in closing.

The Berkshire 13F Thursday — the first under Greg Abel as chief executive — is the audit instrument for the disciplined cohort. The Cerebras pricing Wednesday is the price-discovery instrument for the undisciplined cohort. Both happen this week. The trial wrap is, in this register, the disclosure event the undisciplined cohort's pricing instrument cannot un-read.

-- THEO KAPLAN, San Francisco