ANKARA, TURKIYE - MARCH 14: In this photo illustration, the logo of 'OpenAI' is displayed on a mobile phone screen in front of a computer screen displaying the photographs of Elon Musk and Sam Altman in Ankara, Turkiye on March 14, 2024. (Photo by Muhammed Selim Korkutata/Anadolu via Getty Images)
Anadolu via Getty Images
Jury selection in Musk v. Altman began Monday in Oakland, with roughly $134 billion in damages and the future of OpenAI’s planned IPO riding on the verdict. But the case is litigating a question compute economics already answered — and every frontier AI lab founded in the last five years has already structured around that answer.
The case began with more than two dozen counts and is now effectively about whether a charitable trust was used to build an $852 billion company. Witnesses include the company’s cofounders and investors: Elon Musk, OpenAI CEO Sam Altman and President Greg Brockman and Microsoft CEO Satya Nadella.
Underneath the Shakespearean filings and the social‑media name‑calling sits a much narrower structural question: was OpenAI legally allowed to evolve from a 2015 nonprofit into a capped‑profit subsidiary in 2019, then into a public benefit corporation in October 2025?
Capital has been answering that question for a decade. Frontier training runs cost an estimated $40 million for GPT-4 in amortized hardware and energy, and they’re projected to reach $1 billion by 2027. OpenAI’s 2024 compute spend is widely estimated in the mid‑single‑digit billions. Stargate, the OpenAI-SoftBank-Oracle consortium, has committed $500 billion over four years to build 10 gigawatts of AI capacity in the United States. None of those numbers fit inside a charity; they barely fit inside a public market.
When OpenAI completed its restructure on Oct. 28, 2025, the OpenAI Foundation walked away with a 26% stake in the for-profit PBC — worth roughly $130 billion at the company’s $852 billion valuation. That single stake is now larger than the Gates Foundation’s entire $86 billion endowment. Microsoft holds another 27%, worth about $135 billion. Employees and investors hold 47%. The restructure didn't kill the nonprofit: it capitalized it. The OpenAI Foundation became one of the largest grantmakers on earth, but its grantmaking capacity now depends on the for-profit it was built to oversee.
It's no wonder that when the people who knew OpenAI's structure best went off to build something new, they built it differently. And differently from each other.
Anthropic was founded in 2021, five years before this trial, by seven former OpenAI researchers, including Dario Amodei — the same Dario Amodei who, as OpenAI’s VP of research, helped write much of the charter Musk now wants enforced. He incorporated his next company as a public benefit corporation on day one and layered a Long-Term Benefit Trust on top, with the power to elect a growing share of the board independently of any investor. The structure took compute economics as a given: it didn't pretend frontier AI could be funded any other way.
Musk's answer was simpler. He launched xAI in 2023 — three years before the courtroom — as a straightforward for-profit, with no public-benefit framing and no charitable trust. In February 2026 it merged into SpaceX in a $1.25 trillion all-stock deal — $1 trillion for SpaceX, $250 billion for xAI, the largest merger ever recorded.
Two former OpenAI co-founders chose two different structures. Both for-profit, neither hybrid. Neither tried to preserve the original charter, because both had already done the math the courtroom is about to redo.
Even if Musk wins on all his claims and the judge orders the disgorgement he's asking for, the structural question is settled. Every founder of every frontier AI lab now plans around the same compute gravity — Anthropic built protections in, xAI left them out, OpenAI tried something in-between and is now in court for the in-between. A retroactive verdict reshuffles ownership at one company, but it doesn't shrink a training run.
And if Musk loses, the case still does work. Four weeks of discovery — internal emails, board minutes, Brockman's unsealed personal notes about wanting to become a billionaire, Musk's 2016 email calling Bezos "a bit of a tool" — will enter the public record before either OpenAI or SpaceX prices its IPO. The trial is partly a damages case. It is also a forced disclosure event, on a calendar that lines up almost perfectly with the two largest tech IPOs ever recorded.
The court will tell us whether OpenAI broke the original charter in 2019. It won't tell us whether mission governance can survive the capital required to compete in 2026. That answer is being written every day in billion-dollar training runs and trillion-dollar mergers.
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。