Elon Musk's lawsuit against OpenAI began as a claim about broken promises. What it has become, three weeks into testimony in a federal courthouse in Oakland, California, is something more revealing: a live audit of who actually controls the AI industry, how much they are worth, and what they were really thinking when they built it.
The numbers that came out this week alone are staggering. Ilya Sutskever, the scientist who effectively built the research engine behind ChatGPT, holds around $7 billion in OpenAI shares. Greg Brockman, the co-founder who helped start the company as a nonprofit, holds about $30 billion. Microsoft, which began by offering OpenAI discounted computing services as a goodwill gesture, now holds a 27% stake worth roughly $135 billion, after investing $13 billion over six years. An internal Microsoft email from 2022 showed Satya Nadella warning colleagues that Microsoft would lose $4 billion in the coming year if the OpenAI partnership did not change direction fast.
None of this is what a charity looks like. And that is precisely Musk's point.
Musk donated $38 million to OpenAI in its early years and claims the company's founders promised it would remain a nonprofit. OpenAI and everyone who testified for it say no such promise was ever made in writing, and that moving to a for-profit structure was the only realistic way to raise the billions needed to stay competitive. Sutskever put it plainly on the stand: you cannot build AI at this scale on donations alone.
Sutskever's own story is the clearest illustration of just how strange this whole situation is. He left Google, turning down a $6 million annual salary, to help build OpenAI. He then helped fire the CEO he had worked alongside for years, left the company, and started a competing lab called Safe Superintelligence. That lab, with roughly 20 employees and no product, was valued at $32 billion by investors earlier this year, purely on the strength of Sutskever's reputation. Meta tried to buy it. He said no.
Meanwhile, the very safety team Sutskever led at OpenAI, whose job was to make sure future, more powerful AI systems could be controlled, was disbanded shortly after he left. He mentioned this in court. Nobody from OpenAI disputed it.
The question of who wins the verdict matters, but perhaps less than it seems. Prediction markets currently put Musk's chances at winning at around 40%, down from 60% when the trial started. His own testimony hurt him; he admitted his AI company trained its models using OpenAI's outputs, which violates OpenAI's terms of service. A judge who noticed the irony pointed out that Musk was simultaneously warning about AI dangers while building his own AI company as fast as possible.
If Musk wins in any substantial way, the consequences for businesses that use OpenAI tools are real. A forced structural reversal or leadership change at OpenAI would create serious uncertainty at a company now woven into the daily operations of hundreds of thousands of organizations. OpenAI's planned public share offering would almost certainly be delayed or cancelled.
If OpenAI wins, the ruling would effectively legitimize the broader pattern of AI companies launching as mission-driven nonprofits, attracting public goodwill and top talent on that basis, and then converting to profit-seeking structures once the money gets serious. That pattern is already the norm. A court blessing would make it permanent.
What the trial has already done, regardless of the verdict, is force into public view a set of facts that were previously buried in private emails and locked board rooms. The people who control the most powerful AI systems in the world are simultaneously competitors, former friends, and witnesses against each other. The governance structures they built were, as Microsoft's own CEO described the 2023 Altman firing, amateur. And the financial stakes are so large that almost nobody in the room has a genuinely neutral interest in how AI develops.
For any organization that relies on AI tools built by OpenAI, or is considering doing so, that last point is the one worth sitting with.