Musk vs Altman: How a $1 Billion Nonprofit Became a $500 Billion Courtroom Battle

In December 2015, two men sat together at the launch of an unusual experiment. One was a 30-year-old former Y Combinator president named Sam Altman. The other was a 44-year-old serial founder of car and rocket companies named Elon Musk. They co-chaired the launch of a new entity called OpenAI — a Delaware nonprofit that would, in their own founding charter, exist to "ensure that artificial general intelligence benefits all of humanity." The pledge alongside it was an eye-watering $1 billion in funding.

Eleven years later, those same two co-chairs are on opposite sides of a federal courtroom in Oakland, California.

Musk's lawsuit alleges that the nonprofit they founded together was, in his lawyers' opening words to the jury, "stolen" — that Altman and his colleagues quietly steered a charitable mission toward a $500 billion for-profit colossus once Microsoft's billion-dollar cheques started arriving. OpenAI's defence is equally sharp: that Musk himself wanted the for-profit pivot, drafted his own version of it in 2017, quit when he could not run it, and is now litigating to hobble a competitor to his own AI company, xAI.

This is not a story about who is morally right. It is a story about how structure determines destiny. Every fight in the OpenAI saga — the 2019 capped-profit pivot, the 2023 boardroom revolt, the 2025 PBC restructuring, the 2026 trial — flows from the original 2015 decision to file as a nonprofit. Each subsequent decision carried a consequence that none of the founders fully foresaw at the time.

For every Indian founder building an ambitious company, every board member sitting on a startup or listed-company board, and every long-term investor whose portfolio is increasingly exposed to the AI capex cycle, this case is the single best free education in corporate governance available right now. So let's read it carefully.

1. The Players, The Stakes

Before getting into the timeline, it helps to know who is actually fighting whom — and what each side says they want from the courtroom. Most public commentary treats this as "Musk vs Altman," but the legal pleadings are broader and the relief sought is dramatic.

The judge is Yvonne Gonzalez Rogers of the U.S. District Court for the Northern District of California — the same federal judge who, in earlier rounds, denied Musk's bid for a preliminary injunction to halt OpenAI's restructuring but allowed the core breach-of-trust claims to proceed to a jury trial.

Both characterisations above are partisan. The truth, as it usually does in corporate disputes, lives somewhere in the unread fine print of term sheets, emails and board minutes — many of which are now exhibits in the case.

2. December 2015 — The Original Bargain

OpenAI was incorporated as a Delaware nonprofit on December 11, 2015. The founding cohort included Altman, Musk, Ilya Sutskever, Greg Brockman, Trevor Blackwell, Vicki Cheung, Andrej Karpathy, Durk Kingma, John Schulman, Pamela Vagata and Wojciech Zaremba. Altman and Musk served as co-chairs.

A funding pledge of $1 billion was announced — backed by Altman, Brockman, Musk, Reid Hoffman, Jessica Livingston, Peter Thiel, Amazon Web Services and Infosys. According to nonprofit tax filings later cited by Wikipedia and various journalistic accounts, however, only a small fraction of that pledge — roughly $133 million by 2021 — was ever actually received. This gap between pledged capital and delivered capital is a recurring undercurrent of the entire dispute.

The founding charter committed OpenAI to a single, unusually moralistic goal: ensuring that artificial general intelligence "benefits all of humanity," with an explicit intention to publish research openly. The structure was deliberately designed to keep mission ahead of profit.

That single sentence — Musk himself, in November 2015, before OpenAI was even incorporated — is now Exhibit A for both sides. Musk's lawyers argue he was outvoted on structure but trusted the mission. OpenAI's lawyers argue he knew from day one that a pure nonprofit form was suboptimal for raising capital, and therefore cannot now claim he was deceived when the company eventually pivoted to attract investment.

The single most consequential decision

The most consequential decision of OpenAI's first decade was the choice to file as a Delaware nonprofit corporation rather than a Delaware C-Corp. Every fight that follows — the capped-profit pivot, the Microsoft investment, the boardroom firing, the PBC conversion, and now the lawsuit — flows from the friction between that 2015 charitable filing and the commercial gravity that arrived with ChatGPT seven years later.

This is the first lesson of the case, hidden in plain sight: the form you incorporate in is not just paperwork. It is a legal commitment that reaches forward into every future fundraise, hire, partnership and crisis. It is, as one corporate-law professor put it in a TIME interview, an attempt to "encode values into a structure that capital will respect." Whether such encoding can ever truly hold under real capital pressure is what the entire OpenAI saga is now litigating.

3. 2017–2018 — The First Fracture

By 2017, two problems were colliding inside OpenAI. First, Google's DeepMind was advancing rapidly. Musk has testified at trial that he became increasingly worried OpenAI could not catch up on a donations-only model. In a 2016 email cited at trial, Musk wrote to a colleague at Neuralink that setting OpenAI up as a nonprofit "might, in hindsight, have been the wrong move."

Second, the people running OpenAI began discussing a for-profit subsidiary that could raise conventional venture capital while remaining controlled by the nonprofit parent. According to OpenAI's published timeline and court exhibits introduced by their counsel, Musk himself drafted and incorporated a public benefit corporation on September 15, 2017 as a proposed for-profit vehicle for OpenAI. The proposal would, per OpenAI's account, have given Musk majority control of the cap table and the board.

Musk's testimony in the 2026 trial does not deny the proposal. Instead, he argues — through his lawyers — that any control he might have held would have been temporary and would have diluted as institutional investors came in. OpenAI's lawyers see this differently and have pressed the point repeatedly on cross-examination: that Musk wanted full control, the others refused, and Musk walked out.

One detail is undisputed and matters: Musk had pledged a far larger sum than he ultimately delivered. Musk's lawyers say he poured roughly $38–45 million into OpenAI over five years. OpenAI's lawyers note this was a fraction of the pledge, and that the shortfall left the organisation scrambling for capital — which is exactly what made the Microsoft cheque hard to refuse.

4. 2019 — The Birth of "Capped-Profit"

With Musk gone from the board, OpenAI's remaining leadership designed one of the most unusual corporate structures in modern Silicon Valley history: a capped-profit subsidiary sitting beneath the original nonprofit parent.

Mechanically, here is how it worked at launch:

• OpenAI, Inc. — the original Delaware nonprofit — remained the controlling parent. Its board retained the legal duty to humanity, not to shareholders.
• OpenAI LP (later restructured as OpenAI Global, LLC) — a for-profit subsidiary fully governed by the nonprofit board — was permitted to raise venture capital and issue equity-like interests.
• Investor returns were capped at 100x their initial investment. Anything above the cap reverted to the nonprofit. The cap was later reduced through subsequent rounds, eventually — per third-party reporting cited by The OpenAI Files — falling toward "single digits" with provisions to grow 20% annually from 2025.
• The operating agreement explicitly warned investors that profit was not guaranteed and that they should think of investments "in the spirit of donations," per Northwestern law professor Jill Horwitz's reading of the documents in TIME.

A few months after this structure was announced, Microsoft invested $1 billion. By 2023, that figure had grown to over $13 billion in cumulative cash and Azure compute commitments. The capped-profit mechanism had done exactly what its designers intended — it had unlocked institutional capital while preserving, on paper, the nonprofit's primacy.

OpenAI's structural evolution at a glance

In retrospect — and this is observable from the public timeline rather than disputed — the 2019 structure was an attempt to have it both ways: charitable governance on top, venture economics below. That tension held for a few years. Then ChatGPT shipped on November 30, 2022, and the gravitational pull of the for-profit subsidiary became, in a literal sense, billions of times larger than the parent.

5. The Microsoft Tipping Point

Microsoft's $10 billion follow-on investment in early 2023 was the moment Musk has identified, in his own sworn testimony, as the point of no return. According to CNBC's trial coverage, Musk testified about his reaction in the fall of 2022 in striking terms — calling the deal a "bait and switch" in a text he says he sent to Altman.

OpenAI's documentary defence on this point is that Musk was told about the plan in 2018. At trial, OpenAI's counsel Savitt put a 2018 term sheet up on the courtroom screen — a four-page document that, per CNN's reporting, explicitly mentioned a $10 billion future fundraise target. Musk, on the stand, conceded he "did not read the fine print." Savitt's reply, as reported, was clipped: that the document was, after all, only four pages long.

Musk's argument here is the deepest analytical question the case raises, and one every founder of a mission-driven company should sit with carefully. Capital has its own gravity. No charter clause, profit cap, or board structure permanently neutralises that gravity once the cheque clears. The only real questions are how much friction you build in, and who sits on the board when the friction is tested.

OpenAI's counter is equally strong: that Musk had every opportunity to read the documents that disclosed exactly this trajectory in 2018, and that he chose not to engage with the fine print. Both can be true. The jury will weigh which carries more weight.

6. November 2023 — The Five Days That Shook AI

If the 2019 structure was the experiment, November 2023 was the stress test — and the structure essentially failed in slow motion over five days. This is the single most studied corporate-governance episode of the AI era, and every Indian board member should read it carefully.

What Helen Toner later said

Former director Helen Toner publicly explained the board's rationale months later in a podcast interview. According to her account (cited by Wikipedia's well-sourced summary and contemporaneous reporting), the board alleged that Altman had withheld information — including, she said, about the very release of ChatGPT and his ownership of OpenAI's startup fund. She also said that two other executives had reported what she described as psychological mistreatment.

An independent review later commissioned by the new board concluded that Altman's conduct "did not mandate removal." The independent reviewers did not, however, dispute the underlying accounts of board–CEO communication breakdown. These remain contested characterisations. Altman has consistently denied the more serious allegations.

The point for a corporate-governance student is not who is right. It is that the board's process was so flawed that the substantive question became almost impossible to answer cleanly afterward.

The governance lesson from those five days

Wired's Steven Levy and others compared the episode to the 1985 ouster of Steve Jobs from Apple. Yale governance scholar Jeffrey Sonnenfeld, in widely reported interviews, described the OpenAI board harshly. He is not a neutral commentator. But the underlying critique is hard to dismiss:

• The board did not consult its largest stakeholder (Microsoft) before firing the CEO.
• The board did not pre-coordinate with employees, who held what amounted to a soft veto via collective resignation threat.
• The board did not communicate a specific, defensible rationale at the moment of action. "Not consistently candid" is not a charge sheet.
• The board did not have a transition plan beyond installing an interim CEO.

A board can be right on the substance and catastrophically wrong on the execution. In this case, even if the directors' concerns were entirely justified — and reasonable people disagree — the manner of the removal made it logistically impossible for those concerns to win.

7. 2024–2026 — The Lawsuit Builds

Musk filed his original suit in California state court in February 2024. He withdrew it in June 2024 and refiled in federal court in August 2024, broadening the scope to include Microsoft as a co-defendant on an "aiding and abetting" theory.

The legal core of his claim is breach of charitable trust. The argument, as presented in the federal complaint and reiterated by Musk's counsel Steve Molo in his April 28, 2026 opening statement, is summed up in one line: the defendants, in his telling, "stole a charity." Specifically, the plaintiffs allege Altman and Brockman steered assets and opportunities away from the nonprofit and toward themselves and the for-profit subsidiary in a manner inconsistent with OpenAI's founding charter.

OpenAI's response, as articulated by Savitt at trial, has three pillars:

1. Statute-of-limitations defence: Musk publicly criticised OpenAI's Microsoft relationship as early as a September 2020 X post in which he wrote that OpenAI was "essentially captured by Microsoft." Counsel for Microsoft used this exact post in opening statements to argue Musk has known about the structure for years and should have sued earlier.
2. Documentary evidence Musk endorsed for-profit conversion: The 2017 PBC he himself created, the 2018 emails about $10B fundraising, and the term sheets giving him board control — all introduced as exhibits.
3. Competitive motive: Musk founded xAI, a direct competitor, in 2023. OpenAI's lawyers argue this is the actual reason for the suit. Their phrase, used repeatedly in court: this is not a fight about a charitable mission; it is a fight about who runs frontier AI.

8. 2025 — The PBC Restructuring

In late 2025, OpenAI completed a fundamental restructuring of its corporate stack. The high-level shape, according to OpenAI's own disclosures and Wikipedia's well-sourced summary as of April 2026:

• The for-profit subsidiary was converted from an LLC into a Delaware Public Benefit Corporation named OpenAI Group PBC.
• The original nonprofit was renamed the OpenAI Foundation.
• All directors of the PBC are appointed by the Foundation and removable by it at any time. As of April 2026, all but one Foundation director also sat on the PBC board.
• The 100x profit cap was, in effect, retired. PBCs do not require profit caps.

The new ownership math

The defenders of this structure — including OpenAI itself — argue that nonprofit control is preserved through the appointment power, and that the new structure simply makes capital raising (and a potential IPO, which Altman has mentioned publicly) feasible at scale. The critics — including the coalition of nonprofits and former OpenAI employees who wrote to the California and Delaware Attorneys General in April 2025 — argue something more subtle but important:

Both characterisations are partially true. PBCs in Delaware do require directors to balance shareholder returns against the stated public benefit purpose. But the standard of review is forgiving and the enforcement mechanism is weak. A nonprofit parent with 100% legal control of a subsidiary's commercial decisions is structurally a different beast from a PBC where the nonprofit owns a 26% economic stake and appoints (but cannot day-to-day direct) the board.

9. The 2026 Trial — Where Things Stand

The federal trial began in Oakland, California on April 27, 2026, before Judge Yvonne Gonzalez Rogers. As of early May 2026, here is what has happened in court:

The jury will ultimately be asked to decide several distinct questions. Did OpenAI breach a charitable trust? Were Altman and Brockman unjustly enriched? Did Microsoft aid and abet any breach? And — perhaps most consequentially — should the 2025 PBC restructuring be reversed?

10. Corporate Structure Decoded — In Plain English

Before the lessons, it is worth stripping out the jargon. There are five different corporate forms in this story, and understanding the differences is the entire game.

The signal hidden in the table

Notice the trajectory. OpenAI's structure has migrated down the table — from the strongest mission constraint (nonprofit) to one of the weakest (PBC). Each step was justified by the founders as necessary to attract capital. Each step weakened the legal mechanism that originally protected the mission. The Indian founder reading this should ask: is there any example, anywhere, of a mission-driven company moving in the other direction?

11. Seven Lessons Every Founder & Board Should Internalise

The point of a case study is not to pick a winner. It is to extract durable, transferable lessons. Here are the seven that hold up regardless of how the jury ultimately rules.

12. What Indian Founders & Investors Should Take From This

The OpenAI saga is a U.S. case under U.S. law, but the underlying dynamics are exportable. Three specific takeaways for the Indian startup and investing ecosystem:

For Indian founders

If you are building a company with a stated public-good purpose — climate, health, education, financial inclusion, AI safety — the OpenAI story should make you cautious about overdesigning your governance structure on day one. Bespoke structures that solve for mission preservation tend to break under capital pressure. A simpler design that you can actually defend in board meetings tends to outperform an elaborate one that no one outside the founder fully understands. Pick your hill carefully, paper your founder relationships clearly, and be honest with yourself about whether you want to run a venture-scale business or a foundation-scale one. The two have different operating logics and Indian capital markets will eventually force a choice.

For Indian boards (listed and unlisted)

The November 2023 episode is worth reading carefully if you sit on any Indian board, public or private. SEBI's listing obligations and disclosure requirements (LODR) and the Companies Act 2013 already require independent directors to consider stakeholder interests and document board reasoning. The OpenAI episode is a real-world illustration of why those requirements exist. A board that fires a CEO without a defensible written rationale, a stakeholder consultation, or a transition plan is a board that is one weekend away from a crisis it cannot contain. This applies to family-run promoter boards just as much as to professionally governed ones.

For Indian retail investors

OpenAI is not listed, and most Indian retail investors will not have direct exposure to it. But Microsoft is, indirectly, through international funds and increasingly through GIFT City and US-stock platforms. More importantly, the broader AI capex cycle that OpenAI helped catalyse is now flowing through Indian listed names — power utilities supplying data centres, Indian IT services that resell or build on top of frontier models, semiconductor packagers, and so on. The OpenAI story is a useful reminder that the corporate structure of the company whose technology powers your portfolio matters. Read the 10-K. Understand who controls the board. Pay attention when ownership structures change. The most consequential financial events often happen in governance documents that almost nobody reads.

13. The Verdict — Three Scenarios

Three rough scenarios are worth holding in mind as the trial proceeds.

Scenario A — Defendants win cleanly

The jury finds no breach of charitable trust, no aiding and abetting, no unjust enrichment. The 2025 PBC structure stands. Musk's lawsuit is dismissed. OpenAI proceeds toward a potential IPO, possibly within the next 12–24 months. This is, on balance, the outcome OpenAI's lawyers have set up the strongest documentary case for — particularly given the 2017–2018 emails and Musk's own PBC filing. It would also strengthen the legal viability of the PBC structure as a vehicle for mission-driven AI labs going forward.

Scenario B — Plaintiffs win on a narrow theory

The jury finds a partial breach — perhaps on unjust enrichment, perhaps on a procedural breach of charitable duty — without ordering full reversal of the restructuring. OpenAI pays damages or makes equitable adjustments to the Foundation's stake. Altman and Brockman remain in their roles. The PBC structure largely survives but with a higher Foundation share or stronger control rights. This is, in the view of most external commentators, the most probable middle-of-the-distribution outcome.

Scenario C — Plaintiffs win broadly

The jury orders reversal of the 2025 restructuring, possibly orders Altman's removal, and grants substantial damages. This would be a near-existential event for OpenAI as currently constituted, and would send shockwaves through the entire venture-backed AI sector. It is also the least probable outcome based on the documentary evidence introduced so far. But "least probable" is not "impossible," especially in a jury trial where the central question — was a charitable mission betrayed? — is as much a question of intuition as of law.

14. Frequently Asked Questions

15. Closing Thoughts

Strip away the personalities and the dollar figures, and the OpenAI case is, at its core, an old story told with new technology. A group of co-founders set out to build something ambitious. They disagreed on how to fund it. One left. The remaining ones built something enormous. Now the founder who left is asking a court to reverse the decisions made after his departure.

Stories like this have played out in tech repeatedly — Apple in the 1980s, Facebook in the 2000s, Twitter for most of the 2010s. The OpenAI version is unusually consequential because the technology at issue genuinely could reshape the global economy, and because the corporate structure is unusually creative. But the underlying lesson is unchanged: structure matters, process matters, paper matters, and the people on the board the day a hard decision has to be made matter most of all.

For founders, board members, lawyers and serious long-term investors, this case is the single best free education in corporate governance available right now. Read the filings. Watch the testimony. Form your own view. And — to take Musk's own most painful trial admission seriously — read the four-page document, all four pages, before you sign.