Transcription
Elon Musk is suing OpenAI for $150 billion. And the company that brought us ChatGPT is feeling the heat. Its own former CTO just testified that Sam Altman runs a company defined by chaos, deception, and executives pitted against each other. Co-founder Ilia Sutskever has revealed that he spent close to a year compiling a 52-page dossier documenting a consistent pattern of lying. And OpenAI's president, Greg Brockman, has disclosed a personal stake worth almost $30 billion and been forced to read excerpts from his diary in court. And while all of this was unfolding, rival AI company Anthropic went and leased the entire 220,000 GPU capacity of SpaceX's Colossus 1 data center on the 6th of May. So before retail investors buy a single share of OpenAI at a rumored valuation of $1 trillion, you need to see what is actually inside this deal.
Today, we look at exactly what was said under oath before the presiding judge in the case, why Sam Altman's $2 billion in personal side investments may be the single biggest red flag for institutional allocators, and how the real power moves are no longer happening at OpenAI. My name is Guy, and this is the Finance Bureau.
Now, for clarity's sake, it's important to know that Musk isn't suing OpenAI for personal gain. The $150 billion discordment he's calling for would be paid back to the OpenAI nonprofit entity rather than to Musk personally. The suit was filed in the Northern District of California with closing arguments running this week and an advisory jury of nine about to deliberate. The case alleges breach of charitable trust and unjust enrichment. And its argument is pretty straightforward. According to Musk, Altman and Brockman took intellectual property, talent, and compute originally donated to a nonprofit and rooted it into a for-profit machine for personal enrichment. Microsoft, OpenAI's biggest investor, is named as a third defendant for aiding and abetting. So Musk is not suing for a payday, but rather to prove that the most valuable private company on the planet was built by basically stealing a charity. And so having donated approximately $38 million in seed funding between 2015 and 2017, he's just put OpenAI's most senior former executives on a witness stand to make exactly that case.
Now to put the $150 billion into context, OpenAI closed a $122 billion funding round on the 31st of March at a post-money valuation of $852 billion with Andreessen Horowitz and SoftBank co-leading and Amazon and Nvidia coming in as the largest individual investors. The rumored IPO figure circulating among bankers right now is $1 trillion. So the damages figure represents roughly 15% of that valuation. In any other IPO process, that would be the headline risk factor on the very first page of the S-1. Here though, it's competing for attention with everything else that came out under oath, which brings us to Mira Murati.
Now, Murati was OpenAI's Chief Technology Officer from May 2022 until her resignation in September 2024, having originally joined the company in 2018 in a different senior role. Her video testimony was played in court on the 6th of May. "My concern was about Sam saying one thing to one person and completely the opposite to another person." She testified that Altman was creating chaos, that he was not always candid with the board, and that he actively pitted senior executives against each other to maintain control. The single most concrete allegation she made, though, was that Altman falsely claimed the legal department had cleared a new AI model in order to bypass the safety board. Given the many concerns around OpenAI's ethics and ethos, this amounts to a pretty awkward allegation.
The viral text messages from November 2023 were entered into evidence, too. When Altman, freshly fired by the board, texted Murati asking how things were going, she replied, "Sam, this is very bad." When he asked if he could come into the office, her reply was, "They don't want you to." Now you might be thinking, fine, this is just one disgruntled former executive being weaponized by a rival billionaire. But that is not how the SEC reads sworn testimony from a former C-suite officer. An S-1 filed by OpenAI in the next 18 months will have to disclose this testimony as material litigation risk under the rules. It's a matter of permanent public record and cannot be settled into silence. And it lands directly on the section of the prospectus that institutional allocators read first, which is the management and governance risk disclosure.
Now, before we get into Sutskever, Brockman, Microsoft, and the self-dealing problem that could be the most damaging piece of the entire trial, if you want to make sense of what actually matters across global finance, markets, and macro without drowning in noise, then you should sign up for the Finance Bureau newsletter. It's absolutely free. It lands in your inbox every single week, and we break down the stories that you need to be up to date with. All you have to do is click the link in the description or scan this QR code on your screen. And now back to the witness stand because the next witness up was the man who actually organized the November 2023 board firing.
Ilia Sutskever, OpenAI co-founder and former chief scientist, testified on the 11th of May that he spent approximately one year collecting evidence of what he described as a consistent pattern of lying by Altman. He authored a 52-page dossier presented to the board that triggered the firing. He disclosed his own OpenAI ownership stake at approximately $7 billion. And he confirmed under oath that he ultimately voted to reinstate Altman only as a Hail Mary, in his own words, to prevent the company's destruction during the employee revolt that followed. Remember, this comes directly from the co-founder who personally built the case for removal and then watched the structure he tried to fix snap back into place.
Then there's Greg Brockman, OpenAI's current president, also named as a defendant. Brockman disclosed his personal stake in the company at nearly $30 billion. But the more damaging evidence was his own diary, roughly 100 pages of it entered into the record. A 2017 entry asked in his own writing, "Financially, what will take me to $1 billion?" And in another 2017 entry, Brockman wrote that, "I would actually be warm to steal the nonprofit from him to convert to BC Corp without him." Although this is contradicted by other entries that say, well, pretty much the exact opposite. Whatever the truth is though, being forced to read from his own diary in court is probably not an experience Brockman will look back on all that fondly.
Anyway, on the same day that Sutskever gave evidence, Satya Nadella also took the stand. The CEO of Microsoft, the company holding a 26.79% stake in OpenAI worth approximately $228 billion, characterized the entire November 2023 board crisis as "amateur city." He testified he was never given a clear explanation for Altman's firing and critically that Musk never once contacted him during the entire crisis to raise concerns about Microsoft's investment despite having his contact information. Now, that cuts both ways for the case, but it does not cut at all in the IPO's favor because the largest external investor in OpenAI just told a federal judge that the governance process at the company he is funding is functionally amateur.
Which brings us to the part that institutional allocators are perhaps losing the most sleep over. A court document entered into evidence by Musk's attorney, Steven Molo, revealed that Sam Altman personally holds stakes worth over $2 billion across companies doing business with OpenAI, valued as of the 31st of December 2025. There's Helion Energy where Altman held roughly a one-third stake worth around $1.65 billion and chaired the board while OpenAI signed a power purchase agreement to buy fusion power from Helion. There's Stripe where his stake is worth approximately $632 million while OpenAI uses Stripe for payments. Retro Biosciences where his stake is roughly $258 million is also a business partner. Then there's Cerebras Systems where the structure is even more remarkable. OpenAI's compute deal with Cerebras includes warrants for up to 10% of the chipmaker, creating a circular arrangement where OpenAI's purchasing decisions directly enrich Altman's portfolio company. Altman also has stakes in Reddit, Stoke.ai, and in his own Worldcoin project.
Now, in 2023, Altman testified before the United States Senate that he had no equity in OpenAI. "I have no equity in OpenAI." But trial evidence has now established that he holds an indirect stake via Y Combinator. And on the 8th of May, Republican Congressman James Comer, chair of the House Oversight Committee, sent a formal letter demanding all documents related to the internal audit committee that OpenAI established in 2023 to review exactly these conflicts. Altman's deadline to respond is the 22nd of May. Furthermore, six Republican state attorneys general have separately written to the SEC requesting a formal review of OpenAI's governance ahead of the IPO.
So, let's layer this together. A sworn record of concerns from the former CTO about her boss. A 52-page dossier from the co-founder. A diary from the company president admitting they weren't honest with the original donor. A characterization of governance as amateur from the largest external investor. And $2 billion in personal side investments from the CEO across companies that OpenAI does business with, while a congressional committee actively investigates whether nonprofit capital was diverted to inflate his personal portfolio. Now, any one of those would be a meaningful S-1 disclosure. All of them together though is the kind of governance file that pension funds, sovereign wealth funds, and university endowments specifically have written policies to avoid.
And while every camera in financial media has been pointed at the Oakland courtroom, the actual power consolidation in AI has been happening somewhere else entirely. On the 2nd of February, SpaceX completed an all-stock acquisition of XAI at a combined valuation of approximately $1.25 trillion. XAI no longer exists as an independent entity. It's been absorbed into SpaceX as a division now branded SpaceX. Then on the 6th of May, Anthropic signed a multi-year agreement to lease the entire 220,000 GPU capacity of Colossus 1, the SpaceX-controlled data center in Memphis, Tennessee. That is 300 megawatts of power, a mix of Nvidia H100, H200, and Blackwell GB200 accelerators. Analyst estimates from New Street Research put the annual revenue to SpaceX between $3 and $6 billion.
Now, the reason SpaceX was willing to lease it out is that Colossus 1 was running at only 11% utilization because XAI had migrated all its training workloads to the newer Colossus 2 cluster, which now houses approximately 550,000 Blackwell accelerators. So if we map who actually owns the rails right now, the SpaceX AI cluster controls Grok Colossus 2 for training and is now monetizing Colossus 1 to a competitor. The OpenAI cluster runs on Microsoft Azure, has committed to roughly $600 billion in compute spending over 5 years, and is increasingly tied into Oracle, CoreWeave, and the Stargate joint venture. Meanwhile, the Anthropic cluster has assembled a multi-provider compute fortress: AWS up to 5 GWatt, a Google Broadcom 5 GWatt deal targeted for 2027, $30 billion in Azure capacity, and now Colossus 1 from SpaceX. Anthropic's annualized revenue has reportedly reached around $30 billion as of April, up from roughly $1 to $3 billion a year prior, with some estimates running as high as $44 billion. And its implied secondary market valuation is now running at approximately $1 trillion, ahead of OpenAI's secondary valuation of roughly $880 billion on private market sentiment.
And here is the part the headlines have not fully digested. Elon Musk is simultaneously suing OpenAI for $150 billion while leasing his most powerful idle data center to OpenAI's biggest competitor. Make no mistake, this is a calculated power move. So, OpenAI is no longer the default winner of the AI race, and Sam Altman's institutional stewardship is increasingly viewed as a liability rather than a value-add. The whole AI race is no longer a single horse pulling clear of the field. And the rumored IPO figure can no longer be a clean function of revenue growth divorced from governance disclosures, congressional subpoenas, and federal court rulings. What replaces all of that is a market structure where compute, talent, and capital are consolidating into two or three overlapping clusters that lease infrastructure to each other while litigating each other simultaneously. Where the founder myth at OpenAI has been documented under oath as something significantly closer to a governance crisis. And where the public market vehicle being prepared for retail investors must now disclose all of it.
So here are the three IPO scenarios as we see them. Scenario one is price through. The judge in the Musk case issues limited remedies. The testimony gets disclosed but absorbed. The IPO completes in the fourth quarter of 2026 somewhere between $800 billion and $1 trillion, and S&P 500 inclusion forces every passive index fund and 401k holder to own it whether they choose to or not. The retail buyer arrives at the top of a valuation that early employees were already cashing out of at $400 billion in October 2025, when 600 employees took $6.6 billion off the table.
Scenario two is repriced lower. The judge orders meaningful discordment to the nonprofit. The House Oversight investigation expands. The SEC requires deeper governance disclosures, and institutional allocators demand a discount. The IPO clears at perhaps $500 to $700 billion, materially below the $1 trillion many are calling for.
Scenario three is delayed into 2027. The structural remedies are heavy. The SEC will not clear the S-1 until governance issues are resolved, and CFO Sarah Friar's already public concerns about organizational readiness become the official reason for the delay. Pitchbook analysts are already calling the fourth quarter 2026 target overly ambitious.
Now, the verdict comes the week of the 18th of May. But that verdict itself almost doesn't matter. The sworn testimonies from Murati, Sutskever, Brockman, and Nadella are now permanent public record. Any S-1 filing has to address this. Any institutional allocator has to price it. And any retail buyer who ends up owning OpenAI through an index fund inherits all of it without ever consciously choosing to. The smart money has been positioned in this story since before the trial began. But for those eyeing up the IPO itself as a potential investment, they can now see that OpenAI has some pretty serious inner dysfunction, some worrying concerns around its CEO. And oh yes, its biggest rival just secured a major piece of AI infrastructure. The chips are down, and the stakes are very, very high.
But what do you think? Do all these embarrassing headlines mean OpenAI is overpriced? Or is AI itself so hot that most potential investors just won't care? And is Anthropic now the name to watch as the race to dominate the sector intensifies? Let us know your thoughts down in the comments. And if you want to understand how the broader circular financing model between hyperscalers and AI labs is propping up valuations across the entire sector, then you can check out our video on that right over here. That's all from me for today though. So, thank you as always for watching, and I'll see you again soon. This is Guy signing off.