Musk Loses to OpenAI as Meta Shuffles Bodies Toward the Furnace

The Grudge That Ran Out the Clock

Elon Musk sued OpenAI for $150 billion, claiming Sam Altman converted a charity into his personal rocket sled. A nine-person federal jury, after three weeks of testimony that illuminated very little, decided he should have brought the complaint years earlier. The statute of limitations did what Altman's lawyers could not, which is end the fight without ever actually settling whether the underlying accusation was true.

This is the most important non-finding of the year. Musk lost, OpenAI was cleared, and the question of whether a nonprofit was effectively laundered into the most valuable private company on earth remains formally unanswered. The jury was asked a procedural question and gave a procedural answer. The press, predictably, is calling it vindication.

Functionally, it is. The last credible legal brake on OpenAI's expansion just snapped. Altman walks out of court with a free hand to continue the for-profit conversion, the chip deals, the data center buildout, and the trillion-dollar capex hallucinations he has been narrating to anyone with a checkbook. Protesters outside the courthouse were noted and ignored. The juggernaut, as the Times put it without irony, accelerates.

Musk now returns to building his own version of the thing he claims to fear, which is the dramatic structure he prefers anyway.

Meta's Two-Day Pirouette

While the OpenAI drama sucked up the cameras, Meta executed one of the more honest pieces of corporate choreography in recent memory. On a Monday, the company reassigned roughly 7,000 employees into AI-focused roles. On a Wednesday, it announced layoffs of approximately 8,000, or about ten percent of the workforce.

The sequencing is the message. You are not being laid off because AI replaced you. You are being laid off because you were not reassigned to AI in time. Mark Zuckerberg has stopped pretending these are separate conversations. The labor budget is being rotated, not cut, and the people who survive are the ones standing near a GPU when the music stops.

The rest of the industry will study the playbook. Reassign first, fire second, and the narrative writes itself. No one has to say the word automation out loud.

The Grid Becomes the Story

The quietest item in today's traffic is the loudest in five years. NextEra and Dominion are reportedly exploring a $420 billion merger that would place a significant share of US data center power generation and transmission under one corporate entity. The hyperscalers have spent the past eighteen months signing power purchase agreements like teenagers signing yearbooks. Utilities noticed.

If this deal clears regulators, and that is a serious if, the chokepoint of American AI moves from chips to kilowatt-hours owned by a single counterparty. Every training run, every inference call, every Altman press release about gigawatt campuses runs through a pipe that someone else now prices. The compute oligarchy was supposed to be the story. The electricity oligarchy is the sequel, and it has better leverage.

Regulators will posture. The deal will probably get carved up. But the direction is set. Power is the scarce input, and the people who own power have figured out what their customers are worth.

Tenstorrent and the Nvidia Tax

Intel and Qualcomm are circling Tenstorrent, the Jim Keller-led chip startup that represents roughly the only credible attempt to build training silicon outside the Nvidia-AMD duopoly. A takeover would give either acquirer a fighting chance in a market where Nvidia currently extracts gross margins that resemble a software company's.

The interesting question is not whether Tenstorrent gets bought. It is whether being bought ruins it. Keller's pitch has always depended on open architecture and not being captured by a single platform owner. Intel needs a hit badly enough to promise anything. Qualcomm needs an answer for the data center it has never managed to enter. Whoever wins inherits a credible roadmap and the institutional gravity that has historically smothered credible roadmaps.

Nvidia, watching from a $4 trillion perch, is not nervous yet. It should at least be paying attention.

The Shape of the Day

A legal threat neutralized, a labor model normalized, a power grid consolidating, and a chip challenger going up for auction. None of it is catastrophic in isolation. All of it points the same direction, which is fewer hands on more levers, moving faster, with the last meaningful friction now removed from the largest player in the field.