ELON RINGS THE NASDAQ BELL WITH A $1.77 TRILLION COMPANY THAT ALSO INCLUDES AN AI THAT IS BURNING MONEY AT BIBLICAL SPEED
Today is one of those days you will remember where you were. SpaceX, the company that built rockets in a field in Texas and told everyone they were wrong, is going public this morning on the Nasdaq at a valuation of $1.77 trillion. That is not a typo. One point seven seven trillion dollars. For context, that is more than the GDP of most countries on Earth, and it is for a company that, up until a few years ago, was still figuring out how to land boosters without blowing them up.
But here is the thing that makes this genuinely interesting to follow. SpaceX is not just a rocket company anymore. When Musk merged xAI into SpaceX back in February, he basically plugged Grok, his AI chatbot, and the entire xAI operation into a company that also owns Starlink, a global satellite internet network with more than ten thousand satellites. Now they are pitching something called Macrohard, which is an AI agent service described as an AI-run software company. The idea is that you point it at a coding problem and a team of AI agents handles the whole thing like a digital software house run entirely by robots. Which is either the future or a very expensive science experiment.
The numbers, though, are a bit uncomfortable. The xAI unit burned through $7.72 billion in just the first three months of 2026 and posted a $2.47 billion operating loss in the same period. So Musk is bringing to market, in the same package, the most iconic rocket company in history and an AI division that is currently lighting money on fire at a rate that would make most CFOs cry into their spreadsheets. The IPO is priced at $135 per share and the company plans to raise $75 billion from the offering.
Wall Street is nervous, not because SpaceX is a bad company, but because this thing is so enormous that if it tanks, it is going to take a lot of portfolios down with it. Morningstar came out last week and said SpaceX is worth less than half its $1.75 trillion target price. Several analysts are comparing this moment to the peak of the dot-com bubble. They are calling it the signal that the AI bull run may have officially run out of road.
But then again, the people who said SpaceX could not land a booster were wrong. The people who said Tesla would go bankrupt were wrong. The people who said nobody would pay for satellite internet were wrong. Musk has a documented history of making his skeptics look stupid over a long enough time horizon. Whether that applies to a $1.77 trillion AI-rocket hybrid listed during what might be the frothiest moment in technology market history, that is the bet you are making today if you buy SPCX. It is either the trade of a generation or the kind of decision people tell cautionary tales about at future business school seminars. Probably both, depending on when you got in.
What nobody is saying loudly enough is that we now have SpaceX, OpenAI, and Anthropic all heading toward public markets at roughly the same time, all at valuations that would have seemed fictional two years ago. The sheer volume of capital chasing AI infrastructure is unlike anything in financial history. At some point the music stops. The question is whether you are still dancing when it does.
Source: SpaceX IPO won’t break the bull market, but investors are worried about what comes next — CNBC
BEZOS RAISES $12 BILLION TO BUILD AN AI THAT COMPRESSES A 10-YEAR ENGINEERING PROGRAM INTO ONE
Jeff Bezos has been quiet for a while. Since stepping back from Amazon and briefly going to space in a rocket that everyone made fun of, he has been doing what all serious billionaires eventually do: investing in things that sound insane and hoping to be right before anyone notices. Well, yesterday he stopped being quiet. Bezos, along with Vik Bajaj, a former Google executive, announced that Prometheus, their industrial AI startup, just raised $12 billion in Series B funding at a $41 billion valuation. The company has now raised over $18 billion total.
And look, I know what you are thinking. Another AI company, another absurd valuation, another press release that sounds vaguely like science fiction. But give this one a minute because it is actually trying to do something meaningfully different from the usual “we built a chatbot that writes your emails” pitch.
The idea behind Prometheus is this: physical manufacturing is brutally, painfully slow. If you walk into a jet engine company today and say you want the exact same engine but with ten percent more thrust, they will tell you that is a decade-long program. Not because they are lazy or incompetent, but because the complexity of modern industrial design is genuinely staggering. Bezos wants to build what he calls an “artificial general engineer,” a system that can take an engineer’s dream and compress that build cycle by a factor of ten or more. Instead of a ten-year program for a better jet engine, maybe one year. Maybe faster.
The company, which just dropped the “Project” from its name and currently employs around 150 people, has backing from JPMorgan, BlackRock, Goldman Sachs, DST Global, and Arch Venture Partners. Bezos himself was the largest backer in the Series A and participated again in the Series B, which is not something many founders do and signals he is genuinely committed rather than just lending his name to the letterhead.
What they are not telling you is also interesting. They declined to discuss reports of an affiliated $100 billion holding company that would acquire legacy industrial manufacturers and funnel their data into Prometheus. They also would not say how the AI is being trained, which is a notable gap given that, as they freely admit, there is no “Internet of manufacturing data” to scrape. Unlike the text-and-image data that trained language models, real industrial knowledge lives inside company servers, engineering teams, and decades of institutional memory that nobody has bothered to digitize. Getting it requires actual partnerships with actual manufacturers who have been guarding it jealously for years.
There is also the philosophical tension built into the whole pitch. Bezos and Bajaj insist this will create more engineering jobs, not fewer. Their argument is that when it becomes easier and faster to build things, more things get built, and more people get hired to build them. That might be true. It also might be exactly the kind of thing you say when you are standing in front of cameras after announcing a $12 billion bet on replacing human engineering labor with software. We have heard similar arguments before. They are sometimes right. Time will tell which version of this story holds up.
Source: Prometheus, the industrial AI startup from Jeff Bezos, is now worth $41 billion — Axios
NVIDIA’S JENSEN HUANG DECLINES TO TESTIFY BEFORE THE US SENATE AND OFFERS SENATOR WARREN A TOUR OF HIS HEADQUARTERS INSTEAD
Last week, Senator Elizabeth Warren sent Nvidia CEO Jensen Huang a formal invitation to appear before the Senate Banking Committee and answer some questions about Nvidia’s chip sales to China, the state of US export controls, and how one semiconductor company came to sit at the center of the most important technology competition in modern history. This week, Jensen Huang sent back a polite no.
His official response was that he is “unable to attend” but would be happy to host Warren and other committee members at Nvidia’s campus in Santa Clara. Which is either a genuinely gracious gesture from a busy executive or the most elegantly worded version of “I am not going to sit in a chair and let a senator yell at me for two hours on C-SPAN” that the English language has ever produced. Probably a bit of both.
Warren was not amused. She responded publicly that “the American people deserve answers in a public forum,” which is the kind of statement senators make when they are frustrated and out of options. She is right, actually. Whether or not you agree with her politics, the underlying questions are legitimate ones. Nvidia chips are showing up in Chinese AI systems and, according to various investigative reports, in applications with military adjacency. Congress has been trying to tighten export controls, and Nvidia has been walking a careful line between complying with the rules and finding every legal path to keep selling chips to one of the world’s largest markets.
Here is what makes this situation genuinely remarkable. Nvidia is currently worth about $3 trillion. The company went from being a niche gaming hardware company to the backbone of the entire global AI economy in roughly five years. If you are OpenAI, Anthropic, Google, Meta, Microsoft, or any of the dozens of AI companies currently spending billions of dollars on compute, you are buying Nvidia GPUs. There is essentially no alternative at scale. That gives Jensen Huang a kind of quiet power that is unusual for any single executive in any industry.
And when you have that kind of power, declining a Senate invitation is a statement. Not just a scheduling conflict. It is a signal that Nvidia does not feel particularly obligated to answer to political institutions right now because, practically speaking, nobody can stop them. Congress cannot build a competing chip company in a weekend. The entire AI buildout runs through one man and his products, and everyone in Washington knows it.
Warren’s frustration is understandable. You have the most strategically important hardware company in the world selling cutting-edge technology to a geopolitical rival, the CEO declines to explain himself in public, and the administration has been photographed warmly with that same CEO multiple times. It is a storyline that is not going away and will get louder before it gets quieter.
Source: Nvidia CEO Jensen Huang declines Senate testimony on AI, China and exports — CNBC
THE BANKS ARE QUIETLY DRAWING UP THE LISTS: AI IS COMING FOR FINANCE JOBS AND THE EXECUTIVES ARE BARELY PRETENDING OTHERWISE
There is a specific kind of corporate communication that happens when a company is about to do something unpleasant but does not want to say so directly. You hear words like “efficiency,” “transformation,” “strategic alignment,” and my personal favorite, “rationalizing the workforce.” The banking industry is currently producing this type of communication at an impressive volume, and Bloomberg has been paying close attention.
A report out this week from Bloomberg details what has been happening quietly inside major financial institutions around the world: the banks are laying the groundwork for mass workforce reductions tied directly to AI adoption. This is not shocking if you have been watching the space, but the scale and frankness of what is being planned is worth stopping to take seriously.
Morgan Stanley published analysis estimating that European banks alone could cut up to twenty percent of their headcount as AI enables productivity improvements of around thirty percent. Standard Chartered has already announced plans to eliminate eight thousand support roles over the next four years. An executive at the bank described the plan as a shift from “lower-value human capital” to “technology capital.” That phrase, “lower-value human capital,” is doing an enormous amount of heavy lifting in what is, stripped of the jargon, a sentence about firing people and labeling them as less valuable than a software subscription. Someone in that PR team had an interesting morning when that quote went public.
The roles most at risk are the ones you would expect: back-office processing, compliance checking, trade reconciliation, fraud screening, customer service, and anything that involves taking in structured data, applying rules, and producing a consistent output. Banks have had humans doing these jobs for decades because computers were not quite good enough and humans were cheap enough. AI is now good enough, and the math has changed.
What makes this particular moment sharp is the feedback loop it creates. Candidates applying for banking jobs are now running into AI screening systems before they ever speak to a human recruiter. The hiring process for many roles involves being evaluated by the same category of technology that, if you actually get the job, will spend the next several years automating parts of it. You are being screened by your future replacement on the way in the door.
There are real people on the other side of this calculation. Bank employees in operations, compliance, junior trading, and administrative functions are watching their job descriptions narrow in real time, through news articles and internal briefings that often arrive simultaneously and say very different things. The broader banking industry employs tens of millions of people globally. A twenty percent reduction across the sector is a number large enough to reshape labor markets in multiple countries at once. Nobody knows exactly how fast this unfolds or whether new roles absorb the losses. But the executives in the boardrooms are not waiting to find out. The lists are being drawn up now.
Source: Banks Lay Groundwork for Mass Workforce Cuts as AI Takes Hold — Bloomberg