May 22, 2026 · Money Circuit
THE CHATBOT IS FILING FOR IPO: OPENAI CONFIDENTIALLY SUBMITS PAPERWORK TO GO PUBLIC, GOLDMAN AND MORGAN STANLEY IN TOW
SOURCE: CNBC — May 20, 2026
OpenAI has filed a confidential draft prospectus with the SEC, kicking off the most anticipated and probably most insane initial public offering in the history of American finance. The company is working with Goldman Sachs and Morgan Stanley, which is the Wall Street equivalent of hiring the two biggest bouncers in town to manage the line outside a club where everyone already wants in.
Let us talk about what this actually means for a second. OpenAI was a nonprofit. Then it became a capped-profit. Then it restructured again in February. Now it is filing for an IPO at a valuation north of eight hundred and fifty billion dollars. In four years, Sam Altman took a research lab that was once worried about corporate influence on AI development and turned it into a company that is about to list on a stock exchange alongside Walmart and JPMorgan Chase. That is either the greatest pivot in startup history or the most elaborate joke ever played on Silicon Valley. Maybe both.
The timing is not accidental. Cerebras went public last week and popped sixty eight percent on day one. SpaceX dropped its IPO filing on Wednesday. The market is in full AI euphoria mode right now, and OpenAI’s bankers know you do not throw a party when everyone is sober. You wait until the room is already going, and then you walk in.
What makes this complicated is the financial picture. OpenAI is burning through cash at a rate that makes a Vegas bender look fiscally conservative. The company spent most of 2025 losing money faster than it raised it, even as revenue exploded past forty billion dollars annualized. The compute costs, the talent wars, the infrastructure buildout, the side deals with every large enterprise on earth. It all costs a fortune.
But here is the thing: none of that may matter. We are past the era where IPO investors want to see a path to profitability. They want to see dominance. They want to see market share. They want to see the thing that everyone is talking about, packaged into a ticker symbol they can tell their brother-in-law about at Thanksgiving. OpenAI is all of that and more.
The real question the prospectus will have to answer is this: what does OpenAI actually own that Anthropic, Google, Meta, or a Chinese lab cannot build or copy in eighteen months? The moat is real right now. Whether it holds through a public market cycle where every single number gets picked apart by short sellers and hedge funds is an entirely different matter. This is going to be one of the most watched IPO filings in years. Every word in that prospectus will be treated like scripture by people who have not read an S-1 before in their lives.
Either way, history is being made. The company that essentially started the modern AI era is going public. Grab some popcorn.
ELON MUSK TELLS WALL STREET SPACEX IS AN AI COMPANY NOW: IPO FILING REVEALS $20 BILLION IN AI SPENDING, DATA CENTERS IN SPACE BY 2028
SOURCE: TechCrunch — May 20, 2026 | Bloomberg — May 21, 2026
SpaceX filed its IPO prospectus this week and the document is less of a rocket company filing and more of a very long argument that Elon Musk figured out the future before everyone else did. The headline number is a valuation that could top two trillion dollars, which would make SpaceX one of the five most valuable companies on earth on its first day of trading. But the more interesting number is buried in the financials: SpaceX directed approximately sixty percent of its capital expenditure in 2025 to its AI division. That is about twenty billion dollars. On AI. From the company that makes rockets.
Here is how this happened. SpaceX absorbed xAI, Elon Musk’s AI lab, earlier this year. xAI burned six point four billion dollars from operations in 2025 on just three point two billion in revenue. The filing discloses these losses with the same tone a doctor uses when telling you to cut back on salt. Casual. Almost bored. Yes, the AI division is bleeding billions. Moving on.
The total addressable market SpaceX claims in the filing is twenty six and a half trillion dollars. Of that, the company says ninety three percent of the opportunity comes from AI. Not from launching satellites. Not from sending cargo to the International Space Station. Not from eventually putting humans on Mars. AI. The company that is famous for landing rockets backwards, on drone ships, in the ocean, has looked at its business and decided the real play is artificial intelligence.
The plan to launch data centers into space by 2028 is either visionary or completely unhinged, and the frustrating thing is that both can be true at the same time when Elon Musk is involved. The basic logic is not stupid: if you own the launch infrastructure, you can put computing power in orbit at a cost no competitor can match. Low-latency AI processing from space sounds like science fiction until you remember that five years ago, landing a reusable rocket sounded like science fiction too.
The financial picture is challenging in ways that will give IPO analysts nightmares. SpaceX is profitable on its core launch and Starlink business. xAI is very much not profitable, in the way that a furnace is not cold. The question the filing does not really answer is: at what point does the AI spending start generating AI revenue at the scale required to justify a two trillion dollar valuation?
SpaceX’s answer, implied throughout the document, seems to be: trust us, we have done harder things. Which, honestly, is not the worst answer they could give. This is going to be a fascinating IPO to watch. The two most anticipated listings of 2026, OpenAI and SpaceX, have effectively filed within days of each other. Wall Street is going to need a bigger spreadsheet.
CEREBRAS GOES PUBLIC AND INSTANTLY BECOMES A $95 BILLION COMPANY: THE AI CHIP IPO NOBODY SAW COMING JUST CHANGED EVERYTHING
SOURCE: CNBC — May 14, 2026 | TechCrunch — May 14, 2026
Cerebras Systems walked onto the Nasdaq last week, raised five and a half billion dollars in the largest US tech IPO since Uber went public in 2019, and watched its stock pop sixty eight percent on opening day. The company’s market cap briefly crested one hundred billion dollars before settling back to ninety five billion at close. CEO Andrew Feldman and CTO Sean Lie are now billionaires. Not paper billionaires. Real ones. The kind where you stop checking flight prices.
If you are not familiar with Cerebras, here is the short version: they build really, really big chips. The company’s main product is essentially a wafer-scale chip that is the size of a dinner plate. While Nvidia’s A100 and H100 GPUs are tiny by comparison, Cerebras decided early on that the way to win was to go big and go weird. The chip is so large that it contains over four trillion transistors and eliminates most of the communication bottlenecks that slow down traditional multi-chip AI training setups.
This approach was considered eccentric for years. Then the world decided it needed to train enormous language models at speeds that traditional chip architectures could not match, and suddenly the company that was making dinner-plate-sized chips did not look eccentric anymore. It looked prescient.
The sixty eight percent first-day pop is worth dwelling on because it tells you something important about where institutional investor appetite is right now. This is not retail mania. This was a priced IPO with Goldman Sachs and JPMorgan in the syndicate. Sophisticated money looked at Cerebras, priced it at a range, and then the market said: no, this is worth dramatically more than that. The company bumped its IPO range multiple times in the week before listing and still the stock flew.
Part of what drove the enthusiasm is Cerebras’s customer list. The company has signed deals with Amazon and OpenAI this year, diversifying away from its original customer base of government and research labs. When the two biggest names in AI cloud services decide to buy your chips, the market interprets that as a fairly strong signal about the quality of your product.
The interesting implication for the broader AI market is what this IPO says about timing. Cerebras was supposed to go public much earlier and got stuck in a lengthy national security review tied to its largest customer, G42 in Abu Dhabi. The fact that it cleared those hurdles and still executed one of the best IPO performances in years suggests that the window for AI hardware companies to go public is wide open right now. Expect more. This was the starter pistol, not the finish line.
$700 MILLION FOR A PRODUCT NOBODY HAS SEEN: BRETT ADCOCK’S STEALTH AI STARTUP HARK RAISES A SERIES A BIGGER THAN MOST COMPANIES EVER TOUCH
SOURCE: TechCrunch — May 21, 2026 | Bloomberg — May 21, 2026
Brett Adcock raised seven hundred million dollars this week for a company called Hark, which is building what it describes as a “universal AI interface.” The company is valued at six billion dollars. The product does not exist yet in any publicly available form. The round is called a Series A. Seven hundred million dollars. Series A.
To understand why this is funny and also completely logical, you need to know a little about Brett Adcock. He started a company called Archer Aviation, the electric air taxi startup, and took it public via SPAC. He then started Figure AI, the humanoid robot company, which raised hundreds of millions and got a massive BMW production contract. Now he has started Hark, which he funded with a hundred million of his own money in late 2025 before going out to raise more. The man has a track record of raising large amounts of money for ambitious hardware projects and actually building the things. That is extremely rare. Most people who raise this kind of capital at this stage are pitching a vision. Adcock has a body of work.
The Hark investor list reads like someone ran a search for “companies that know things about the future.” Nvidia put money in. So did AMD, Qualcomm, Intel Capital, Salesforce Ventures, and Brookfield. When five separate chip and enterprise technology giants all agree to write checks into the same stealth startup, that is not coincidence. That is a coordinated bet on something they have seen in private that the rest of the market has not.
The description Adcock has used publicly is that Hark is building an agentic AI system that will serve as a universal interface with the digital world. The company recruited Abidur Chowdhury, the former Apple designer behind the iPhone Air, to run design. Whatever this product is, it is going to look extremely good while doing extremely unclear things.
Here is the thing about funding rounds at this scale for companies at this stage: they are not just bets on a product. They are bets on a person. The investors in Hark are essentially saying, in writing, with nine-figure checks: whatever Brett Adcock decides to build next, we want to own a piece of it. That is the kind of founder reputation that takes a decade and a series of genuinely difficult technical achievements to earn. Adcock has apparently earned it.
Seven hundred million for a Series A is not normal. Even in 2026, even in AI, even with a track record, this is a remarkable number. The company expects to release its first multimodal models this summer. Between now and then, it is one of the most expensive secrets in Silicon Valley.
ANDREESSEN HOROWITZ BETS $250 MILLION THAT GOOGLE SEARCH IS DEAD: EXA LABS RAISES AT $2.2 BILLION TO BUILD A SEARCH ENGINE FOR THE AI AGE
SOURCE: Bloomberg — May 20, 2026 | TechCrunch — May 20, 2026
Exa Labs just raised two hundred and fifty million dollars at a two point two billion dollar valuation in a round led by Andreessen Horowitz. The company is five years old. Six months ago it was valued at seven hundred million dollars. That is a three-times markup in half a year. In the AI investment environment of 2026, this is considered a modest gain.
What does Exa actually do? It builds a search engine, but not the kind you are thinking of. Traditional search engines like Google are built around matching keywords to pages and ranking those pages by a combination of relevance signals. That works reasonably well for finding information if you know what words to use. Exa is built differently. It does what it calls meaning-based retrieval, which means you can ask it a complex question and it will find documents that contain the answer rather than documents that contain your search terms. That distinction sounds small and is actually enormous.
The reason this matters specifically right now is AI agents. When an AI model needs to look something up in real time, it cannot just type queries into Google and hope for the best. It needs a search system that understands what it is actually looking for and can return clean, structured, accurate data. Exa has been building exactly that infrastructure since 2021. The timing has turned out to be excellent.
Google is aware that companies like Exa exist and are presumably not thrilled about it. The search giant has spent billions building its own AI search features into its core product, with mixed results. Every time Google’s AI Overview confidently tells you something incorrect, Exa’s pitch to enterprise customers gets a little easier.
The Andreessen Horowitz bet is interesting because a16z has been careful about where it puts money in the search space. This is a firm that understands the difference between a genuine technology shift and a demo that looks good in a slide deck. When they write a lead check of this size into a search startup, they are saying something specific: they believe the infrastructure layer of AI search is up for grabs, and Exa has a real shot at owning it.
The previous valuation of seven hundred million came from a round led by Lightspeed. Now a16z has led at three times that number. The signal from the venture community is clear: whoever builds the search layer that AI systems rely on will be worth an enormous amount of money. Exa thinks it is them. Two and two tenths billion dollars’ worth of investors agree.
ANTHROPIC POSTS $10.9 BILLION IN ONE QUARTER AND EYES A $900 BILLION VALUATION: CLAUDE’S MAKER IS ABOUT TO BECOME ONE OF THE MOST VALUABLE COMPANIES ON EARTH
SOURCE: CNBC — May 20, 2026
Anthropic is about to post its first profitable quarter. The company that makes Claude generated ten point nine billion dollars in revenue in Q2 2026. Annualized, that is roughly forty three billion dollars a year. The company is in talks to raise money at a nine hundred billion dollar valuation, which would make it the second most valuable startup in the world behind only SpaceX, and ahead of the current estimated value of every NBA franchise combined.
Let us put the number in context. Anthropic was founded in 2021 by Dario Amodei, his sister Daniela, and several other former OpenAI researchers who left because they were worried about AI safety. The company’s original pitch was essentially: someone needs to build frontier AI carefully, and that someone might as well be us. That pitch attracted billions from Amazon, Google, and a parade of sovereign wealth funds. It has now apparently attracted revenue at a pace that is hard to fully comprehend.
Ten point nine billion dollars in a single quarter means Anthropic is generating roughly a billion two hundred million dollars per week. For a company that did not exist five years ago, that is the kind of growth rate that makes venture capitalists go quiet for a moment and just stare at the ceiling. Amazon, which invested up to eight billion dollars in Anthropic in 2023 and 2024, is looking like one of the more prescient bets in the history of corporate venture investing. Google, which matched roughly the same amount, is feeling similarly clever.
The profitability angle is genuinely significant. The knock on all the frontier AI labs has been that they are essentially burning money at rates that no revenue model can sustain. OpenAI was losing money on ChatGPT usage for a long time. The compute costs for running these models at consumer scale are brutal. Anthropic appears to be proving that if you focus aggressively on the enterprise, charge real prices, and build models that businesses actually want to pay for and integrate into workflows, the economics work. Claude’s uptake in legal, medical, software development, and customer service contexts has been strong enough to actually move the needle.
The nine hundred billion dollar valuation figure is where things get genuinely surreal. Amazon’s market cap is approximately two point two trillion. Anthropic, a five-year-old startup, may be about to raise money at nearly half that number. If the IPO follows, and sources suggest it is coming within the year, the Amodei siblings are going to find themselves in a very strange position: the people who left OpenAI because they were worried about moving too fast may end up running one of the largest public companies in history.
The race between OpenAI and Anthropic to go public just got a lot more interesting. Both companies are now valued above eight hundred billion dollars on the private market. One of them is going to go first. Right now, the smart money says OpenAI gets there sooner. But Anthropic’s revenue growth suggests that if Dario Amodei decides to move quickly, the market will be ready to receive him.
Keywords: OpenAI IPO 2026, SpaceX IPO AI bets xAI, Cerebras IPO Nasdaq $95 billion, Hark AI Brett Adcock $700 million Series A, Exa Labs Andreessen Horowitz AI search, Anthropic $900 billion valuation Q2 revenue, AI startup funding May 2026, Money Circuit weekly AI finance