What would you do with a company that grew annualized sales from a mid-single-digit-billion pace to more than sixty-five billion in a matter of months, then started talking about a public listing that could stretch toward two trillion dollars? I keep coming back to that question because the scale feels almost rude. It is not a cute startup story anymore. It is a market event that could crowd out other listings, pull attention away from slower growers, and force every public-market investor to decide whether enterprise artificial intelligence is a durable cash engine or a very expensive habit.
Why The November Window Suddenly Matters
People close to the process now point to November rather than an earlier autumn slot. That extra month is not just calendar fluff. It gives the company time to walk prospective buyers through third-quarter numbers before anyone has to put a firm price on the shares. In my experience, that kind of pause is less about nerves and more about presentation. You do not ask the market for a record-setting check while the latest quarter is still a rumor.
There is also the political calendar. Some voices have floated waiting until after U.S. midterm elections. Others insist the vote itself should not swing the listing. Both can be true at once. Markets hate surprises more than they hate politics. A quieter tape after the ballots are counted can still be useful even if the election is not the real reason for the delay.
No registration statement is public yet. Until that packet lands, the exchange, the ticker, the banks, and the exact share count remain unconfirmed. That vacuum is where rumor thrives. It is also where serious investors should stay calm and wait for audited language rather than hallway math.
The Size Of The Ask
Figures circulating among investors put the possible valuation near two trillion dollars, with proceeds that could reach one hundred billion. Those numbers would place the deal among the largest offerings on record. They are also still fluid. Valuation, dilution, and final proceeds can all move once roadshow demand is tested.
I have found that giant deals live or die on three boring questions. Can the company keep growing after the lockup stories fade? Can it fund the machines that produce that growth without constantly tapping the market? And can public holders live with a story that mixes commercial speed and safety caution? Miss one of those and a headline valuation becomes a trap.
A listing of this scale would force investors to treat frontier AI like a core holding, not a side bet.
Revenue That Refuses To Sit Still
Annualized revenue passed sixty-five billion dollars by the end of July, up from roughly nine billion at the close of the prior year. That figure is a run rate, not cash already banked across twelve tidy months. Still, the slope is hard to ignore. Some investors now expect the annualized total to climb above one hundred ten billion by year-end.
Most of that money comes from organizations, not weekend hobbyists. Subscriptions, programming interfaces, and multi-year business contracts sit at the center of the model. Teams use the systems for software work, research, support desks, and a pile of quieter internal tasks that never make a product announcement.
Perhaps the most interesting aspect is how quickly the mix can shift. One quarter the story is developer tools. The next quarter it is a handful of huge enterprise seats. Concentration risk hides inside that success. Public filings would have to say so in plain English.
How The Growth Case Gets Tested
Roadshow meetings will not be polite coffee chats. Buyers will poke at four pressure points.
- How fast sales can keep rising after the easy wave of first-time enterprise adoption.
- How much of each dollar is eaten by chips, power, and cloud partners.
- Whether a small set of customers accounts for too much of the book.
- What it costs to train the next model and keep older ones running without looking cheap.
Those are not gotcha questions. They are the difference between a growth stock and a science project with a ticker. I would rather see an honest slide on compute inflation than a glossy chart that pretends electricity is free.
Competition Is Already In The Room
Rival systems keep grabbing mindshare among business buyers. One competing model recently took a larger slice of tracked enterprise AI spend than the company’s own flagship line in at least one private data set. That does not end the story. It does mean the listing narrative cannot rest on a single brand name.
There has been talk of another model release timed against that pressure. Product cadence is both a commercial weapon and a safety headache. Release too slowly and customers wander. Release too fast and you inherit every argument about untested systems. Public investors will have to live with that tension in quarterly letters, not just in conference-stage speeches.
Even with that fight, the company’s July run rate was still described as higher than a major rival’s reported forty-billion pace. Long-range internal views have mentioned something near one hundred ninety to two hundred billion in 2028 revenue. Forecasts that far out are not promises. They are maps drawn in pencil.
The Power Bill Behind The Product
Serving this demand takes metal, land, and megawatts. Investors cited in market conversations expect access to about five gigawatts of computing capacity by the end of 2026, then close to twice that a year later. Read that again. This is industrial scale dressed up as software.
Expansion of that size raises capital questions. Who pays for the halls of servers? How much sits on the company’s balance sheet? How much is wrapped into cloud contracts that can look like either cost or revenue depending on the fine print? A public filing would have to unpack those commitments. Private-market storytelling can wave at them. Public-market storytelling cannot.
I’ve found that investors often underprice the electricity story until a hot summer or a delayed substation makes it real. Chips get the headlines. Power and cooling keep the lights on. If the listing happens, that unglamorous pair should sit near the front of the risk section.
Safety Talk Meets Shareholder Math
The chief executive has kept calling for tighter controls on advanced systems even while the company prepares to sell stock. He has argued that developers should slow the release of ever more capable models until governments and firms strengthen safeguards. That stance is principled. It is also commercially awkward.
Slower launches can cut near-term growth. Faster launches can raise the chance of a messy incident. Prospective holders have to decide which pain they prefer. There is no tidy slide that removes the choice.
Safety policy is not a side essay. In a listed company it becomes a financial variable.
The firm also joined a multi-year commitment of at least two billion dollars with a major consulting group to fund independent evaluation of frontier models. Outside teams would run tests, adversarial reviews, and alignment work, with access close to what internal staff already have. The idea is simple. Distant reviewers miss things. Close reviewers still need independence.
If shares go public, that program stops being a press note and starts being an operating cost with a governance story attached. Investors will ask who picks the evaluators, who publishes the ugly findings, and what happens when a test result collides with a sales target. Those are fair questions. They should be asked early.
What A U.S. Listing Would Actually Change
A domestic offering would give ordinary and institutional buyers a packet they do not get while the company stays private. A registration statement typically covers the business model, audited financials, risk factors, management, large holders, and use of proceeds. After trading starts, periodic reports keep that pipeline open.
One public-company chief who recently took a payments firm through a New York listing argued that markets impose audited accounts, regular reporting, independent board oversight, and a thicker layer of accountability. He also drew a useful line. Securities rules are not a substitute for government rules on advanced systems. They answer different questions. One set asks whether investors were told the truth. The other asks whether society can live with the product.
That distinction matters. A ticker does not retire privacy law, cyber rules, or competition policy. It adds another stack: material-risk disclosure, quarterly numbers, and the unblinking gaze of people who mark a book every afternoon.
How Circle’s Path Colors The Conversation
The same executive pointed to his own firm’s June 2025 debut, which raised about 1.05 billion dollars after an upsized book. The comparison is imperfect. A stablecoin issuer and a frontier-model lab do not share the same cost curve. Still, the cultural point lands. Once you are listed, partners and institutions can judge you with familiar tools. Some buyers will not write a large check until those tools exist.
He also said markets look jittery, the calendar looks crowded, and the valuation chatter is loud. Fair enough. Crowded calendars punish companies that cannot explain themselves in one sitting. A two-trillion conversation has to be almost boringly specific or it will sound like folklore.
What Public Investors Should Demand In The Fine Print
If this deal is real, the first filing should not read like a keynote. It should read like a contract with strangers. Here is the checklist I would keep on the desk.
- A clean split between subscription, interface, and large contract revenue, with concentration called out by name or by band.
- A compute map that shows owned capacity, leased capacity, and take-or-pay cloud terms.
- Training spend versus inference spend, because those two beasts do not move together.
- The policy for delaying or withholding a model, and who inside the firm can force that delay.
- Related-party and strategic-investor rights that could shape the board after the opening bell.
- A use-of-proceeds line that admits how much of the raise is really for power, land, and chips.
- Litigation, export-control, and safety-incident language that is specific enough to be useful.
Miss those and the market will invent its own answers. Invented answers are rarely kind.
A Simple Scorecard For The Valuation Debate
| Investor Question | Bull Case | Bear Case |
| Revenue durability | Enterprise seats keep expanding across workflows | Budgets freeze after the first wave of pilots |
| Margin path | Scale and better utilization lift contribution | Power and chip costs stay sticky |
| Competition | Brand and safety trust win long contracts | Share shifts with every rival release |
| Policy risk | Clear rules reduce accident risk | Release limits cap near-term sales |
| Capital need | Listing cash funds capacity on better terms | Follow-on raises become a habit |
None of those rows is a morality play. They are just the places where a two-trillion story can bend.
Market Timing Is A Personality Test
Some buyers want the company to wait for a calmer tape. Others want the liquidity now, while the growth print still looks violent. I tend to side with process over folklore. If third-quarter numbers are the missing piece, November is a reason, not an excuse. If the real issue is price discovery at a shocking multiple, extra weeks will not invent a new buyer universe. They will only give existing buyers more time to argue.
A crowded new-issue calendar is real. So is fatigue with stories that need a hundred-page footnote. The firms that get through that weather usually do one thing well. They pick three numbers and defend them until the room is bored. Everything else is decoration.
The Human Texture Behind The Machines
It is easy to talk about gigawatts and run rates and forget the messy middle. Sales teams are selling a product that changes every few months. Safety staff are asking engineers to wait. Finance staff are modeling cash needs that look like utility projects. Legal staff are drafting risk language for events that have not happened yet. That is a lot of friction inside one building.
Public markets do not remove the friction. They put a clock on it. Every quarter someone has to explain why the friction was worth it. That discipline can be healthy. It can also push a lab to sound more certain than the science allows. Watch the tone of the first few letters. Certainty theater is a tell.
What This Means If You Already Own Tech
A deal this large would not live in a vacuum. It would compete for specialist analyst time, index inclusion chatter, and the finite patience of growth funds. Other software names could look small by comparison, even if their cash conversion is cleaner. That relative-value squeeze is easy to miss when everyone is staring at one logo.
It would also reset the conversation about what “expensive” means. A company growing this fast can look cheap on next year’s sales and insane on last year’s. Both statements can be true before breakfast. The useful question is whether the customer is renting a tool or rebuilding a workflow. Tools churn. Workflows linger.
The Disclosure Dividend
For U.S. holders, the quiet prize is not the opening print. It is the paper trail. Audited statements. Segment clues. Risk pages that name compute suppliers and policy fights. Board structure that can be compared with other large issuers. Private markets can be brilliant and still leave you guessing. Public markets are ruder. Rudeness has value.
That said, disclosure is not regulation of the models themselves. Anyone who treats a 10-K as a safety charter is reading the wrong book. The filing tells you what managers fear. It does not certify that the fear was priced correctly.
A Few Practical Scenarios
Think in branches, not slogans.
- Clean November print: third-quarter numbers land, demand is deep, the raise is large, and the first-day tape is orderly.
- Stretched timetable: the company keeps talking, files later, and uses the extra time to settle a valuation fight.
- Partial raise: the headline valuation slips, proceeds come in below the loudest rumors, and the firm still gets a liquid currency.
- Stay private a while longer: market windows slam, strategic holders fill the gap, and the public story waits for a calmer year.
I would not marry any single branch. The only lazy take is pretending the outcome is already written.
Where I Keep Getting Stuck
Two thoughts will not leave me alone. First, the revenue slope is real enough that dismissing it as hype feels sloppy. Companies do not stumble into a sixty-five-billion run rate by accident, even if the measure is annualized rather than trailing. Second, the capital intensity is also real enough that treating this like a classic high-margin software name feels sloppy in the other direction.
So the honest stance is a bit uncomfortable. This can be both a generational franchise and a balance-sheet project. Public investors who need it to be only one of those things will have a long year.
Working frame I keep on a notepad: Growth is the magnet Compute is the tax Safety is the governor Disclosure is the price of the ticket
How To Read The Next Few Weeks
Watch for three tells. Detailed financials shared with prospective holders. Language about capacity that uses numbers instead of adjectives. And any shift in how leadership talks about release speed. If those three move together, the November window is more than a rumor. If they stay foggy, the valuation conversation is still theater.
Also watch the supporting cast. Banks do not staff a deal of this size for fun. Strategic holders do not stay quiet if they fear dilution. Enterprise buyers do not sign multi-year paper if they think the model roadmap is about to stall. Those side plots often leak the truth before the prospectus does.
A Note On Hype Versus Homework
Giant offerings attract a carnival. Social feeds will flatten the story into a single number. Resist that. Two trillion is a conversation starter, not an investment thesis. The thesis lives in unit costs, contract length, customer concentration, and the willingness to delay a flashy launch when a test looks ugly.
I would rather own a slightly slower compounder that can explain its power bill than a legend that treats electricity as a rounding error. That bias will not fit every portfolio. It fits mine.
The Accountability Argument, Without The Sermon
There is a decent case that public markets make powerful labs easier to inspect. Auditors show up. Independent directors acquire a paper trail. Material events have to be named. Critics of listing worry that quarterly pressure will punish caution. Both worries can be true. The job is to design a board and a reporting cadence that do not turn safety into a rounding item on an earnings call.
That design work is unglamorous. It is also the part that will matter five years after the opening print is forgotten. Tickers age. Governance either holds or it does not.
Closing The Loop Without Pretending We Know The Price
So where does that leave a reader who is not sitting in the roadshow room? Treat the November talk as a live process, not a done deal. Treat the two-trillion figure as a bid-ask argument, not a fact of nature. Treat the revenue spike as evidence of demand, not proof that margins will behave. And treat the safety stance as part of the product, not a charity side quest.
If the company files, read the risk pages before the glossy summary. If it waits, ask whether the wait was about numbers, about politics, or about price. Those are different stories. They deserve different reactions.
One last thought, and then I will let the kettle boil. Markets can digest a huge listing when the issuer is willing to be specific. They punish vagueness at any size. Specificity is the real scarce asset here. Revenue can be large. Capacity can be vast. The offering can still stumble if the story stays abstract. November will tell us whether the company is ready to trade poetry for line items. That, more than the headline multiple, is the test I want to see.
