Pre-IPO options
$ANT: The Show Must Go On
In two weeks the market went from a coin flip on a $2 trillion Anthropic to single digits. The date recovered and the price did not. What Anthropic options, Polymarket and Nvidia's straddles say about the biggest IPO ever.
On Sunday afternoon, with every stock exchange on earth shut, Nvidia's board put out a statement saying it was investigating reports that its founder had been the target of a foreign intelligence effort. The whole AI boom runs through his chips. Nobody could sell the stock, so people sold the only thing that was open, and on Hyperliquid lost more than 6% before New York opened. By Monday morning, with The Information and CNBC reporting a roadshow set for this week, Anthropic was pressing ahead anyway with what would be the largest IPO ever.
Two weeks ago, the morning after it filed, Polymarket gave Anthropic of closing its first day worth more than $2 trillion. Today it's 9%. In between, semiconductors had their second-worst day since 1994, the lost a fifth of its value in two sessions, and early on Friday the market gave a listing by November 20 about a one-in-three chance. The date has come back to 70%. The $2 trillion hasn't.
We think this deal happens by the end of November, and at a price cheaper than anyone hoped a month ago: below $1.5 trillion at the IPO, and below $2 trillion when the first day ends. Options on Anthropic opened on Monday, and they show where the market still disagrees with itself. To see why, you need to know what's being priced.
The video and audio versions use an AI voice.
Background
Two numbers decide what this deal is worth, the share count and the price, and the S-1 gives you enough to estimate both before Anthropic prints them. Below is where we come out, and the order in which the market finds out whether we're right.
The share count is the one that matters most for reading the markets in this piece, because every dollar figure for the perp depends on it. We convert at the S-1's 16.9 billion fully diluted count throughout. The Series H mark implies about 16.4 billion, so treat that as the low end. On the post-deal count, the perp is already pricing a first-day close well above where Polymarket thinks it lands.
The amendment is the date to circle. Until it arrives, every market here is guessing at the same two blanks. Once it does, the perp and Polymarket's valuation ladder either close the gap between them or one of them is wrong by a lot.
The perp is the cleanest read, so start there.
The price came down in steps
Polymarket's closing market cap market, which has traded $4.2 million, shows the whole distribution, and the perp shows where the money that can actually trade it sits. Here they are together on four days.
In late October the single likeliest result was a first-day close between $1.75T and $2T. Today it's $1.5T to $1.75T, and the bucket under that has grown from 6% to 30%. The odds of $2T collapsed by about four fifths while the middle of the distribution only moved about 20%. That's how a tail gets repriced: the top falls away and the median just drifts.
The ANT perp says the same thing in dollars. It went from about $1.98T on Nov 4 to $1.63T on Friday, and it's back near $1.75T. That's still above the $1.25T to $1.75T the FT's sources were talking about, and a long way above the $1.2T to $1.3T Bloomberg had from buyers.
Each headline took a step off the price, and none of the steps came back. The filing moved the date and barely touched the price, because that price was already in. Thursday's crash hit the date. Friday's price talk hit the price.
The level fell. The date did something different.
The date wobbled and came back
Polymarket also runs a separate contract for each possible listing date, which makes a calendar you can price.
had been 68% on Oct 27. By 4am New York on Friday they were around 35%, with the crash and the valuation stories hitting at once. By Monday's close they were back at 70%, and by-Nov-30 at 85%. Almost nobody thinks this deal is going away. They're arguing about what it costs.
The test of that came on Sunday.
Nvidia's Sunday
The New York Times reported on Sunday that Nvidia had opened an investigation after Jensen Huang told the board about a relationship with a business executive. The espionage part is an allegation in the reporting, which attributes it to US intelligence. China has denied it, per a post from Kalshi. The board's own statement says it has "no reason to believe that either national security or NVIDIA secrets were compromised," and that independent directors with outside counsel are running the review. Huang stays CEO. House committee chairs have asked him for a closed-door briefing on Nov 18, the day after earnings.
The biggest AI headline of the year hit while the only place to trade it was Hyperliquid. If Anthropic were simply a derivative of the Nvidia trade, it should have gone down with him. It went up.
The S-1 gives you the reason. Anthropic trains and serves Claude across four chip families: AWS Trainium, Google TPUs, Nvidia GPUs and AMD GPUs. It has committed about $895.4 billion, nearly all of it on compute, including up to $120.5 billion of Nvidia-powered capacity through SpaceXAI. A company on the buy side of a supplier's bad weekend has a different problem from the supplier.
Nvidia's own options barely flinched. The stock lost about 5% by Monday's close, and its Nov 27 straddle went from 11.7% of the stock on Friday to 11.4%. The stock got repriced and the price of further surprise didn't.
This cuts against what we wrote on Friday. We said the same setup was forming around Anthropic and that the trade would all go together. On Monday it didn't. If the deal happens cheaper, the next question is what that does to day one.
A cheaper deal pops
Polymarket prices the IPO valuation and the first-day close as two separate bets.
The likeliest IPO price is $1.25T to $1.5T. The likeliest close is $1.5T to $1.75T, and Polymarket gives 84% to the stock closing above its IPO price. Those are the same trade. Bankers price a deal so it trades up, and a lower number gives them room.
Options opened on Monday. Do they agree?
Options still pay for $2T
Hypercall's ANT options went live on Monday, synthetic and cash-settled. We bill them as the first options market on the name, and we haven't found another listed venue, though that's our own claim. Indicative quotes are up on every strike from Nov 10 to Nov 27, all from one provider, so this is one firm's opinion priced in public.
Implied volatility tells you how much movement an option has priced in. Line up the expiries and you see when.
Listing week runs at about 116% through Nov 20, and on its own that week is about 129%, a one-standard-deviation move of about 18%. The week after is priced at about 15%. We said on Friday the risk was the day after and the market would price it like nothing. This curve says otherwise, so we got that wrong, at least so far.
Option prices can also be turned into odds. A call spread's cost, divided by the gap between its strikes, is roughly the chance of finishing above.
At the mid, options give about 26% to ANT finishing above the $2T price on Nov 27. Polymarket gives 9% to a $2T close. At $1.75T it's 48% against 25%. But the quotes are wide: going into Monday's close the bid on that 10-point spread was 0% and the ask about 56%. The dates differ, and options carry what people pay to insure the tail. You can't pick this off. What you can say is that the perp, and the options built on it, are the most expensive way of looking at Anthropic, and that gap closes when the amendment lands.
That disagreement is tradable.
Trades, our call, what to watch
Each trade takes one piece of the view, using Monday's quotes. You'd pay near the ask, so the table shows both.
Anthropic lists by Nov 30. It prices below $1.5 trillion, somewhere around $1.3T to $1.45T, and it closes its first day below $2 trillion.
The perp is near $1.75T on the S-1's share count and higher after the deal's new shares, while the reported book sits well below that. Every piece of information since Friday has been read the same way: this is coming, and it's coming cheaper.
We could be wrong. Monday disagreed with our own Friday note twice, once when the deal didn't go down with Nvidia and once when listing week turned out not to be priced like a cliff with a quiet day after it. The S-1 shows $16.3 billion of first-half revenue, and no single customer is above 10%. If the roadshow turns up far more orders than shares, a big first day could reach $2T without much effort.
The amendment. It sets the range and the share count. If it brackets $1.25T to $1.5T, we're on the right track. If the range starts with a 2, we're wrong.
Nov 20 odds. They should climb toward 80% as the roadshow goes. If they fall on a day when the market's fine, something's off with the book.
The perp against the range. Watch the gap between ANT near 103 and the top of the range. We think it closes from above.
The Nov 20 straddle. About 16% of the forward today. If it trades down on a good roadshow, the post-event pattern has started early.
We'll score this call in public within a week of the listing, with the actual pricing and first-day close next to everything printed here, right or wrong.
This article is for information only and is not financial advice.