NeuroAI NEUROAINEUROAI.SITE
China AI, reported clearly
1,300+ verified stories on models, robotics, BCI and the business of AI.
Browse stories
ESC

Two Days After the IPO, the $83 Price Is Worth 7% More

Five cents. On November 20 the stock traded as low as $83.05.

2026-11-21 · 4586 words · NeuroAI
Two Days After the IPO, the $83 Price Is Worth 7% More

Five cents.

On November 20 the stock traded as low as $83.05. The offer price was $83. The gap between the two is 5 cents, or 0.06% of the offer price, thinner than most brokers' commission.

It closed at $88.91, down 8.5%, after ranging between $83.05 and $92.56, a spread of 11.5%.

The open was the high

On the first day, November 19, the stock opened at $111.50, 34% above the offer price, and closed at $97.17.

The opening print was the high of the day.

Anyone who bought at $111.50 was down 12.9% at the close, a loss of $14.33 a share. The next session took another 8.5%. From $111.50 to $88.91 is 20.3% in two days. The bid that stood at $111.50 on day one did not stand on day two.

Two bases, two true numbers

The 17.07% gain reported on day one compares $97.17 with $83. The -12.86% loss compares $97.17 with $111.50. Same stock, same session, both numbers correct, opposite directions.

The 17% in the headline is measured from the offer price. The -12.9% an actual buyer feels is measured from the trade price. Different bases let one session be reported as a rally or as a double-digit loss. Both stories are true. The one worth using here is the second, because the low on day two, $83.05, is measured against the offer price, not the open. Going back to the start means returning the gains of both days at once.

What 7.1 percent is

$88.91 divided by $83 is 1.0712. Two sessions after pricing, the premium to the offer price is 7.1%.

34%, then 17%, then 7%.

Day one gave back half. Day two gave back most of the rest. 7.1 divided by 34 is 0.209, so about one-fifth of the opening premium survives.

That speed is not a valuation moving; it is the anchor moving. The anchor shifted from "what is an AI company worth" to "what did it earn this quarter," and when an anchor moves, the price jumps rather than walks.

The stock is still above $83 and has not closed below it. But the distance is $5.91, and the low of November 20 already touched $83.05.

Who sold

The base offering was $81.34 billion: 980 million Class A shares, of which the company sold about 956 million and existing holders about 23.65 million.

At $83, the selling shareholders took about $1.96 billion, or 2.4% of the base deal. Nearly all of the money went to the company, not to insiders cashing out.

After the over-allotment was exercised in full, the total reached about $93.5 billion, including the secondary shares. The difference between $93.5 billion and $81.34 billion is $12.16 billion.

Two prices, one company

A private placement of about $26 billion closed alongside the IPO, with Microsoft, Nvidia and Google among the participants. That is 32% of the base offering.

Public markets bought float. The private placement bought new shares. Two pockets, two prices, one company priced twice in the same week. Whether a discount exists between the two has not been disclosed.

Add the two raises: $119.5 billion. The company's entire 2025 revenue was $4.585 billion. The ratio is 26.

One fundraising round equals 26 years of revenue. The money arrives first and the revenue arrives later, and the gap between them is what buyers paid for.

Two share counts, one market cap

At $88.91 the market capitalization is about $1.175 trillion. Published estimates for the company range from $1.27 trillion to $1.76 trillion, a spread of $490 billion.

The difference comes from which share count is used. That is the only public explanation, and no breakdown of the share classes has been published. The range has to be quoted with that caveat attached. Without it, $1.27 trillion and $1.76 trillion read as two markets disagreeing about one company, when they are two arithmetic conventions applied to the same price.

At $83 the figure would be about $1.1 trillion, an offer-price valuation rather than a market value.

Notice that $1.175 trillion is $95 billion below the bottom of the published range. Same price, two share counts, $95 billion apart. None of it came from trading.

$1.175 trillion is 39% of Nvidia's $3.0253 trillion. Revenue of $4.585 billion is 4% of Nvidia's $115.8 billion quarter. Both numbers are on the page; dividing them requires no estimate.

Put the two ratios side by side and the pricing becomes legible. Nearly 40% of the market cap on 4% of the revenue. The 35 points in between are the market's quote for what this becomes next year. That quote is not a lie or a promise; it is a number that changes daily, and it is already 20% lower than it was at the opening print.

256 times

In 2025, consumer revenue was $3.796 billion, or 82.79%, and subscription revenue $789 million, or 17.21%. Together, $4.585 billion.

Divide November 2026's market capitalization by 2025's revenue and you get 256.

That multiple is not a verdict. It is two different years placed in one fraction. The denominator has already happened. The numerator has not.

A company spends months between filing a prospectus and setting a price. It took two sessions to travel from a 34% premium to 7%.

The first graded item is not profit. It is whether growth can carry 256 times. How many quarters $4.585 billion of revenue needs to support $1.175 trillion has no published answer. What has an answer is the other number: 34% to 7%, in two sessions.

The company reports quarterly. The market settles daily. When the two timetables collide, price moves first and the filing follows.

At the close on November 20: $88.91, $5.91 above the offer price.

*Disclaimer: This article is based on publicly available information and is for reference only. It does not constitute investment advice.*

Related coverage

Cite this article

NeuroAI (脑机网), "Two Days After the IPO, the $83 Price Is Worth 7% More," 21 November 2026, https://neuroai.site/a/na-stock-anthropic-second-day-1121

Machine-readable: BibTeX · RIS · JSON · Markdown

More in “stock” → · Back to home · Markdown version

China AI, reported clearly
1,300+ verified stories on models, robotics, BCI and the business of AI.
Browse stories