What the Price Asks
Anthropic's amended filing of 17 November set a range of $70 to $78 a share on 825 million shares, 801 million of them new. This piece was written on 18 November, before the IPO priced at $83.00, and runs the method Chan Ahn gave Benzinga in August at the $74 midpoint. On the filing's adjusted count of about 18 billion shares, the midpoint implies roughly $1.33 trillion. On the August method that needs revenue growth of about 24% a year, to roughly $475 billion by 2036, at a 10% cost of equity. Treating the roughly $213 billion of pro forma cash as surplus lowers the bar to about 22% a year. Even from the $81 billion October run-rate, the following nine years still need about 19% to 22% a year, and the company says its growth will slow. The piece then covers the cost structure, the filled-in lock-up calendar, the SpaceX precedent and what a buyer takes on. It does not value Anthropic and takes no view on how the listing will go.

19 November 2026. Written on 18 November, before Anthropic priced its IPO at $83.00 a share (final prospectus, 18 November 2026). Figures are from its amended registration statement of 17 November 2026 and use the $70–$78 range then on offer, unless the text names another source.
Interest disclosure: T-OpenAI and T-Kalshi are tokenized loan participation rights providing economic exposure linked to those companies' pre-IPO valuations, each issued through a dedicated issuing subsidiary. T-SpaceX, issued through its own dedicated issuing subsidiary, provides economic exposure linked to SpaceX; following the SpaceX listing, the investment underlying T-SpaceX is being divested. Redemption opens only once a Redemption Start Date is announced, and as of 19 November 2026 none has been. Tessera's interest in how Anthropic's listing is received runs in more than one direction. OpenAI competes with Anthropic, so on one side sits a competitor's valuation. On the other sits a supplier: Anthropic's filing records compute agreements of up to $120.5 billion with SpaceXAI, a SpaceX subsidiary, and those agreements depend on Anthropic paying. SpaceX's own prospectus also names Anthropic among its competitors, and on 14 August 2026 SpaceX acquired Anysphere, the company behind the coding tool Cursor. A well-received listing could also affect the window for other private AI companies, OpenAI among them. Tessera has no arrangement with Anthropic, issues no product referencing it, and none is planned. It has no arrangement with Amazon, Google, Microsoft or NVIDIA. Tessera's founder, Chan Ahn, worked earlier in his career at Goldman Sachs, JPMorgan and Credit Suisse, now part of UBS. Goldman Sachs is a representative of the underwriters of this offering, and JPMorgan and UBS are among the underwriters; Tessera has no arrangement with Goldman Sachs, JPMorgan or UBS. Tessera, and entities under its control, hold related exposures. Read what follows with that in mind.
Anthropic's amended filing of 17 November sets a range of $70 to $78 a share on 825 million shares. Of those, 801 million are new shares and 24 million are being sold by existing holders. The IPO is expected to price this week (Renaissance Capital, 10 November), with trading slated to begin on Thursday 19 November (Renaissance Capital, 13 November). In August, Benzinga asked Chan Ahn, Tessera's founder, what a reported $2 trillion valuation would require, and on 3 November he scored that answer against the first filings, and some of it went against him. Here the same exercise is run at the range now on offer, at its $74 midpoint. After that come the risks, the unlock calendar, the SpaceX precedent, what a buyer takes on and what the result will be read to mean. This piece does not value Anthropic, and it takes no view on whether the listing will go well or badly.
What the Range Implies
The test starts with the whole company. The filing's table of shares outstanding after the offering, once options, restricted stock units, Amazon's convertible notes, the donation matching plan and about 0.85 billion shares held by Anthropic's own public benefit subsidiary are added, comes to about 18 billion. Multiplied by the $74 midpoint, that is roughly $1.33 trillion (our arithmetic, from the 17 November filing). Renaissance Capital puts the fully diluted figure at $1.2 trillion, but leaves out $111 billion of convertible debt (Renaissance Capital, 10 November). Count that debt as shares and the two figures roughly agree.
The method is the one Chan used in August, applied without change so the price gets a fair test. Revenue grows at a constant rate from a 2026 base of $54 billion. The free cash flow margin rises in a straight line from zero to 25% over ten years, every year's cash flow counts, and the final year's cash flow is capitalized at 25 times. On that basis, in August, $2 trillion needed growth of about 30% a year at a 10% cost of equity, taking revenue to roughly $725 billion by 2036. About $1.33 trillion needs about 24% a year, to roughly $475 billion. At a 13% cost of equity it needs about 28% a year, to roughly $620 billion.
The August arithmetic left out cash. The filing's pro forma figures add the offering's net proceeds at the midpoint to the balance at 30 September, and on that basis Anthropic would hold about $213 billion of cash and short-term investments. If all of it is treated as surplus, the business itself is priced at about $1.12 trillion, which needs about 22% a year, to roughly $400 billion. Much of the cash is earmarked for compute, so a fair figure lies somewhere between the two. The margin assumption pushes the answer either way. At $1.12 trillion, a 35% free cash flow margin needs about 18% a year; a 20% margin needs about 25% a year, to roughly $500 billion (all our arithmetic).
The bar is therefore lower than it was in August. The price is below the $2 trillion reported then, and lower again with the cash counted. What has changed about the base is its source, which is now the company itself. Revenue for the first nine months of 2026 was $33.2 billion. A fourth quarter matching the third, at $17.0 billion, would take the year to about $50 billion. Run-rate revenue was about $81 billion as of 31 October, up $10 billion in a month, which points higher (S-1/A, 17 November). That run-rate annualizes a 28-day window, though, and the company states that it does not expect revenue "will continue to grow at comparable rates".
A bull would answer that the October run-rate already exceeds what this path needs in 2027, so most of the growth is behind it. Starting from the run-rate does not remove the requirement. The nine years that follow still need about 19% a year to reach $400 billion and about 22% to reach $475 billion (our arithmetic), and the company says its growth will slow.
Margin is the other half. Adjusted operating margin was 19% in the second quarter and 15% in the third. Under GAAP, the second quarter remains the only profitable one the filing shows, and the third lost $567 million. Anthropic attributes most of the fall in adjusted margin to $5.0 billion of short-term compute, bought when customer demand or research needs outran its long-term capacity, and it expects its use of short-term compute "to continue to increase in the near term and to in turn compress our operating margin" (S-1/A, 17 November). As Chan noted on 3 November, this sits slightly against the simple form of Chan's argument, because the third quarter suffered from too little committed capacity. The new language moves it a little further the same way. When the offering completes, Anthropic will also record about $14.2 billion of stock-based compensation at once. The first amendment, on 30 September figures, gave $8.2 billion, the number used on 3 November, with a further $24.1 billion still to be recognized; part of that $24.1 billion is now inside the new charge, so the two figures cannot simply be added.
In Chan's view his August line still holds, with the numbers moved: "You can underwrite the growth or you can underwrite the margin. Underwriting both at once is the leap being asked of public investors" (Benzinga, 24 August 2026). At this price the leap is shorter. It still has to cover both.
The Risks in the Filing
The cost structure remains the risk Chan has written about most, and the amendments since 3 November have moved its figures while leaving its shape alone. About $70.7 billion of commitments, substantially all of them for compute and infrastructure, falls due in the twelve months after 30 September, against the $69.8 billion used on 3 November. Contracted revenue not yet recognized stood at $13.8 billion, about three-quarters of it expected to be recognized within a year. As the earlier piece said, that is a floor and falls short of the split Chan asked for, since it leaves out what committed customers have already used. Three agreements alone carry minimum spending of about $121.3 billion with Google, $110.0 billion with Amazon and $47.2 billion with Microsoft, payable "regardless of actual usage." Against those fixed sums, the company describes its revenue as "primarily consumption-based, and much of it is not subject to long-term contractual commitments" (S-1/A, 17 November). That revenue is the floating leg, and it turns over quickly: models released in the previous six months produced over 80% of third-quarter revenue, a measure of how much depends on the latest release.
The concentration Chan wrote about in October shows up more in the sales channel than in the customer list. Cloud marketplaces carried about 44% of revenue for the nine months and 52% of the amounts customers owed at 30 September. The five largest customers account for a falling share, about 36% of revenue in the first quarter and about 21% in the third (our arithmetic, from the filing's figures). That counts in the company's favor.
The share itself carries the last risk. Class A voting power is "limited to an aggregate of 49.9%". The co-founders' entity holds a single share carrying the rest, and a trust can generally appoint most of the board. Anthropic is a public benefit corporation, so its directors must balance "the pecuniary interests of our stockholders" against the interests of those affected by its conduct and its stated purpose (S-1/A, 17 November). All of this is disclosed, and all of it is deliberate. It also means a public buyer has close to no influence over any of the risks above.
The Unlock Calendar
The lock-up is where the filing has changed most. On 3 November its share counts were blank. They are now filled in, and the schedule runs partly on day counts and partly on earnings dates (S-1/A, 17 November; every figure is an "up to").
Up to 315 million shares are released from the first day of trading, and up to 1.3 billion more across days 27, 48, 62 and 76, so up to about 1.6 billion in all (our sum) before the first annual results, if those arrive after day 76. That is about twice the number of shares sold in the offering. On the second full trading day after the 2026 annual results, up to 1.7 billion more are released, followed by five releases of up to 480 million each over the next ten weeks. When first-quarter 2027 results come out, or on day 180 if that is earlier, up to 3.0 billion are released, then four releases of up to 447 million. Second-quarter 2027 results bring up to 558 million, then five releases of up to 502 million. The longest lock-up ends when third-quarter 2027 results come out, or on day 365 if that is earlier, with up to 1.9 billion.
Goldman Sachs and Morgan Stanley may waive the restrictions early. Eligibility also differs from sale. Each release turns a restriction into a choice for each holder, and the price on the day decides it. The pattern to expect is a staircase, with its tallest steps tied to earnings. The first annual results matter most: they will be the first full year Anthropic reports to holders who owe it nothing, and the first large release follows two trading days later. In SpaceX's case, as Chan wrote on 3 November, the decline came with the earnings and not on the day the shares became free to sell.
The SpaceX Precedent
SpaceX is the natural comparison on two counts, size and the timing of the first large release. It raised $75 billion in June, $86 billion with the greenshoe, and Anthropic's base deal of about $61 billion would be the second largest on record (Renaissance Capital, 10 November). On Renaissance's figures, SpaceX rose 19% on its first day and peaked on its third at 57% above its offer price. It then fell 49% from that peak to a low on its 37th trading day, and as of 12 November it stood 4% above its offer (Renaissance Capital, 13 November). By Chan's count, that low came on 5 August, the first session after SpaceX filed its first results as a public company, which it did after the close on 4 August (SpaceX, Form 8-K, 4 August 2026). Most of the fall had come earlier; on 27 July Chan noted the stock trading around $115. SpaceX's first large lock-up release came on the second trading day after those results, as Chan set out on 2 September, and Anthropic's first large release falls in the same place.
SpaceX was no outlier. Renaissance's table shows all three of the largest US listings, SpaceX, Alibaba and Visa, rising strongly at first, peaking early within two months and then trading below their offer price. Chan made the same point in July: "Demand for an allocation is not the same thing as demand for the stock." Anthropic differs in one respect that matters. Its offering is almost entirely new money, with about 97% of the shares coming from the company (our arithmetic). Holders who want out are mostly not selling in the IPO, and if they sell, they will do it down the staircase above.
What a Buyer Takes On
An IPO is usually an exit for the holders who sell into it. Here most of them are not selling yet, so the question a buyer faces moves from who is selling today to who can sell later, and when. The calendar above answers that, and it is worth reading before the first trade.
The filing answers the rest. A buyer gets an economic interest and very little say. Reported profit will take a charge of about $14.2 billion in the quarter the offering completes; it catches up on stock awards already partly earned and is not a cost of that quarter's business. The company, the selling holders and the underwriters negotiate the offer price, and the filing says it "will not necessarily reflect" where the shares trade. The ties to suppliers run in both directions. After the offering, Amazon will hold about 1.92 billion of Anthropic's non-voting shares and Google about 1.79 billion. Amazon's stake would be 23.8% if its notes converted in full, ignoring an ownership cap it has chosen not to waive. Google and NVIDIA have each agreed to buy about $10 billion in the private placement alongside the IPO, and Anthropic's compute commitments to Amazon and Google, set out above, run over the next decade or so (S-1/A, 17 November). For a portfolio that already holds those suppliers, Anthropic adds exposure to which model wins and very little to whether the spending continues.
This is a case for reading the calendar and the share classes before buying in the IPO, and nothing in it argues against buying.
What the Result Will Be Read to Mean
Whatever the first weeks bring, they will be read as a verdict on more than one company, and we think they should not be. US IPOs of at least $50 million in market value are down 35.9% in number from the same point last year, while proceeds are up 314.7% (Renaissance Capital, as of 17 November). In October, working from Renaissance's figures to early October, Chan wrote: "That is not a shut market. It is one where two issuers set the totals and, behind them, fewer and larger deals get done." Anthropic would make three. Renaissance's index of recent listings was up 1.2% for the year as of 12 November, against 11.8% for the S&P 500.
A strong debut will be taken as the window opening for other large private AI companies, and a weak one as a reason for their holders to wait. Either reading will say more about price than about the businesses. In August Chan wrote that "The option to remain private has become a valuation input in its own right." This listing will be treated as a price on that option for every holder still exercising it. Tessera has an interest in that window, as the disclosure above states, and does not guess which way it will go.
For the wider AI build, the read-across runs through balance sheets. By the filing's own account, Anthropic's largest suppliers are also its shareholders and its competitors, Google above all, and $26 billion of the money raised alongside the IPO comes from suppliers and partners at the IPO price. In August Chan asked of the AI build's revenue not whether it exists but "what share of it is independent of the capital that created it." Filings can answer that, and a share price cannot. The listing adds a daily public mark on the stakes those suppliers hold, set by people who owe Anthropic nothing.
Sources: Anthropic, PBC, Form S-1/A, SEC EDGAR accession 0001628280-26-074338, filed 17 November 2026, read via the StockTitan reproduction on 17 and 18 November 2026: cover page; capitalization and dilution, including the adjusted share table; key metrics (Run-Rate Revenue); quarterly results; MD&A (key factors, Adjusted Operating Income, liquidity, contractual obligations and infrastructure service arrangements); revenue note (remaining performance obligations); risk factors; description of capital stock; principal and selling stockholders; concurrent private placement; shares eligible for future sale and lock-up. SEC EDGAR submissions list for Anthropic, PBC, checked 18 November 2026. Renaissance Capital, "Anthropic sets terms for $61 billion IPO", 10 November 2026; "How the largest US IPOs performed after listing and what it means for Anthropic", 13 November 2026; US IPO weekly recap, 13 November 2026; US IPO market statistics as of 17 November 2026. Space Exploration Technologies Corp., Form 8-K (Item 2.02), 4 August 2026. Benzinga, Namrata Sen, 24 August 2026. Chan Ahn, "Record on the Headline, Thin Underneath", 27 July 2026, "Eligible Was Not Sold", 10 August 2026, "Who Counts the Capex", 17 August 2026, "The Money Goes Round", 28 August 2026, "The Announcement Outlasts the Holder", 2 September 2026, "Half Through the Marketplace", 5 October 2026, "Who Was Selling", 7 October 2026, and "Fixed Against Floating", 3 November 2026, as published. Valuation arithmetic is Tessera's, on the method described in the text.
This is market commentary, not investment advice. It is not a recommendation to acquire, hold or redeem any Tessera product, or to take or avoid exposure to any company mentioned.
T-Tokens are tokenized loan participation rights, not equity. High risk. DYOR. Not financial advice. Not available in the US or other restricted territories. tessera.pe/terms
