Company Overview

Anthropic IPO Valuation, Share Price and First-Day Trading Explained

Jon Sinclair using Luminix AI
Jon Sinclair using Luminix AI Strategic Research

Anthropic's IPO valuation was set on November 18, 2026, when the company behind the Claude AI models priced the largest initial public offering on record. Its shares began trading on Nasdaq the next day under the ticker ANT1,3. The listing matters beyond one company. It is the first time public investors have put a price on a frontier AI lab, and it now serves as the reference point for OpenAI. Investors bid up shares to above the marketed range, then gave back about half of a large opening jump by the end of the first session. Whether the price holds depends on revenue keeping pace with the compute bills Anthropic has already signed.

What was the Anthropic IPO price?

A company going public shows investors a price range during its roadshow and sets the final price once orders are in. Anthropic marketed its Class A shares at $70 to $78 and priced them at $83, according to its final prospectus1. Pricing above the top of the range means the order book was deep enough to accept a higher price than investors had been shown.

Anthropic also enlarged the deal. It added 155 million new shares at the last minute, taking the base offering to 980 million shares1. All of the added shares were new stock issued by the company. The seven co-founders sold a small block that stayed the same size, and the company receives none of that money. Gross proceeds from the IPO itself came to $81.34 billion, more than SpaceX raised in its June 2026 listing1,3.

Many headlines cite a larger total of $107.3 billion. That figure adds a separate $26 billion private placement of non-voting Class N stock, sold at the same price to seven chip and cloud companies. Google and Nvidia each bought about $10 billion1. Goldman Sachs led the deal. Morgan Stanley is the stabilization agent, the bank that can buy or supply shares to steady early trading1.

What is Anthropic's valuation at the IPO price?

Market capitalization is the share price multiplied by the number of shares, and Anthropic's prospectus supports several answers to how many shares there are. The company has two classes with equal economic rights. Class A shares are listed. Class N shares carry no vote and are held mainly by Amazon, Google and the placement buyers. A fully diluted count goes further. It adds employee stock options, restricted stock units (shares promised to staff that vest over time) and Amazon's convertible notes that have not yet turned into stock. Using the final Form 424B4 prospectus, the figures at the offer price line up as follows2.

Share basis Shares Value at $83
Class A only 7.58B $629B2
Basic: Class A plus Class N 13.22B $1.097T2
Basic plus options and RSUs ~16.0B $1.33T2
Prospectus fully diluted total 18.14B $1.505T2
Above plus unissued plan reserves 20.19B $1.675T2

Reuters and Axios reported the basic and prospectus fully diluted figures, rounded, which is why the same deal produced two headline valuations2. Some data screens show the Class A figure, which leaves out the Class N block and so does not value the whole company. Outlets that kept citing about $1.08 trillion after pricing were using the share count from before the deal was enlarged2.

The basic count includes 846.8 million Class N shares, worth about $70.3 billion, held by Anthropic's own Public Benefit Fund, a subsidiary that finances its public-benefit work2.

The largest economic owners have no vote. Amazon holds 14.6% and Google 13.4% of the company, all in Class N stock. Amazon's stake would be about 23.6% if a cap on converting its notes did not apply2,5.

How did Anthropic stock trade on its first day?

Trading opened shortly before noon Eastern time on November 19. The first trade came at $111.50, about 34% above the offer price. The stock climbed a little further, then slid through the afternoon to close at $97.173.

The broader market was falling at the same time. The S&P 500 and Nasdaq each lost more than 2%, and it was the S&P's sixth losing session in a row, as investors sold AI stocks during a controversy around Nvidia3. Anthropic was still the most active stock on Nasdaq that day.

At the close, the FT valued the company at $1.3 trillion on outstanding shares and $1.75 trillion fully diluted2. The "$2 trillion" headlines from the morning applied the opening price to the diluted share count2,3.

Morningstar's fair-value estimate is $92 a share, and its analyst Malik Ahmed Khan called the close roughly fairly valued6. The open market settled near that estimate and did not hold the opening premium.

Large technology listings commonly rise on their first day. Across 2026 the average gain was about 20%3.

First-day close vs. IPO price
CoreWeave0%
Anthropic+17%
SpaceX+19%
Arm+25%
Reddit+48%
Figma+250%
Gain from offer price to first-day close.3

The first day has said little about what followed. Figma now trades about 51% below its offer price, while Arm went on to strong multi-year gains. Of the 15 largest tech IPOs of 2026, 13 trade below their first-day close3.

Is Anthropic stock expensive compared with OpenAI and Nvidia?

Young, fast-growing companies are usually valued as a multiple of revenue, because their profits are small or erratic. Anthropic's prospectus reports run-rate revenue, its recent monthly revenue annualized, of about $81 billion as of October 31, 20264,6. This is a company metric and differs from revenue booked under generally accepted accounting principles (GAAP). For next year, Goldman Sachs estimates about $136 billion of 2027 revenue. Goldman is also the deal's lead bank, and the prospectus itself gives no guidance4.

Company and basis Revenue base Revenue multiple
Anthropic, IPO price, basic shares Run-rate / Goldman 2027 estimate 13.6× / 8.1×4
Anthropic, IPO price, fully diluted Run-rate / Goldman 2027 estimate 18.5× / 11.0×4
OpenAI, $1.25T private round ~$50B run-rate, net of partner share ~25×4
Nvidia, ~$3.5–3.65T market value ~$362B trailing revenue ~10×4

On run-rate, public investors priced Anthropic below OpenAI's latest private valuation. The two are hard to compare, though. OpenAI reports revenue after the share paid to cloud partners, while Anthropic's run-rate is a gross measure. OpenAI's round is also only partly closed, with about $11 billion wired by November 64. Sources disagree on whether OpenAI's valuation counts the new money, and that question is still open.

Nvidia trades at a lower revenue multiple than either lab, and it earns a 74.8% GAAP gross margin and a 66.9% operating margin4. Anthropic's reported gross margin of about 85% leaves out the compute used to train models. Counting all compute and infrastructure, costs ran at about 57% of nine-month revenue4,6.

That cost load showed up in the third quarter, when Anthropic swung to a GAAP operating loss of $567 million4,6. The revenue base also changes the multiple sharply. On the $35.3 billion of GAAP revenue booked over the last four quarters, Trefis puts the IPO at about 31 times sales6.

Anthropic's largest financial obligation does not appear in its market value. The company has about $901 billion of mostly non-cancelable compute commitments, roughly 82% of its basic equity value. About 79% of its revenue is consumption-based, meaning customers pay for what they use and can scale back4,6.

How binding those commitments are remains unclear. One account based on Better Markets readings suggests about $215 billion may carry termination rights or counterparty conditions6.

The case for and against

The bull case

Anthropic's growth is large and well documented. Nine-month revenue rose about 1,200%, and net revenue retention, which compares what existing customers spend now with a year earlier, was about 900%6. Reliance on the top five customers fell sharply during the year, and William Blair reads nine-month free cash flow as positive. Melius Research has a $155 price target, and Scott Galloway argues the fully diluted multiple is about where Broadcom and AMD trade6.

The bear case

Anthropic owes $70.7 billion of its commitments within twelve months, many large customers have no long-term contracts, and Goldman Sachs is cited secondhand as not expecting positive free cash flow until the end of the decade6. Amazon and Google are both shareholders and the main compute suppliers, which leads Trefis to compare the arrangement with vendor financing at Cisco during the dot-com boom. Morningstar finds the largest IPOs of 2010 to 2025 had a median one-year return of −31%6. Public buyers also have little say, because a single share held by the co-founders carries a fixed 50.1% of the general vote and the Long-Term Benefit Trust elects most of the board5.

What to watch

  • Mid-December 2026The underwriters' 30-day option to buy up to 147 million more shares from the company expires. Exercise had not been confirmed as of November 20, and the outcome shows whether the banks supported the price in early trading1.
  • Early 2027About 76 days after pricing, legacy shares eligible for sale reach roughly twice the size of the IPO, and the largest single lock-up release is tied to the FY2026 results that follow5.
  • Late May 2027The day-180 lock-up gate, tied to first-quarter 2027 earnings, lets founders sell up to 29% of their shares, and the placement shares become eligible for resale under Rule 144 around the same time5.
  • 2027OpenAI's possible listing, which investors will price against the run-rate multiple Anthropic achieved in public markets3,4.

Get Custom Research Like This

Start Your Research

Source Research Reports

The full underlying research reports cited throughout this analysis. Tap a report to expand.

Report 1 Research the Anthropic IPO's final offer price versus the $70 to $78 marketed range, including whether it was priced above, within, or below range and any upsizing. Pull from the SEC S-1 and 424B4 prospectus, Reuters, Bloomberg, and WSJ coverage. Report shares sold (primary vs. secondary), the greenshoe, gross proceeds, listing exchange, ticker, and lead underwriters, in a table with sources. If no IPO has priced, state that clearly and describe the latest verified status.

The IPO has priced. On November 18, 2026, Anthropic, PBC set the Class A offer price at $83.00, above the marketed range of $70.00–$78.00, and upsized the deal. It is not still in the marketed range, and it is not unpriced.

The mechanism was a late, demand-driven reset of both price and size. The November 10, 2026 S-1/A (kept in the November 17 S-1/A) marketed 825 million Class A shares at $70–$78, with a 15% greenshoe of 123,750,000 shares. After the book was covered several times, the company raised the price $5 above the top of the range and added 155 million primary shares, taking the base deal to 980 million. Secondary size did not change. The greenshoe was reset to 15% of the new base, 147 million shares. Registration of the extra shares was done on Form S-1MEF on November 18; the final terms are in the Rule 424(b)(4) prospectus filed November 19. The SEC declared the S-1 effective November 18. Shares began trading November 19 on Nasdaq as ANT. Delivery was expected on or about November 20, 2026.

At $83, the base offering’s gross proceeds are $81,340,000,000 ($81.34 billion), the figure printed on the 424B4 cover. That is the IPO gross only. Headlines that say about $107.3 billion are adding a separate $26.0 billion concurrent private placement of non-voting Class N stock at the same $83 price. That placement is not part of the registered IPO and is not included in the $81.34 billion.

Deal terms

Item Final terms Marketed terms (S-1/A, Nov. 10 and Nov. 17, 2026) Sources
Offer price $83.00, above range ($5, or about 6.4%, above the $78 top; about 12% above the $74 midpoint) $70.00–$78.00 424B4 cover; company pricing release; Reuters; Bloomberg; WSJ
Above / in / below range Above — Same
Shares sold, total 980,000,000 Class A 825,000,000 Class A 424B4; company release
Primary (company) 956,351,354 801,351,354 424B4; company release
Secondary (selling stockholders) 23,648,646 (seven co-founders; company receives none of these proceeds) 23,648,646 (unchanged) 424B4; Bloomberg
Upsize +155,000,000 primary shares (+18.8% vs. 825 million; Reuters described it as about 19%) — 424B4 vs. S-1/A; Reuters; WSJ (“sold more shares than planned”); Bloomberg
Greenshoe (overallotment) 30-day option for up to 147,000,000 additional Class A shares from the company at the $83 IPO price. No underwriting discount on option shares. Full exercise would add $12.201 billion of gross proceeds (147 million × $83). Not known to have been exercised as of the Nov. 19 prospectus and Nov. 18 pricing release. Up to 123,750,000 shares (15% of 825 million) 424B4 cover footnote; company release
Gross proceeds (base IPO) $81,340,000,000 (980 million × $83). Cover table: proceeds before expenses to Anthropic $78,840,438,875; to selling stockholders $1,949,565,525; underwriting discount $550,000,000 ($0.56122 per share). $57.75–$64.35 billion at the range on 825 million shares 424B4 pricing table; Reuters ($81.3 billion); WSJ ($81 billion)
Exchange Nasdaq Global Select Market Nasdaq Global Select Market 424B4; company release
Ticker ANT ANT 424B4; company release
Lead underwriters Goldman Sachs & Co. LLC and Morgan Stanley & Co. LLC as representatives. Goldman was lead left; Morgan Stanley was second and the stabilization agent. Syndicate of 22 banks. Same lead pair named in the S-1 424B4 cover; Bloomberg (Oct. 26 roles); Reuters syndicate count; company release (prospectus copies from those two banks)

Primary gross at $83 is $79,377,162,382 (956,351,354 × $83). Secondary gross is $1,962,837,618 (23,648,646 × $83). Those two sum to the prospectus total of $81,340,000,000. The “proceeds before expenses” lines are after the $550 million discount, which is why they do not equal the gross.

What “above range and upsized” actually means

Pricing above the top of a filed range, rather than merely at the top, is the stronger signal. The company did not just clear a soft book at $78. It moved the clearing price outside the range investors had been shown on the roadshow, and it did so while also selling 155 million more primary shares. WSJ reported on November 17 that advisers were already discussing a price above $80 and a fully diluted value around $1.5–$1.6 trillion, versus roughly $1.33 trillion at the midpoint of the filed range. The print at $83 landed in that zone.

Reuters, Bloomberg, and WSJ all describe the same outcome: $83, above $70–$78, and a larger share count than planned. Bloomberg’s count of the upsize matches the filings: the company had marketed 801.4 million shares and the co-founders 23.6 million; the final sale was 956.4 million by the company and 23.6 million by the co-founders. Secondary was not upsized. The extra capital is new primary issuance.

The greenshoe is structurally unusual in one respect that is on the face of the 424B4: underwriters receive no discount or commission on the option shares. They can still buy up to 147 million shares at $83 for 30 days, which is the stabilization tool, but the fee pool of $550 million is only on the base 980 million shares. That is the same concession reported on SpaceX’s June 2026 IPO.

What is not in the $81.34 billion

Immediately after the IPO, and contingent on it, seven strategic buyers agreed to purchase $26.0 billion of non-voting Class N stock at $83: Google and NVIDIA about $10.0 billion each, AMD about $2.5 billion, Broadcom, Micron, and Microsoft about $1.0 billion each, and TSMC about $500 million (313,253,007 Class N shares in total). Underwriters take no fee on those shares. Bloomberg and Reuters add that placement to the IPO and report a combined $107.3 billion. The 424B4 estimates net proceeds to Anthropic from its IPO shares plus the placement at about $104.8 billion, or about $117.0 billion if the greenshoe is exercised in full. Those figures are not IPO gross proceeds.

Latest verified status (as of November 20, 2026)

The offering has priced and has begun trading. It is not a pending deal.

  • November 18, 2026: priced at $83; registration statement effective; 30-day greenshoe granted.
  • November 19, 2026: 424B4 filed; first trade on Nasdaq Global Select Market under ANT. Shares opened at $111.50 (about 34% above $83) and closed at $97.17 (about 17% above the offer price), per Reuters and WSJ live coverage.
  • On or about November 20, 2026: expected closing and delivery, subject to customary conditions, as stated in the company release and the 424B4. Whether the greenshoe has been exercised is not in the pricing release or the November 19 prospectus.

Fully diluted value at the $83 offer price was reported at about $1.5 trillion by Reuters, Bloomberg, and WSJ. That is a valuation at the offer price, not the first-day trading value.


Recent Findings Supplement (November 2026)

Anthropic priced its IPO above the marketed range on November 18, 2026, at $83 per Class A share (vs. the $70–$78 range set in the November 10 S-1/A amendment). The offering was upsized by ~19% to 980 million shares amid strong demand (book reportedly ~6x oversubscribed), making it the largest IPO on record by proceeds raised.[1][2]

Trading began on the Nasdaq Global Select Market under ticker ANT on November 19, 2026. The company also raised $26 billion via a concurrent private placement of non-voting Class N shares to seven strategic partners (Google, Microsoft, Nvidia, AMD, Broadcom, Micron, TSMC) at the same $83 price, for a combined total of $107.3 billion.[3][4]

Key offering details (from final pricing, 424B4 prospectus, and contemporaneous coverage):

  • Final price vs. range: $83/share, above the $70–$78 marketed range (and above the ~$74 midpoint).[2]
  • Upsizing: Base offering increased from 825 million shares to 980 million shares (+155 million primary shares via S-1MEF).[5]
  • Shares sold:
    • Primary (Anthropic): 956,351,354 Class A shares.
    • Secondary (selling stockholders/co-founders): 23,648,646 Class A shares.[6]
  • Gross proceeds (IPO only): ~$81.3 billion (980 million shares × $83).[2]
  • Greenshoe/underwriters’ option: Up to 147 million additional Class A shares from the company (30-day option).[7]
  • Listing/exchange/ticker: Nasdaq Global Select Market; ANT.[2]
  • Lead underwriters/representatives: Goldman Sachs & Co. LLC (lead left) and Morgan Stanley & Co. LLC (stabilization agent); broader syndicate of ~22–23 banks.[8][9]

The 424B4 final prospectus (filed ~November 19, 2026) and related S-1 amendments/S-1MEF filings on EDGAR (CIK 0002130357) contain the definitive terms, consistent with Bloomberg, Reuters, and other coverage from November 18–19, 2026.[10]

No material changes or new regulatory updates to these terms have been reported since pricing. Shares opened at $111.50 (34% pop) on November 19 before closing at ~$97.17 (up ~17%).[11]

Sources for table/facts above: Primarily Bloomberg (Nov 18–19, 2026), Reuters (Nov 18–19, 2026), company pricing release/424B4 details via SEC filings and aggregators (Nov 19, 2026), and contemporaneous WSJ/FT/Forbes reporting. All post-date May 20, 2026, and reflect the most recent verified information as of November 20, 2026.

Report 2 Establish the implied market capitalization at the offer price on basic shares outstanding and on a fully diluted basis (options, RSUs, warrants, convertibles). Cite the prospectus capitalization and share count tables and reconcile differences between press-reported figures. Show the calculation steps and flag discrepancies among outlets.

Anthropic’s $83 IPO price does not imply one market capitalization. The final prospectus (Form 424B4, dated November 18, 2026) states 13,219,642,276 shares of all classes outstanding after the offering and the concurrent private placement. Stripping the two non-economic shares leaves 13,219,642,274 economic shares and an implied equity value of $1.097 trillion. The same document’s share-count table, once options, RSUs, remaining convertible notes, and the vested donation-matching obligation are added, produces about 18.14 billion economic claims and $1.505 trillion. That is the gap behind the “$1.1 trillion vs. $1.5 trillion” headlines.

Basic shares at the offer price

The binding count is in “The Offering” of the 424B4, not the pre-pricing S-1/As. After the upsized sale of 956,351,354 new Class A shares and the $26 billion Class N placement (313,253,007 shares at $83), the company states:

  • Class A outstanding: 7,583,103,996 (7,730,103,996 if the 147,000,000-share underwriters’ option is exercised in full)
  • Class N outstanding: 5,636,538,278
  • Class T and Class F: one share each; Class E: zero
  • All classes: 13,219,642,276 (13,366,642,276 with the option)

Class T and Class F have no economic rights, so the economic basic count is Class A + Class N = 13,219,642,274.

Calculation:

  • 13,219,642,274 × $83.00 = $1,097,230,308,742 ($1.097 trillion)
  • Class A only: 7,583,103,996 × $83.00 = $629,397,631,668 ($629.4 billion)
  • With the greenshoe: 13,366,642,274 × $83.00 = $1,109,431,308,742 ($1.109 trillion)

Class N has the same economic rights as Class A and converts into Class A on a sale to an unaffiliated buyer. A “listed shares × price” figure of ~$629 billion therefore understates the claim on the business by the entire Class N block, including Amazon and Google.

The 5,636,538,278 Class N figure includes the Initial Public Benefit Issuance of 846,840,960 Class N shares (5% of fully diluted shares as of June 30, 2026, assuming full conversion of the Amazon notes). Those shares are legally outstanding and sit inside the 13.22 billion, but they are held by a wholly owned subsidiary and are not treated as outstanding for accounting purposes. At $83 they are worth about $70.3 billion of the basic market cap.

What “fully diluted” means in this filing

The 424B4 does not publish a single fully diluted share count or a fully diluted valuation. It does two things that have to be kept separate.

First, the outstanding-share footnote excludes, as of the September 30, 2026 capitalization date:

  • 1,774,342,693 Class A shares under options, weighted-average exercise price $2.91
  • 1,026,660,010 Class A shares under RSUs whose service condition is not met by November 19, 2026
  • 47,041,287 Class A shares under RSUs granted after September 30
  • 1,920,580,692 Class A shares reserved under the 2026 Equity Incentive Plan (plus up to 3,401,163,438 more as 2021 Plan awards return to the pool)
  • 128,038,713 Class A shares reserved under the ESPP
  • 933,698,210 Class N shares reserved under the Donation Matching Plan, of which 553,441,297 is the vested matching obligation
  • 27,991,850 Class N shares already issued after September 30 as matching contributions
  • Small post-September 30 acquisition issuances (561,963 Class A and 848,882 Class N) and 1,590,190 acquisition holdback shares

Second, the share-count table’s “Additional Adjustments” row applies the treasury-stock method to the options and adds the instruments the company treats as incremental claims. At $83, exercise proceeds of about $5.2 billion repurchase 62,208,882 shares, so net option dilution is 1,712,133,811. That row’s totals are:

Component Shares
Class A (incl. post-9/30 acquisition and holdback) 7,585,256,149
Class N (incl. public-benefit shares, rollover notes, matching issuance, vested matching obligation) 7,766,496,616
Options, treasury-stock method 1,712,133,811
RSUs (unvested at 9/30 plus grants after 9/30) 1,073,701,297
Economic total 18,137,587,873

The Class N increase versus the 5,636,538,278 already outstanding is exactly the rollover notes (1,547,676,309), the vested matching obligation (553,441,297), the post-September 30 matching issuance (27,991,850), and a small acquisition issuance. The rollover-note line is the Amazon convertible notes that were not converted at the offering because Amazon did not waive the ownership cap. If those notes converted in full without the cap, Amazon would hold 3,484,128,662 Class N shares (23.6% of capital stock) rather than the 1,936,452,353 it holds after the capped conversion.

18,137,587,873 × $83.00 = $1,505,419,793,459 ($1.505 trillion).

Narrower and wider constructs, all at $83, using only numbers in the 424B4:

  • Options and RSUs only, gross (no treasury-stock method, no notes, no matching obligation): 16,067,686,264 shares → $1.334 trillion
  • Options on a treasury-stock basis plus RSUs only: 16,005,477,382 shares → $1.328 trillion
  • Additional-adjustments total above: 18,137,587,873 → $1.505 trillion
  • That total plus the initial 2026 Plan reserve and the ESPP reserve (2,048,619,405 shares), still excluding the contingent 2021 Plan rollover of up to 3.4 billion: 20,186,207,278 → $1.675 trillion

There is no material warrant overhang in the exclusions. The convertible that matters is the Amazon note complex, already partly in the basic count via the capped conversion of 292,768,653 Class N shares and partly in the diluted count via the 1.55 billion rollover shares.

How the press figures line up — and where they don’t

Reuters, on the night of pricing, is the cleanest match to the prospectus: about $1.5 trillion fully diluted (options and restricted stock) and about $1.1 trillion non-diluted. Axios used the same pair (“nearly $1.1 trillion” / “around $1.5 trillion”). Both round $1.097 trillion to $1.1 trillion and $1.505 trillion to $1.5 trillion. AP, PBS, and Business Standard led with $1.5 trillion without always repeating the diluted qualifier in the headline.

Discrepancies that are arithmetic, not disagreement about the price:

  • $1.08 trillion vs. $1.10 trillion basic. The November 10–17 filings said 13,086,672,593 shares would be outstanding after an 825 million-share deal at the $74 midpoint. That count × $83 is about $1.086 trillion. The priced deal added 155 million primary shares and cut the private-placement share count (because $26 billion buys fewer shares at $83 than at $74). Net, the final basic count is 13.22 billion, not 13.09 billion. Outlets still citing ~$1.08 trillion after pricing were using the pre-upsize denominator.
  • Class A-only market cap (~$629 billion). Some data displays multiply the Nasdaq-listed class by price and ignore Class N. That is not a market capitalization of the company.
  • “Fully diluted” is not one number. Pre-pricing, CNBC’s $1.405 trillion and The Information’s $1.26–1.40 trillion at $70–$78 were built on roughly 18 billion shares that, in several reconstructions, included unissued 2026 plan reserves. At the final price, the $1.5 trillion headline matches the prospectus additional-adjustments total (~18.14 billion, notes and vested matching in, unissued plan reserves out), not “options + RSUs only” ($1.33 trillion) and not “plus empty option-pool reserves” ($1.68 trillion). AInvest’s post-pricing note that a still-wider stack (reserves, donation reserve, notes) reaches about 20.6 billion and ~$1.71 trillion is the plan-reserve case, not the Reuters case.
  • Index-provider counts. One post-listing compilation of index price documentation used 13,116,075,286 Class A and Class N shares — about 104 million below the 424B4 — and a convertible figure of 1,563,617,444, which is the pre-final rollover-note amount rather than the 424B4’s 1,547,676,309. Treat index shares as a vendor convention, not as a correction of the prospectus.
  • Trading-day valuations changed the price, not the method. Reuters reported a debut valuation of about $2 trillion fully diluted and as much as $1.5 trillion on outstanding shares during the session. The FT, at the $97.17 close, put the company at $1.3 trillion on outstanding shares and $1.75 trillion fully diluted; the Straits Times used about $1.3 trillion and $1.8 trillion. Those are $97.17 × ~13.22 billion and × ~18.1 billion, not a new share count. The stock opened at $111.50, which on the basic count is about $1.47 trillion before any dilution add-on.

Aswath Damodaran’s November 18 workbook (primary ~14.88 billion, fully diluted ~17.68 billion, option-adjusted value $76.98) was built on the amended S-1 before the upsize to 980 million shares and before the $83 price was set. It is a valuation of the claims, not a reading of the final outstanding table, and it subtracts option value rather than simply multiplying diluted shares by price.

Capitalization context, not equity value

The 424B4 capitalization disclosure (as summarized from the filing) shows cash and short-term investments of about $131.8 billion actual and about $232.4 billion pro forma as adjusted, with convertible notes falling from about $130.5 billion actual to about $109.7 billion pro forma after the capped conversion. Dilution math in the same filing puts pro forma as adjusted net tangible book value at about $128.2 billion, or $9.70 per share, and dilution to new investors at $73.30 per share versus the $83 offer price. Those figures explain how much cash arrived; they are not an alternative market cap. Net proceeds to the company from the IPO shares plus the placement are about $104.8 billion ($117.0 billion if the option is exercised in full). Gross IPO proceeds are $81.34 billion; selling stockholders (the seven co-founders, 23,648,646 shares) receive about $1.95 billion before expenses and the company receives none of that.

What to use

For a conventional post-deal equity value at the offer price, use $1.097 trillion: 13,219,642,274 economic shares outstanding × $83. State that this includes 846.8 million subsidiary-held public-benefit shares and excludes the greenshoe. For the figure Reuters, Axios, and the company-adjacent “$1.5 trillion” coverage are using, use $1.505 trillion: the 424B4 additional-adjustments total of 18,137,587,873 × $83, which folds in net options, RSUs, the unconverted Amazon notes, and the vested matching obligation, and does not fold in the empty 2026 plan and ESPP reserves. Anyone competing on the headline should label the denominator. The $400 billion gap is not a reporting error; it is the difference between shares that exist and claims the filing itself says can still be created.

Report 3 Document the opening price, first-day high, low, and close, percent change from the offer price, volume, and market cap at close. Cover the trading trajectory since, as of the latest data available. Compare with recent large tech IPOs (e.g., Arm, Reddit, CoreWeave, Figma) on first-day pop and subsequent performance.

Anthropic’s (NASDAQ: ANT) November 19, 2026 debut was the largest IPO on record by proceeds, but the first-day tape was a classic open-and-fade rather than a sustained re-rating. The stock priced at $83 (above the $70–$78 range), opened at $111.50, briefly touched $115.10, then sold off into a $97.17 close as the broader market dropped more than 2%. With only one session of history, the “trajectory since” is limited to that fade plus after-hours weakness.

First-day tape: open, range, close, volume, and valuation

Anthropic and selling stockholders sold 980 million Class A shares at $83 on November 18, raising about $81.3 billion—eclipsing SpaceX’s June 2026 haul. Trading began shortly before noon ET on November 19 under the ticker ANT.

  • Open: $111.50, roughly 34.3% above the offer price.
  • Intraday high: $115.10 (about +38.7% versus the offer).
  • Intraday low: $92.72 (about +11.7% versus the offer).
  • Close: $97.17, up 17.1% (or +$14.17) from the $83 offer and about 12.9% below the opening print.
  • Volume: Roughly 846–857 million shares (sources cluster around 846 million and 857.3 million), the most active name on Nasdaq and well above SpaceX’s first-day volume of about 522 million shares.
  • After-hours (Nov. 19): Around $92.58, down roughly 4.7% from the regular close.

Market-cap figures diverge by share-count convention, which matters at this scale. On approximately 13.22 billion shares outstanding after the offering and concurrent private placement (Class A plus non-voting Class N), the close implies roughly $1.28 trillion. Reuters and company-related reporting put the fully diluted value near $1.5 trillion at the $83 offer price (including options, RSUs, and related instruments); some coverage put the fully diluted figure near $1.75–1.8 trillion at the $97.17 close. The gap is structural, not a data error: outstanding shares and fully diluted shares are different bases. At the open near $111.50, headlines briefly cited a roughly $2 trillion valuation on the higher base.

The mechanism of the fade was straightforward. Early demand pushed the opening auction well above the offer, but selling intensified as the S&P 500 and Nasdaq each fell more than 2%—their sixth straight losing session for the S&P—and investors rotated out of the AI complex amid Nvidia-related controversy and bubble concerns. Anthropic’s debut did not lift the tape; it coincided with the day’s lows in the major indexes.

What “since the IPO” means right now

As of the latest available data (close and after-hours of November 19, 2026; today’s date is November 20), there is no second regular session to report. The only post-close signal is the after-hours dip toward the mid-$90s. Underwriters have a greenshoe option (reported around 147 million shares) that can be exercised in the days after pricing, which can supply stock and mute further upside in the near term. Lock-up dynamics and the path of AI sentiment will dominate the next several months more than the first-day print.

How the first-day pop compares with recent large tech IPOs

Anthropic’s +17% close-to-offer pop sits in the middle of the recent distribution: larger than CoreWeave’s flat debut, similar to SpaceX, smaller than Arm and Reddit, and far smaller than Figma’s extreme outlier.

Arm (September 14, 2023). Offer $51. Opened around $56.10 and closed at $63.59, up about 24.7% from the offer. The stock later became a multi-year winner: as of the November 19, 2026 close it traded near $177, roughly +247% from the offer price and well above the first-day close, even after a sharp pullback from 2026 highs above $400. A moderate pop did not prevent substantial subsequent gains once the AI cycle matured.

Reddit (March 21, 2024). Offer $34. Opened at $47, traded as high as about $57.80, and closed at $50.44, up 48% from the offer. Performance since has been volatile but still positive versus the IPO price: the November 19, 2026 close was about $89.88, roughly +164% from the offer and about +78% from the first-day close. It has given back a large share of its peak (52-week high near $263), illustrating that a strong first day and a strong first year do not immunize a stock from later drawdowns when growth narratives shift (in Reddit’s case, traffic and search concerns).

CoreWeave (March 28, 2025). Offer $40. Opened at $39 and closed exactly at $40—a 0% first-day return after pricing below the original range. As of November 19, 2026 the stock closed at $44.91, only about +12% from the offer, after having traded as high as roughly $138. A flat debut left less “air” to deflate, but the subsequent path has still been a round trip driven by AI-infrastructure financing and demand concerns. Modest day-one pricing did not guarantee a smooth ride.

Figma (July 31, 2025). Offer $33. Opened at $85 and closed near $115.50, a roughly +250% first-day pop—the largest among the comparables. By November 19, 2026 the stock traded around $16.26–$16.29, about 51% below the offer price and roughly 86% below the first-day close. The extreme pop transferred value from the company and selling shareholders to day-one buyers who then faced a long unwind as growth expectations normalized and lock-ups expired. It is the clearest recent counterexample to the idea that a large first-day gain predicts durable outperformance.

SpaceX (June 2026), for additional context, priced at $135, closed its first day up about 19% near $161, and by mid-November 2026 was only modestly above its offer price and below its first-day close—another case of a solid pop that did not compound.

Across a broader set of 2026 technology and AI listings, first-day gains averaged around 20%, yet most of the largest names were already trading below their first-day closes within months, with the group down on the order of 25% from those closes on average. A double-digit pop is common; holding it is not.

What the pattern implies

Anthropic’s debut mechanics—pricing above range, a 34% opening pop that halved by the close, and enormous volume—show strong allocation demand but limited willingness to pay a sustained premium once the stock was freely tradable into a weak tape. That is closer to a well-executed large-cap IPO (Arm, SpaceX) than to a scarcity-driven blowout (Figma).

For anyone underwriting the next frontier-AI listing (notably the anticipated OpenAI process), three practical lessons stand out:

  • The first print is not the valuation. Fully diluted figures near $1.5 trillion at $83 already embed aggressive assumptions on revenue scale-up against very large compute commitments. The market’s refusal to hold the $111–$115 zone on day one is an early signal that those assumptions will be tested trade by trade, not granted at the open.
  • Pop size is a poor predictor. Arm’s ~25% day-one gain preceded large multi-year gains; Figma’s +250% preceded a collapse below the offer price; CoreWeave’s flat open still produced a boom-bust. Subsequent performance has tracked operating delivery, capital intensity, and the broader AI multiple more than the opening gap.
  • Context dominates the first sessions. Anthropic listed into a multi-day equity sell-off and heightened skepticism about AI spending. A similar fundamental story in a risk-on tape would likely have produced a larger and stickier premium. Near-term trading will also be shaped by the greenshoe, any early lock-up releases, and whether AI sentiment stabilizes after the Nvidia-related shock.

The data available as of November 20, 2026 stop at one session. The useful comparison is not “will Anthropic follow Figma or Arm?” but whether revenue growth and margin trajectory can justify a valuation that, even after the fade, remains among the largest in the market. History of the names above says the first-day percentage is the least informative input to that question.


Recent Findings Supplement (November 2026)

Anthropic (ANT) completed the largest IPO in history on November 18-19, 2026, pricing at $83 per share (above the $70–$78 range) for 980 million shares and raising approximately $81.34 billion, surpassing SpaceX’s June 2026 record of $75 billion.[1][2]

It began trading on Nasdaq under ticker ANT on November 19 amid intense demand (including a concurrent $26 billion private placement from tech giants such as Google/Alphabet, Nvidia, Microsoft, and others at the IPO price). The debut featured a strong but volatile first-day performance on a broader market down day.[3]

First-day trading details (November 19, 2026):
- Opening price: $111.50 (approximately 34.3–34.34% above the $83 offer price).[1][4]
- Intraday high: Approximately $115.10 (or up to $114.99 per some reports).[4][5]
- Intraday low: Approximately $92.72–$92.76.[4][5]
- Closing price: $97.17 (up 17.1% from the offer price).[1][2]
- Volume: Roughly 846–852 million shares traded (making it the most active Nasdaq name that day and exceeding SpaceX’s first-day volume of 522 million shares).[4][5]
- Market cap at close: Approximately $1.27–1.28 trillion (based on post-offering and placement shares outstanding of roughly 13.22 billion; fully diluted valuations were cited higher, around $1.6 trillion or more at points during the day, with the open briefly implying near or above $2 trillion).[1][6]

The stock opened with a massive pop and high-volume first trade (one print reportedly involved ~108 million shares worth ~$12 billion), peaked early, then retraced significantly in the afternoon while holding well above the offer price.[7]

Trading trajectory since debut (as of latest available data on or around November 19–20, 2026): Too early for a full post-IPO arc, but the pattern aligns with other large 2026 AI/tech debuts—strong initial enthusiasm followed by partial giveback amid broader market caution over AI spending, costs, and interest rates. One post-close snapshot showed shares around $94–$97 in after-hours or subsequent indications.[6]

Comparisons to recent large tech IPOs on first-day pop and performance (focusing on 2025–2026 examples):
- Cerebras Systems (May 2026, AI chipmaker): Opened ~89% above its $185 offer price after raising $5.55 billion—one of the strongest pops among recent deals.[8]
- SpaceX (June 2026): More modest 11.1% open above $135 offer after its record $75 billion raise at the time.[8]
- Figma (July 2025): Priced around $33; first-day close reportedly ~$115.50 (massive pop exceeding 200–250% in some accounts), but has since declined sharply to around $16 by mid-November 2026 (down ~85% from debut levels and well below day-1 close).[9][10]
- CoreWeave (March 2025): Priced at $40; has shown modest net gains since (trading around $44–45 recently) but with volatility and recent pressure.[11][12]

Broader context: 2026 tech/AI IPOs averaged a ~20.5% first-day pop, yet 13 of the 15 largest now trade below their day-1 closes (group average down ~25.9%).[13] Anthropic’s 34% open and 17% close represent a solid but not outlier pop for its enormous size, with the retracement fitting the post-pop cooling seen in peers. Longer-term precedents (e.g., Arm’s strong multi-year gains) vary widely by company fundamentals and market conditions.[14]

Implications for market participants and competitors: Anthropic’s record raise and debut validate intense investor appetite for frontier AI labs as a new public-market category, potentially setting a high bar (and benchmark) for anticipated 2027 listings like OpenAI while highlighting risks of volatility and valuation compression after the initial frenzy.[15][16] For entrants or rivals, the data underscores the value of strong strategic backing (e.g., the $26B placement) and the challenge of sustaining momentum amid macro and sector-specific pressures. Subsequent performance will likely hinge on execution, AI monetization progress, and broader sentiment. Data is current as of November 19–20, 2026 reporting; real-time quotes should be checked for updates.

Report 4 Calculate the valuation as a multiple of run-rate revenue and of forecast 2027 revenue, citing the source of each revenue figure (prospectus, company statements, or "publicly estimated" analyst figures). Compare with the OpenAI private round at $1.25 trillion (its implied multiples, terms, and date) and with Nvidia (market cap, revenue multiple, margins, growth). Include a comparison table and note differences in growth, gross margin, and compute costs.

Anthropic’s public valuation, set by its November 18, 2026 IPO at $83 and its first-day close, prices the company at roughly 14–19× its own October run-rate and about 8–11× bank forecasts for 2027 calendar revenue—materially cheaper on current run-rate than OpenAI’s contemporaneous private mark, and in the same neighborhood as Nvidia once Nvidia’s much higher margins and free-cash-flow conversion are ignored. The gap that matters is not the headline multiple; it is that Anthropic’s revenue is still a fraction of the compute it has already contracted to buy.

Anthropic’s valuation multiples

Anthropic priced 980 million Class A shares at $83 on November 18, 2026 (above the $70–$78 range), raising $81.3 billion in the IPO plus a concurrent $26 billion private placement from chip and cloud suppliers. On the roughly 13.22 billion Class A and Class N shares outstanding after the offering and placement, that price implied about $1.10 trillion of equity value; on a fully diluted basis (options, RSUs, and convertibles) the figure cited by Reuters and others was about $1.5 trillion. Shares closed the first session (November 19) at $97.17, a market capitalization of about $1.27–$1.28 trillion on the same outstanding-share count. [1] [2] [3] [4]

Run-rate revenue is a company figure from the prospectus, not an analyst construct. The amended S-1 states that Run-Rate Revenue—last 28 days of usage fees annualized (×13) plus a month of subscription fees annualized (×12)—was approximately $81 billion as of October 31, 2026, up $10 billion from September 30. Nine-month 2026 GAAP revenue was $33.2 billion and third-quarter revenue was about $17.0 billion. The prospectus contains no forward guidance. [5] [6] [7]

Publicly estimated 2027 revenue comes from the banks, not the company. Goldman Sachs (a lead underwriter) projected about $136 billion of 2027 revenue in materials discussed with investors; other sell-side work clustered at $135–137 billion for 2027 and $270–280 billion for 2028. Those are not prospectus forecasts. [8] [9] [10]

Valuation basis Equity value ÷ $81B Oct run-rate (prospectus) ÷ ~$136B 2027E (Goldman / sell-side)
IPO price, shares outstanding ~$1.10T ~13.6× ~8.1×
IPO price, fully diluted (reported) ~$1.5T ~18.5× ~11.0×
Nov 19 close, shares outstanding ~$1.28T ~15.8× ~9.4×

Pro forma cash, cash equivalents, and short-term investments after the IPO and placement were about $232 billion in the final prospectus. Subtracting that cash (and ignoring modest debt) drops enterprise value at the close to roughly $1.05 trillion, or about 13× the October run-rate and about 7.7× the 2027 bank forecast. Multiples above are equity values unless noted; cash-rich balance sheets make EV the cleaner comparison to Nvidia. [11]

OpenAI’s $1.25 trillion private round

OpenAI’s latest private mark is a partial close, not a completed round and not a public filing. Bloomberg reported that roughly $11 billion had been wired by November 6, 2026, as the first tranche of a round valuing the company at $1.25 trillion, with the total potentially reaching about $20 billion. That was a step down from earlier talk of $30 billion or more at $1.4 trillion. Reported participants included Coatue, T. Rowe Price, SoftBank, Sequoia, Founders Fund, Andreessen Horowitz, and UAE-linked capital. OpenAI did not confirm the figures. Sources differ on whether $1.25 trillion is strictly pre- or post-money; the number attached to the round in the November reporting is $1.25 trillion. [12] [13] [14]

Revenue is company communication to investors, not a prospectus. OpenAI told investors its annualized revenue was approaching $50 billion at the end of September 2026 on a net basis (after partner share on some cloud sales), and that it expected to reach or exceed $70 billion annualized by year-end 2026, driven by enterprise. Earlier “grossed-up” figures near $70 billion mixed methods with Anthropic’s. There is no audited public revenue line. [15] [16]

Implied multiples on the $1.25 trillion mark:

  • About 25× the ~$50 billion end-September run-rate (company statements to investors).
  • About 18× the company’s own year-end 2026 target of ≥$70 billion annualized.
  • No equivalent bank 2027 calendar forecast is in the same public circulation as Anthropic’s $136 billion figure. Independent trackers (for example 06.ai’s base case of a ~$200 billion end-2027 run-rate) are model estimates, not company or underwriter numbers, and should not be treated as comparable to Goldman’s Anthropic case.

On the September figure OpenAI is priced richer than Anthropic’s outstanding-share IPO and close. On a year-end $70 billion target the gap narrows toward Anthropic’s fully diluted IPO multiple. The accounting difference matters: OpenAI’s disclosed run-rate is net of partner share; Anthropic’s prospectus run-rate is its own gross measure.

Nvidia

Nvidia’s market capitalization was about $3.5–3.65 trillion at the November 19, 2026 close (share price near $146, roughly 24.1 billion shares outstanding; quote services differ by a few percent on the exact cap). [17] [18]

Trailing revenue to the quarter ended October 25, 2026 is on the order of $362 billion. The just-reported Q3 FY2027 quarter was $115.8 billion, up 103% year over year, with data-center revenue of $107.6 billion. GAAP gross margin was 74.8% and operating margin 66.9%. Q4 revenue was guided to $130 billion (±3%), with gross margin guided to 71–72%. Management raised the following fiscal year’s revenue-growth outlook to about 75%. [19] [20] [21]

At ~$3.5–3.65 trillion, Nvidia trades at roughly 10× trailing twelve-month sales and in the high-single-digits on the fiscal year that includes the $130 billion quarter (management’s path implies fiscal 2027 revenue near $420 billion if the guide is met). That is a lower revenue multiple than either lab on current run-rate, earned with mid-70s gross margins, mid-60s operating margins, and tens of billions of quarterly operating cash flow—none of which the model labs currently produce.

Comparison

Anthropic OpenAI Nvidia
Valuation ~$1.10T IPO outstanding / ~$1.5T fully diluted IPO / ~$1.28T Nov 19 close $1.25T private mark (partial close ~Nov 6, 2026; round up to ~$20B) ~$3.5–3.65T market cap (Nov 19, 2026)
Revenue base used $81B Oct 31 run-rate (prospectus) ~$50B end-Sep annualized, net (company to investors); ≥$70B targeted by end-2026 ~$362B TTM; Q3 $115.8B
Multiple on that base ~14–19× equity on run-rate (lower on EV after ~$232B cash) ~25× on $50B; ~18× on $70B target ~10× TTM sales
2027 revenue ~$136B publicly estimated (Goldman / sell-side); prospectus silent No comparable public bank forecast cited Path consistent with ~75% growth into the next fiscal year (company outlook)
Multiple on 2027E ~8–11× equity (~8× EV at the close) Not cleanly computable from public bank figures High-single-digit on near-term forward revenue
Recent growth Q3 revenue ~12× year earlier; run-rate +$10B in October alone Company cited ~70% quarterly growth in annualized revenue (not a filed number) Q3 revenue +103% YoY from a far larger base
Gross / operating margin ~85% gross margin on cost of revenue (9M 2026); total compute & infrastructure ~57% of 9M revenue; Q3 GAAP operating loss after a Q2 profit Not publicly filed; still scaling losses against compute GAAP gross margin 74.8%; operating margin 66.9% (Q3 FY2027)
Compute cost structure ~$901B aggregate commitments, substantially all compute and infrastructure; ~$71B due in the 12 months after Sept 2026 Large multi-year compute contracts; dollar total not in a public filing comparable to the S-1 Supplier: captures high margin on the labs’ spend; customer concentration and custom-silicon risk are the offset

Sources for the table cells are the same as the narrative above: Anthropic prospectus and pricing reports, OpenAI funding and investor-revenue reporting, Nvidia’s November 2026 results and market quotes.

What the differences actually mean

Growth is not the scarce input. Anthropic’s third-quarter revenue was about twelve times the year-earlier quarter, and October alone added roughly $10 billion of run-rate—more than twice its entire 2025 revenue. OpenAI’s investor materials also describe very rapid quarterly expansion. Nvidia is still doubling from a base already larger than either lab’s run-rate. A revenue multiple that looks “cheap” versus OpenAI can still be expensive if the growth rate that justifies it has to be sustained while a fixed compute bill arrives.

Gross margin is the misleading comparison. Anthropic’s prospectus supports an ~85% gross margin for the first nine months of 2026 because cost of revenue is mostly customer-serving compute. Training, research, and free-tier compute sit elsewhere. Total compute and infrastructure expense was $18.8 billion on $33.2 billion of nine-month revenue (about 57%), and it rose to about 58% of revenue in the third quarter, coinciding with a swing back to an operating loss after a profitable second quarter. Short-term compute, which the company says costs significantly more than capacity under multi-year deals, was a material part of that quarter. Nvidia’s mid-70s gross margin and mid-60s operating margin are earned after the full cost of designing and selling the chips; they convert into cash. The labs’ reported gross margins do not.

Compute commitments are the binding constraint on the multiple. Anthropic has disclosed about $900.9 billion of aggregate purchase commitments and leases, substantially all compute and infrastructure, with roughly $70.7 billion falling due in the twelve months after September 30, 2026, and several hundred billion more before the end of 2030. That stack does not shrink if demand slows. The IPO and placement left the company with on the order of $230 billion of cash and short-term investments, which covers the near-term bill several times over and is why EV multiples are lower than equity multiples—but the commitments extend well past the cash. Nvidia’s economics are the mirror image: it is paid, at high margin, for a large share of that capacity. OpenAI faces the same structural issue (revenue must grow into contracted compute) without a public commitment schedule as detailed as Anthropic’s S-1.

For anyone underwriting the stock or a competing lab, the relevant question is not whether 9–16× a fast-growing run-rate is lower than OpenAI’s 25×. It is whether 2027–2028 revenue, at sustainable margins after the capacity lands, covers a commitment schedule already signed at a scale that rivals the equity value itself. Bank cases that get Anthropic to the mid-$100 billions in 2027 and toward $200 billion-plus in 2028 are what make the current price look like a mid-single-digit to high-single-digit forward revenue multiple. Those cases are publicly estimated analyst figures, not company guidance, and they assume the usage growth that has to keep filling capacity whose cost is already largely fixed.


Recent Findings Supplement (November 2026)

xAI (now integrated as SpaceXAI within SpaceX following its February 2026 all-stock merger) carries a $250 billion standalone valuation from the merger (implied value in the deal that valued the combined entity at ~$1.25 trillion). This sits at roughly 8.3× its most recent reported revenue run-rate of $30.2 billion (annualized from Q3 2026 Super Intelligence segment revenue of $7.56 billion per SpaceX's SEC 8-K filing as of September 30, 2026).[1][1]

This multiple reflects a blend of core Grok/API revenue (earlier standalone estimates around $500 million ARR mid-2026) plus significant compute-rental and infrastructure income (including deals with Anthropic, Google, and others), which has scaled the segment rapidly but introduces different economics than pure model labs.[2][3] No specific company or analyst 2027 revenue forecast for xAI/SpaceXAI appears in recent post-May 2026 sources; prior standalone targets (e.g., "north of $2 billion" for 2026) predate the merger and compute-leasing expansion.[4]

OpenAI's November 2026 funding round closed an initial ~$11 billion tranche at a $1.25 trillion pre-money valuation (with the round potentially reaching ~$20 billion total; first close wired by November 6, 2026).[5][6] This values it at ~18× on a $70 billion claimed end-2026 run-rate target or ~25× on the $50 billion net annualized run-rate reported for end-September 2026 (net of partner shares; the higher figure reflects grossing up for comparability with peers like Anthropic).[7][8][9] The round (involving SoftBank, MGX, Coatue, T. Rowe Price, etc.) came after scaling back from higher targets (~$1.4 trillion at up to $30 billion) amid AI stock volatility.[10]

Nvidia's market capitalization stood at approximately $3.5–3.75 trillion as of mid-to-late November 2026 (share price ~$146–$156). Its TTM revenue reached ~$362 billion (through October 2026), implying a revenue multiple of roughly 10×.[11][12] Q3 FY2027 (ended October 25, 2026) revenue hit $115.8 billion (+103% YoY), with data center at $107.6 billion; GAAP gross margin was 74.8%.[13][14] The company guided Q4 FY2027 revenue to $130 billion (±3%) and raised its FY2028 (fiscal year ending January 2028) revenue growth outlook to ~75% YoY (from prior ~70%).[15]

Comparison table (using latest consistent post-May 2026 figures; multiples are valuation/market cap divided by the cited revenue run-rate or TTM):

  • xAI/SpaceXAI: Valuation $250B (Feb 2026 merger-implied); Run-rate revenue $30.2B (Q3 2026 annualized, SpaceX 8-K); Multiple ~8.3×; 2027 forecast: N/A in sources.
  • OpenAI: Valuation $1.25T (Nov 2026 round); Run-rate revenue $50B net (end-Sep 2026) or $70B target (end-2026); Multiple ~25× or ~18×; 2027 forecast: Limited data; trajectory from $50B run-rate suggests potential further growth but no explicit figure.
  • Nvidia: Market cap ~$3.5–3.75T (Nov 2026); TTM revenue ~$362B; Multiple ~10×; 2027/ FY2028 outlook: ~75% YoY growth (raised guidance).

Key differences in growth, gross margins, and compute costs: xAI's lower multiple stems partly from its hybrid model (product + large-scale compute rentals/infrastructure, now embedded in public SpaceX disclosures), contrasting with pure-play labs like OpenAI that command higher multiples on consumer/enterprise model revenue despite heavy losses.[16] OpenAI and Anthropic show explosive growth (e.g., OpenAI Q3 quarterly revenue reportedly >$10B, up ~50% QoQ in some reports) but face high compute commitments and negative or low margins; Anthropic's S-1 highlighted improving unit economics (revenue per compute dollar rising sharply) yet massive future infrastructure spend.[17][18] Nvidia enjoys superior ~75% gross margins as a chip seller (vs. the compute buyers' high costs), with sustained high growth from AI demand and strong cash generation.[13]

For competitors or entrants: xAI's integration into SpaceX provides capital and infrastructure advantages (e.g., compute leasing revenue stream) that pure startups lack, but it trades at a discount to OpenAI's valuation multiple. Nvidia's position as the margin-rich enabler of the ecosystem highlights why hardware moats differ from application/model layers. New entrants should note the sensitivity of multiples to accounting (net vs. gross revenue) and the capital intensity of scaling, where compute rental or efficiency gains can alter economics faster than model performance alone. Data is drawn from SEC filings, Bloomberg/Reuters reporting, and company disclosures post-May 2026; 2027 forecasts remain sparse outside Nvidia's guidance.

Report 5 Identify the largest shareholders after the offering (Amazon, Google, employees and founders, the Long-Term Benefit Trust, VCs and other investors), with percentage stakes and voting power. Explain the Long-Term Benefit Trust's governance rights, any dual-class or special share structure, and how they affect public investors. Detail the lock-up terms: duration, staged or early-release triggers, exceptions, and the dates of potential supply overhang.

Anthropic’s November 2026 IPO (Nasdaq: ANT) split economics from control. Amazon and Google are the largest economic holders, almost entirely in non-voting stock. Voting power and the board sit with a founder LLC and the Long-Term Benefit Trust, neither of which has meaningful economic rights. Public Class A buyers get a pro-rata claim on profits and a structurally capped, minority vote.

Largest holders after the offering

The final prospectus (Form 424B4, priced November 18, 2026 at $83) puts Amazon and Google at 14.6% and 13.4% of outstanding capital stock, all in non-voting Class N. Together that is about 28% of the company and zero general votes. The deal was upsized late—from 825 million Class A shares marketed at a $70–$78 range to 980 million shares at $83—so these percentages are slightly below the 14.7% / 13.7% figures in the November 17 S-1/A, which assumed a $74 midpoint and a smaller deal.

Holder Class Economic stake after the offering Voting power
Amazon and affiliates Class N 1,936,452,353 shares; 14.6% of capital stock None, until sold to an unaffiliated buyer (then converts to Class A)
Google / Alphabet Class N 1,774,959,745 shares; 13.4% None, same conversion rule
Seven co-founders Class A, plus one Class F share via Founder LLC About 12.7–12.8% of capital stock on the pre-upsize count (1,673,460,594 Class A shares after each sold 3,378,378 shares); each still holds roughly 237–241 million Class A shares ~11.2% through their own Class A, 61.3% including the Class F share (filing figure on the pre-upsize base; Class F itself stays fixed at 50.1%)
Anthropic Public Benefit Fund (wholly owned subsidiary—not the Trust) Class N 846,840,960 shares; about 6.4–6.5% of legal capital stock (5% of fully diluted shares as of June 30, 2026) None
IPO buyers Class A 980 million shares; about 7.4% of ~13.22 billion shares outstanding A slice of the Class A pool, which as a class is capped at 49.9% of the general vote
Other strategic placement buyers (NVIDIA, AMD, Broadcom, Micron, Microsoft, TSMC Partners) Class N Remainder of the $26 billion placement after Google’s $10 billion; on the order of 1.5% None
Employees, VCs, and other investors Mostly Class A (some preferred that converted) The residual, on the order of ~44% of legal capital stock after the groups above Whatever Class A they hold, inside the 49.9% class cap

Share counts for Amazon and Google, and the Class A / Class N totals (7,583,103,996 and 5,636,538,278), are from the final prospectus as reported from the 424B4. Founder share counts and the 11.2% / 61.3% voting split are from Amendment No. 3 (November 17), which used 13,086,672,591 shares outstanding at the $74 midpoint; the upsize to ~13.22 billion shares dilutes those economic percentages slightly and does not change the Class F 50.1%.

Amazon’s figure is capped. It elected not to waive an ownership cap on its convertible notes. Full conversion without the cap would take Amazon to about 3.48 billion Class N shares, or roughly 23.6–23.8% of capital stock. Google’s 2026 agreement stops new issuance that would push it above 14.99% on a post-money, as-converted basis unless it waives that cap. Google’s 13.4% already includes its $10 billion slice of the concurrent private placement.

No venture firm is a greater-than-5% holder of voting securities. Spark Capital is the named venture presence because general partner Yasmin Razavi sits on the board. Earlier rounds (Lightspeed, ICONIQ, Fidelity, GIC, Coatue, Altimeter, Dragoneer, Greenoaks, Sequoia, and others) sit inside the residual with employees and other financial investors. Employee economics are mostly unissued or unvested: as of September 30, 2026 the company had options on about 1.77 billion Class A shares at a $2.91 weighted-average exercise price, plus more than a billion shares underlying RSUs. Those are not all in the “outstanding” count, so employee ownership of legally outstanding stock is much smaller than the option/RSU pool. A secondary breakdown circulating on November 17 put venture and other investors near 34% and employees and others near 4% of the pre-upsize count; that split is not in the filing’s principal-stockholder table and should be treated as approximate.

The Public Benefit Fund is easy to confuse with the Long-Term Benefit Trust. The Fund is a wholly owned LLC that holds Class N shares to finance public-benefit work. The Trust holds one Class T share, no economics, and the board-election right described below. Co-founders have each pledged 80% of personal equity to philanthropy; the filing describes those pledges as voluntary and revocable, not binding obligations.

How the Long-Term Benefit Trust and the share classes actually work

Anthropic authorized five classes. Only Class A and Class N have real economic rights, and they are equal per share on dividends and liquidation. Control is in two single shares that have essentially no claim on profits.

  • Class T (one share), Long-Term Benefit Trust. The Trust, set up in 2023, has five financially disinterested trustees who hold no Anthropic equity. Through Class T it elects and can remove the Class T directors, who are a majority of the authorized board (four of seven). Named Class T directors in the filing include Reed Hastings (lead independent), Vas Narasimhan, and Christopher Liddell, with one seat vacant. Removing the designated “Specified Class T Director” without cause needs nominating-committee concurrence or a Class E vote. The Trust does not vote on ordinary stockholder matters and takes no dividends.
  • Class F (one share), Founder LLC. Issued immediately before the IPO to an LLC whose members are the seven co-founders (Dario Amodei, Daniela Amodei, Tom Brown, Jack Clark, Jared Kaplan, Sam McCandlish, Christopher Olah). It carries a fixed 50.1% of total voting power on matters submitted to stockholders generally, including election of the three directors the Trust does not appoint. Issuing more Class A or Class N does not dilute that 50.1%. Inside the LLC each founder has one vote and a simple majority decides how the share is cast. The share has no economic rights.
  • Class A. One vote per share. Sold in the IPO and also held by founders, employees, and financial investors. As a class, Class A is capped at 49.9% of general voting power for as long as Class F exists. The filing states that Class A holders generally cannot affect the outcome of any matter submitted to stockholders or the election of directors, regardless of how they vote.
  • Class N. No vote except where Delaware law requires a class vote. Held primarily by strategic partners (Amazon, Google, the placement buyers, the Public Benefit Fund). It converts automatically into Class A on a transfer to an unaffiliated third party. A large strategic sale therefore does two things at once: it adds supply, and it adds votes inside the already-capped Class A pool, diluting existing Class A holders’ share of that 49.9%.
  • Class E. Reserved for employees other than the co-founders. Votes only on limited matters (including certain changes to the Trust arrangement). No economic rights. None outstanding immediately after the offering.

Together, the Trust and the Founder LLC can seat the entire board. Class F plus Class A, voting together, elect the other three directors (initially Dario Amodei, Daniela Amodei, and Yasmin Razavi). That structure makes Anthropic a Nasdaq “controlled company,” with the related exemptions from some board-independence rules. The company has said it intends to use the exemption only for the nominating and governance committee.

The arrangement is hard to unwind and not permanent. Class F sunsets if fewer than three continuing founders remain in the LLC for 90 consecutive days, after a transition period, at which point Class F and Class E are redeemed for no consideration. Amending or redeeming Class T requires high thresholds: with the trustees, 80% of authorized directors and 80% of trustees in office; without them, founder approval (unanimous in the first three years, or two-thirds thereafter) plus Class E and Class A votes; if no Class F remains, 85% of outstanding Class A. Directors of the public benefit corporation must balance stockholder financial interests against the stated purpose—“the responsible development and maintenance of advanced AI for the long-term benefit of humanity”—even if control of Class T changed hands.

What this means for public investors

A Class A share bought in the IPO has the same per-share economic claim as Amazon’s or Google’s Class N share. It does not have a comparable governance claim.

Public holders cannot outvote the Founder LLC on ordinary matters, cannot elect a majority of the board, and cannot, by themselves, remove the Trust’s directors. Their aggregate ceiling is 49.9%, and that 49.9% is shared with founders, employees, and VCs who already hold Class A. On the post-deal Class A count of about 7.58 billion, the 980 million IPO shares are roughly 13% of Class A, which translates to on the order of 6% of total general voting power—useful for a class vote where the law requires one, not for control.

Three further dilutions sit on top of that. First, Class N converts to voting Class A when Amazon, Google, or other strategics sell to unaffiliated buyers, so the public’s share of the 49.9% shrinks exactly when those holders exit. Second, the Public Benefit Fund can receive up to another 2% of the company, at board discretion, each time value rises 50% above a baseline that starts at the IPO price (with trading-day tests). Third, the company is a PBC: the board is required to weigh the public-benefit purpose against returns, and further share contributions to that purpose dilute everyone economically, including the strategics.

The commercial overlay matters as much as the cap table. Amazon and Google are simultaneously large shareholders, the main compute suppliers, and distributors of Claude. The prospectus discloses minimum infrastructure commitments on the order of $110 billion to Amazon (through 2036) and about $121 billion to Google (through 2033), inside a broader ~$901 billion of committed spend, mostly compute. Their equity is capped and non-voting; their commercial contracts are not. A public holder is long the residual after those bills.

Lock-up: duration, stages, exceptions, and the supply calendar

About 15.5 billion share equivalents are under market-standoff or lock-up agreements. The base term is up to 180 days after the prospectus date (November 18, 2026). Co-founders and stockholders who held more than 0.5% of fully diluted shares before the offering are locked for up to 365 days. Both clocks can end earlier, on the second full trading day after a specified earnings release. There is no price-based early release. Goldman Sachs and Morgan Stanley, as representatives of the underwriters, may release any of these securities in whole or in part at any time; a release for some holders can trigger a pro-rata release for certain others.

The private-placement Class N shares ($26 billion; about 313 million shares at $83) are not subject to the lock-up. They are restricted securities. A non-affiliate can resell under Rule 144 only after a six-month holding period once Anthropic has been a reporting company for 90 days, which points to roughly late May 2027, near the day-180 gate. Google also has separate registration rights, contractually delayed until the earlier of two trading days after Q1 2027 earnings or day 180.

Four holder tracks, from the lock-up exhibit and the “shares eligible for future sale” section:

  • Non-founder employees, officers, directors, and consultants. Up to 10% from the first trading day; a further 35% on the second full trading day after FY2026 results; the rest at the earlier of Q1 2027 results or day 180 (May 17, 2027 at the latest).
  • Other stockholders (about 2.7 billion shares in the schedule summaries). None on day one. Up to 6% on each of day 27 and day 48, 7% on each of day 62 and day 76, then a large block after FY2026 results, fully free by the Q1 2027 / day-180 gate.
  • Holders above 0.5% (about 8.0 billion shares). Smaller steps on the same calendar—on the order of 1.8–2.85% early, about 26% cumulative before day 180, plus a further slice at the 180-day gate—and fully free only at the earlier of Q3 2027 results or day 365. From day 195 (around June 1, 2027) some of them may hedge up to 20% of still-locked stock.
  • The seven co-founders. Nothing until the Q1 2027 / day-180 gate, when each may sell up to 29% (with three business days’ notice to the company and the underwriters). The rest stays locked until the earlier of Q3 2027 results or November 18, 2027. Their IPO sales—3,378,378 shares each, 23.6 million in total, about $280 million apiece at $83—already happened and are not part of the lock-up.

Potential supply overhang, using the final-amendment schedule as reconstructed against a November 18 prospectus date. Counts are maximums that become eligible, not forecasts of what will be sold. Dollar figures are at the $83 IPO price.

Window Date Shares that can become eligible Value at $83
First trading day (employee/officer 10%) November 19, 2026 Up to 315 million ~$26 billion
Day 27 December 15, 2026 Up to ~303 million ~$25 billion
Day 48 January 5, 2027 Up to ~303 million ~$25 billion
Day 62 January 19, 2027 Up to ~353 million ~$29 billion
Day 76 February 2, 2027 Up to ~353 million ~$29 billion
Cumulative through day 76 ~1.6 billion ~$135 billion
Second trading day after FY2026 results, then five steps every 14 days Not set; 10-K deadline for a non-accelerated filer is March 31, 2027 ~1.7 billion, then ~480 million × 5 ~$141 billion, then ~$40 billion each
Q1 2027 results or day 180, whichever is earlier May 17, 2027 at the latest ~3.0 billion; employees fully free; founders may sell up to 29% ~$249 billion
After Q1, then after Q2 Mid to late 2027, earnings-dated Further ~1.8 billion, then ~3.1 billion in steps Hundreds of billions at $83
Q3 2027 results or day 365 November 18, 2027 at the latest Last ~1.9 billion, including the rest of founder and large-holder stock ~$158 billion

By early February 2027, eligible legacy supply can already be about twice the 980 million-share offering. The single largest wave is tied to FY2026 results, not to a calendar cliff. Day-one float is the 980 million IPO shares plus up to 315 million employee shares, roughly a tenth of the company, against a scheduled path that opens most of the 15.5 billion locked equivalents by November 2027.

For someone underwriting the stock, the binding constraints are not the Amazon and Google percentages. They are the 50.1% founder vote, the Trust’s board majority, the Class N conversion mechanic, and a lock-up that puts employee supply into the market on day one and keeps the largest economic holders—and the founders—on a one-year staged release rather than a single 180-day cliff.


Recent Findings Supplement (November 2026)

Anthropic completed its IPO on November 18, 2026 (priced at $83 per Class A share; trading began November 19 under ticker ANT; expected close November 20), with detailed ownership, voting, board control, and lock-up structures disclosed in its final S-1/A filings (November 17–19, 2026).[1][2]

These disclosures provide the first public post-offering breakdown and supersede prior private-company information.

Post-Offering Ownership Stakes

Amazon holds the largest stake at approximately 14.7% (1,920,511,218 Class N shares). Google holds about 13.7% (1,789,598,269 Class N shares, including a portion from a concurrent $26 billion private placement at the IPO price involving Google and other strategic partners such as Nvidia and Microsoft).[3][4]

The seven co-founders collectively hold roughly 12.8% (about 1.67 billion Class A shares after each sold ~3.38 million shares in the IPO, for a combined ~$1.96 billion). Each founder beneficially owns roughly 3.1–3.2% of Class A shares (individual holdings in the 232–245 million range post-split).[3][5]

The Anthropic Public Benefit Fund holds about 6.5% (846.8 million Class N shares). Public/IPO buyers received the marketed shares (part of the ~980 million Class A shares sold in the offering). Employees, VCs, and other investors hold the balance, with employees controlling a large but fragmented pool now subject to staged unlocks.[6]

Implication for competitors/entrants: Amazon and Google together control ~28.4% economically (non-voting), creating deep commercial ties but no direct governance influence. Founders retain significant economic skin in the game alongside structural control.

Voting Power and Share Classes

Class A shares (publicly traded) carry one vote each but are capped at 49.9% of total voting power as a class. Class N shares (primarily Amazon and Google) have full economic rights but no general voting rights.[6][7]

A single Class F share (held by the co-founders’ Founder LLC) carries a fixed 50.1% of votes on general stockholder matters (including election of the three non-Class T directors), which cannot be diluted. This gives the co-founders ~61.3% voting power overall when combined with their Class A holdings.[8][3]

A single Class T share (held by the Long-Term Benefit Trust) elects and removes the four Class T directors, forming a board majority (out of seven total directors). The company qualifies as a Nasdaq “controlled company,” exempting it from certain independence and committee requirements (it expects to use the exemption only for the nominating/governance committee).[1][9]

Implication: Public Class A investors have meaningful but minority voting power and cannot elect a board majority or easily influence mission-related decisions. The structure prioritizes long-term control by founders (via votes) and the Trust (via board seats) over standard shareholder democracy.

Long-Term Benefit Trust Governance Rights

The Trust (established 2023) holds the Class T share with no economic rights or claim on profits. Its five independent trustees (e.g., chair Neil “Buddy” Shah, Ben Bernanke, and others; one seat vacant as of the filing) elect/remove the four Class T directors. Trustees receive a fixed $400,000 annual cash payment (independent of company performance) and are barred from holding Anthropic stock.[1][1]

The Trust can vote on limited matters (e.g., redemption of Class E shares or certain amendments to Class F rights) but has no general stockholder vote except as required by law. It was designed to balance profit, stakeholder impact, and Anthropic’s public-benefit purpose (“responsible development and maintenance of advanced AI for the long-term benefit of humanity”). Amendments to Trust terms or Class T rights require high thresholds (typically 80% of board + trustees, or founders’ Class F consent plus other approvals).[9][10]

Implication: The Trust insulates board control from short-term investor pressure or founder turnover, but changes are difficult. Public investors cannot easily sue or redirect the company toward pure profit maximization.

Lock-Up Terms and Supply Overhang

Lock-ups cover ~15.5 billion share equivalents and use a staggered schedule rather than a single cliff (detailed in the November 17 S-1/A and underwriting agreement).[8][11]

  • Non-founder employees, officers, directors, and consultants: Up to 10% (~315 million shares, ~$26 billion at $83) eligible starting first trading day (November 19, 2026); additional 35% after FY2026 results release; remainder by mid-May 2027 (earlier of Q1 2027 results or day 180).[2][12]
  • Other stockholders (~2.7 billion shares): Staged releases beginning December 15, 2026 (e.g., 6% on day 27, additional tranches on days 48/62/76), with larger portions after FY2026 results.
  • Holders of >0.5% and co-founders: Extended lock-up up to 365 days (through ~November 18, 2027), with partial releases (e.g., co-founders see ~29% eligible after Q1 2027 results or day 180).[12]

Underwriters can waive restrictions early. Private placement shares are not locked up. Potential overhang dates include November 19, 2026 (initial employee tranche); December 2026–February 2027 (multiple staged releases totaling over 1.6 billion shares); major waves after FY2026 results (early 2027, >1.7 billion shares); and further releases through May–November 2027.[13]

Implication: Significant early supply from employees (starting immediately) and staged releases through 2027 could pressure the stock, especially around earnings dates. Founders and large holders face the longest restrictions, limiting near-term overhang from them.

Report 6 Gather analyst, investor, and commentator views on whether the IPO price is justified. Cover bull arguments (growth, enterprise share, margins) and bear arguments (compute spending commitments, customer concentration, dependence on cloud partners, competition, regulatory and safety risk, valuation versus the AI bubble narrative, governance limits). Highlight evidence that would show the pricing was too high or the reported figures unreliable, and identify historical analogues of IPO overpricing. Note which claims come from named sources and which are speculative.

Anthropic priced the largest IPO on record at $83 a share on November 18, 2026—above a $70–$78 range—and public investors paid up for growth that is real on the page and still unproven against nearly $901 billion of mostly non-cancelable compute bills. Named bulls treat the price as a discount to hyperscaler-like multiples on extraordinary enterprise growth. Named skeptics treat it as a pricing event that already embeds a multi-year revenue miracle, with public shareholders holding little control if that miracle slips.

What the market actually paid

The book was multiple times oversubscribed, so Anthropic and its co-founders sold 980 million shares at $83, raising $81.3 billion and implying roughly $1.5 trillion fully diluted (about $1.1 trillion on a non-diluted count near 13.2 billion shares). A concurrent private placement of about $26 billion to chip and cloud partners (including Google, Microsoft, Nvidia, AMD, Broadcom, Micron, and TSMC) took total capital raised to about $107 billion. That topped SpaceX’s June 2026 base deal. Demand was strong enough to price above the range even after earlier talk of a $2 trillion valuation had already been cut amid an AI-stock selloff. [1] [2] [3] [4]

On November 19 the stock opened at $111.50 (+34%), traded as high as about $115 and as low as about $92.72, and closed at $97.17, up 17% from the offer. Jay Ritter (University of Florida) called both the pop and the volatility unsurprising given how far opinions diverge on future profitability. Morningstar’s Malik Ahmed Khan kept a $92 fair-value estimate and called the close roughly fairly valued. The fade from the open is the first public signal that the offer price was not obviously “too cheap,” even if IPO allocations still made money. [5] [6] [7]

Implication: Day-one buyers who paid the open paid a full ~$2 trillion peak valuation for a business whose own prospectus says growth will not continue at recent rates. Allocations at $83 were the scarce asset; the open market was already discriminating.

Bull case: growth, enterprise share, and margins

The growth numbers in the S-1 are not in serious dispute. Revenue was $41 million in 2023, $386 million in 2024, and $4.6 billion in 2025. For the nine months ended September 30, 2026, revenue was $33.2 billion, up roughly 1,200% year over year. Third-quarter revenue was $17.0 billion, up 1,127% year over year. The company reported an October 31 annualized run rate of about $81 billion, up $10 billion from September 30. More than 11,900 customers generated at least $100,000 of trailing-twelve-month revenue (versus more than 950 a year earlier); more than 1,900 were above $1 million and about 250 above $10 million. Net revenue retention was reported around 900%. Top-five customer share fell from about 36% in the March quarter to about 21% in the September quarter (about 25% for the nine months); no single customer exceeded 10% in the nine months. About 79% of nine-month revenue was consumption-based and 21% subscriptions. [8] [9] [10] [3]

Margins improved as serving costs were spread over that volume. Footnote cost of revenue of $4.8 billion against $33.2 billion of nine-month revenue implies an ~85.5% gross margin, up from ~29% a year earlier—though Anthropic does not present gross margin on the face of the income statement, and model-training compute sits largely in R&D rather than cost of revenue (The Information and secondary analyses of the filing). Adjusted operating margin reached 19% in Q2 2026 and 15% in Q3; nine-month adjusted operating income was about $4.1 billion (12% margin). GAAP operating results were weaker: a Q2 profit, a Q3 loss of $567 million, and a nine-month operating loss of about $1.3 billion. The company has guided toward long-term compute at 45–50% of revenue and adjusted operating margins around 35%. William Blair’s Arjun Bhatia highlighted that cash burn from 2023–2025 was only about $10 billion and that free cash flow in the first nine months of 2026 was positive (about $3.6 billion on his reading; operating cash flow in the filing is about $6.6 billion). [11] [12] [13] [14] [15]

Named sell-side initiations treated that combination as underpriced relative to growth:

  • Melius Research’s Ben Reitzes: Buy, $155 (about 87% above the $83 offer), arguing Anthropic may hold up better than the broader AI trade. [1]
  • Needham: Buy, $150, placing it in the same “dominant bucket” as Amazon, Google, and Microsoft. [1]
  • Wolfe Research: Outperform, $135; Citizens: Market Outperform, $125. Both, with Melius, look for revenue above $400 billion by 2029. [1] [16]
  • New Street Research’s Pierre Ferragu: Buy, $105; roughly 8.5 times next year’s sales versus hyperscalers, addressing a claimed $36 trillion knowledge-work market, with a path above $500 billion of revenue by 2030. [17] [18]
  • William Blair’s Arjun Bhatia: at the marketed range, about 7–8 times his 2027 revenue estimate while growing ~150%, versus a peer median near 11 times on ~20% growth; a 17–20 times revenue multiple would not be unreasonable in his view ($130–$150). [11] [19]
  • Morningstar: fair value $92 ($1.7 trillion equity), so the $70–$78 range was undervalued; bull case $188, bear case as low as $25 in one write-up and $500 billion in a podcast discussion. [20] [21]
  • Scott Galloway (Prof G Markets): at about $1.5 trillion, roughly 19 times the $81 billion run rate, “about where Broadcom and AMD trade,” while Anthropic’s nine-month growth (~1,200%) dwarfs those suppliers’ latest quarterly growth. [22]
  • Battery Ventures’ Dharmesh Thakker: demand and the revenue figures look fairly priced; the excitement is evidence the market is real. [23]

Implication for anyone competing with or against ANT: The bull mechanism is not “AI is big.” It is enterprise usage that compounds inside existing accounts (900% NRR) while inference economics improve. A rival that cannot show both usage expansion and a falling compute-to-revenue ratio will not get this multiple, even with a strong model.

Bear case: the bill, the partners, the vote, and the bubble

The load-bearing risk is a duration mismatch. As of and after September 30, 2026, Anthropic disclosed about $900.9 billion of purchase commitments and leases, substantially all for compute and infrastructure, running through 2045. About $70.7 billion is due within twelve months and roughly $452 billion before the end of 2030. Named minimums include at least $121.3 billion with Google (through 2033, regardless of utilization in the reported terms), at least $110 billion with Amazon, and about $47–$49 billion with Microsoft. Revenue is mostly consumption-based; the filing warns that many large customers have no long-term contracts and can cut or stop spending at any time. The data-center counterparty cannot. Jim Cramer framed the IPO as a bet that revenue keeps outrunning an exploding cash compute bill. Goldman Sachs, in secondary coverage of the filing, was cited as not expecting positive free cash flow until 2030. [24] [25] [15] [26] [22]

Dependence on cloud partners is structural, not incidental. A large share of enterprise distribution runs through Amazon and Google marketplaces (one secondary reading put marketplace-related revenue near 44% of nine-month sales). Those same firms are suppliers, convertible-note holders, and now Class N equity buyers in the $26 billion placement. Trefis noted the 1999–2000 pattern in which equipment vendors financed customers who bought their gear; when customers failed, vendor sales and loans failed together (Cisco fell 89% from its 2000 peak). The differences now are real—Anthropic has actual, fast-growing revenue, and the placement is equity at the IPO price, not a loan—but the circularity is still the mechanism: partners fund capacity they also sell. [27] [9] [28]

Competition and the AI-bubble frame sit on top of that. OpenAI remains private (CFO commentary has pointed to 2027) at a recent primary valuation around $1.25 trillion, so it does not face the same quarterly scrutiny. Axios’s Dan Primack summarized the bear case as enterprise budgets that may have overshot, open-source alternatives, the $900 billion obligations, and a rival that can still operate without public-market glare. New Constructs CEO David Trainer—known for the pre-IPO WeWork warning—called this “the most ridiculous IPO of 2026,” said the firm does not have a viable business once open-source models are in the market, and put value nearer $150 billion, on the order of 90% below the debut target. That note predates the final price and used earlier loss figures; it is a named, extreme bear, not consensus. [4] [29] [30] [31]

Regulatory and safety risk is in the company’s own words. The prospectus warns that models could pose “catastrophic or existential risks to humanity” if misused or inadequately secured, and that each new Claude version accelerates the next. Days before the IPO, Anthropic activated part of its responsible scaling policy. Safety researchers, including a statement associated with the Center for AI Safety, argued the company is racing toward the recursive self-improvement it itself flags. The Atlantic’s Rogé Karma called the investor pitch dependent on the strategy the company has said could threaten humanity. A March 2026 Trump administration “supply chain risk” designation was later limited by courts, but it is a reminder that political risk is not hypothetical. Damodaran builds a regulatory drag into target margins (he raised that drag after the U.S. midterms). [32] [33] [34] [4] [35]

Governance limits are explicit. Seven co-founders, through a Founder LLC and a single Class F share, hold 50.1% of voting power on key matters, including election of some directors. The Long-Term Benefit Trust, via Class T stock, elects a majority of the board. Class A public shareholders share in economics but not effective control. Anthropic is a public benefit corporation and a Nasdaq “controlled company,” which it is using to seat president Daniela Amodei on the nominating and governance committee. Morningstar stated plainly that IPO buyers cede voting power to the founders and that PBC status may limit pure shareholder-return maximization. Damodaran called the structure self-serving and said he would sit out rather than short a price well above his estimate. [36] [37] [38] [39]

Valuation versus the bubble narrative is where named sources diverge most. Damodaran’s published base case moved from about $1.475 trillion to about $1.45 trillion after Q3 figures, then to about $1.26 trillion after he cut sales-to-capital and raised the regulatory margin drag—roughly $69 a share on one fully diluted count, just under the bottom of the marketed range, and below the $83 clearing price. His worst case (about $200 billion of 2036 revenue, 10% margins, harsh regulation) is negative equity; his best case is about $4 trillion. Crowd inputs to his spreadsheet clustered nearer $1 trillion than $1.5 trillion. An independent rebuild on his framework (Indmoney) landed near $580 billion under stricter operating assumptions—that is the author’s alternative, not Damodaran’s number. A probability-weighted model published in Chinese (Techub) put value near $53 a share and rated the $70–$78 range a sell. Investing.com’s model aggregator flagged about 19% downside to fair value at $83 before the open. Julius Baer’s Chinello treated the cut from $2 trillion talk to the filed range as markets discriminating between growth and returns, not as proof the “AI wheel is broken,” and as a benchmark for an expected OpenAI listing in 2027. [40] [35] [41] [42] [43] [44]

Implication: A competitor does not need to “beat Claude” to hurt the equity story. It needs either (a) enough open or cheaper closed capacity that enterprise token prices fall faster than volume rises, or (b) a pause in consumption that leaves Anthropic paying for unused gigawatts. Governance means public holders cannot force a slowdown if management chooses the race.

What would show the price was too high—or the figures unreliable

These are observable tests, not forecasts. Named sources already flag several of them.

Multiple compression on recognized revenue, not run rate. At $83, Trefis put the company at about 31 times the $35.3 billion of sales booked over the last four quarters, versus roughly 12–19 times the $81 billion October run rate depending on share count. The filing defines run rate as an operating metric (trailing 28 days of consumption annualized, plus subscriptions)—explicitly not GAAP revenue and not a forecast. A quarter that annualizes well below the October pace would reprice the “19x suppliers” comparison Galloway used. Trefis’s rule of thumb: every giant U.S. listing it studied that priced above 20 times sales fell at least 26% below its first close within a year (median low about 34% in the above-20x group). That is Trefis’s historical screen, not a prediction, and the sample of true giants is small. [45] [9] [46]

Cash versus the next twelve months of bills. Nine-month operating cash flow was about $6.6 billion against $70.7 billion of commitments due within a year. Cash and short-term investments were about $132 billion at September 30, before IPO proceeds. If utilization lags and short-term compute stays expensive—the company already blamed $5 billion of short-term compute for compressing Q3 margins—the “capital-efficient” roadshow slide stops matching the cash statement. [25] [47] [14]

Adjusted versus GAAP. The $104 billion nine-month net loss is mostly non-cash fair-value marks on convertibles, SAFEs, and derivatives as the private valuation rose (about $94 billion of that line). Ignoring it is correct for cash. Ignoring dilution is not: Amazon’s notes alone convert into a very large Class N stake. Adjusted operating income adds back stock-based compensation, non-cash charitable share donations, and the Bartz copyright settlement. A sustained gap between 15% adjusted and negative GAAP operating margin, without a clear path to the 35% long-term target, would mark the IPO narrative as accounting-dependent. [14] [48] [49]

Customer optionality versus supplier rigidity. Top-five concentration is falling, which is bullish on its face (Trefis). The unreliable part would be a reversal: a handful of $10 million-plus accounts cutting usage while minimum-spend clauses with Google, Amazon, and others still bill. The filing already says many large customers can cease spending for any reason. [10] [50]

Circular capital that does not stick. If partner equity purchases and “funding support” tied to capacity milestones are what keeps the commitment book financeable, a partner pullback or a renegotiation would show the $901 billion figure was a capacity option dressed as a firm order. Better Markets noted press readings that a slice (on the order of $215 billion in one account) may be subject to termination rights or counterparty conditions; the large majority is still described as binding, and the Microsoft/Google/Amazon/Broadcom block is largely non-cancelable outside breach. That conditionality is the footnote to watch, not a reason to dismiss the bill. [51] [52]

Governance stress. Evidence the price ignored control would be a safety-driven slowdown, a related-party compute decision, or a charitable/mission use of cash that public holders cannot block—and a stock that reprices once that becomes visible. That is structural, disclosed, and still speculative as a near-term catalyst.

Trading path. Talley Leger (Wealth Consulting Group) said history shows retail buyers of hot tech IPOs on day one underperform broad indexes over three years, and that Anthropic is not a buy-at-any-price. The open-to-close fade (111.50 to 97.17) is mild evidence of that, not proof. A close below $83 after lock-up (Trefis flags about May 17, 2027) or after the first two public quarters would be stronger evidence the clearing price was demand-driven rather than fundamental. [53] [45]

Historical analogues of IPO overpricing

These are patterns from named research, not destiny.

  • Morningstar: of the 20 largest IPOs from 2010 to 2025, only four traded above their day-one close a year later; six were down more than 50%; the group’s median one-year return was −31%. [54]
  • Trefis: in its giant-listing screen, names priced above 20 times trailing sales each fell at least 26% below the first close within a year. Separately, seven famous U.S. listings from Google (2004) through SpaceX all traded below their first close within six months, by a median of about 25%. SpaceX itself, at a $135 offer and a strong first close, was later cited trading below that first print. [46] [55] [4]
  • Cashu Research’s sample of 19 recent venture-backed U.S. IPOs (March 2025–June 2026): 16 of 19 finished day one above the offer; by six months only five of 17 with data were still above it. Median peak was 66% above the IPO price; many later sat well below both peak and offer (Klarna, Bullish, Firefly, Cerebras among the named laggards). [56]
  • Longer winners still crashed first. Trefis’s table of famous IPOs: Amazon −94% in the dot-com bust, Google about −65% in 2007–08, Facebook −54% within four months of the 2012 IPO. The winners were bought at roughly 16–28 times trailing revenue and still required surviving a halving. Anthropic’s trailing multiple is higher than that band; its run-rate multiple is inside or near it, depending on share count. [57]
  • Vendor-financed customers, 1999–2000: Trefis’s explicit analogue for the $26 billion partner placement and the broader commitment web. Not identical—equity at the IPO price, and real revenue—but the failure mode (capacity ordered ahead of durable demand) is the same shape. [28]
  • WeWork: Trainer’s comparison, useful as a process warning (a research firm that was early on a failed listing) and weak as a business comparison. Anthropic has $17 billion quarters; WeWork did not. The analogue is “insiders need a public bid,” not “the product is offices.” [30]

A Carson Investment Research statistic circulated on X (91.7% of 35 major IPOs since 2004 popped on day one; average maximum drawdown in the first twelve months −55.4%) is secondary social distribution of a research claim, not a primary filing. Treat it as illustrative of the day-one versus year-one gap, not as a verified Anthropic input. [58]

Named claims versus speculation

Claim Status
$83 price, $81.3B raised, ~$1.5T fully diluted, $26B placement, Nov 19 open $111.50 / close $97.17 Named reporting (Reuters, company statement via multiple outlets, Forbes, CBS, Morningstar)
$33.2B nine-month revenue, $17B Q3, $81B Oct run rate, 11,900+ $100k customers, ~25% top-five, ~$900.9B commitments, ~$70.7B due in 12 months Company filing, as reported by Reuters, Forbes, Trefis, SaaStr, and others
~85.5% implied gross margin; 15% Q3 / 12% nine-month adjusted operating margin; large non-cash fair-value losses Derived from S-1 footnotes and company non-GAAP definitions; not a GAAP gross-margin line
Buy ratings and targets from Melius ($155), Needham ($150), Wolfe ($135), Citizens ($125), New Street ($105), William Blair commentary Named analysts, reported by Reuters, Livemint, CNBC/Benzinga, Investing.com
Morningstar $92 base / $188 bull; Galloway “about where Broadcom and AMD trade” Named
Damodaran base roughly $1.26–$1.48T depending on date and inputs; worst case negative; will sit out rather than short Named; inputs are his assumptions, not facts
New Constructs ~$150B and “not a viable business” Named, outlier, partly pre-final-price
Revenue above $400B by 2029 or $500B by 2030; 35% long-term margins; FCF positive only in 2030 Forecasts (brokerages, company roadshow, Goldman as cited)—speculative until printed
OpenAI listing in 2027; partner deals as a hidden backstop; existential risk as a near-term valuation driver Speculative or qualitative; safety language is in the prospectus, timing is not
“Will fall 26% within a year” Historical tendency in a small Trefis sample, not a forecast

Bottom line for pricing judgment: The $83 offer was justified in the only sense markets require—investors oversubscribed it, partners bought alongside them, and the first close held a premium. It was not justified as a low-risk claim on current earnings power. Bulls are underwriting a continuation of enterprise token demand and a further collapse in inference cost per dollar of revenue, at a multiple they say matches slower chip suppliers. Bears are underwriting a non-cancelable cost stack, cancelable customers, founder control, and a trailing sales multiple that has historically punished day-one buyers of giant listings. The next two reported quarters—recognized revenue versus the $81 billion run rate, and cash compute versus the $70.7 billion twelve-month bill—are the evidence that will decide which reading was the price and which was the story.


Recent Findings Supplement (November 2026)

Anthropic priced its record $81.3 billion IPO at $83 per share on November 18, 2026 (above the $70–$78 range), raising the largest amount ever and valuing the company at ~$1.5 trillion fully diluted; shares opened at $111.50 (up ~34%) before settling around $97–$98 (up ~17%).[1][2]

This debut tests whether frontier AI economics justify trillion-dollar multiples amid ~$900 billion in long-term compute commitments, explosive but unprofitable revenue growth, and heavy reliance on cloud partners that are also rivals.

Bull arguments center on hyper-growth in enterprise adoption and improving unit economics, with named analysts projecting massive scale.[3]

  • Q3 2026 revenue reached ~$17 billion (up sharply YoY, with 9M 2026 at $33.2 billion); ARR hit $81 billion by end-October (up $10 billion from end-September); 11,900 customers spending ≥$100k TTM as of Sep 30 (vs. ~1,500 end-2025), with net revenue retention ~900%.[2][4]
  • New Street Research (Pierre Ferragu) initiated with Buy/$105 target, forecasting >$500 billion annual revenue by 2030 and comparing to early cloud leaders at ~8.5x next-year sales.[5]
  • William Blair’s Arjun Bhatia called $1.5 trillion attractive at 7–8x 2027 revenue (with ~150% growth) and sees potential for $1 trillion revenue by 2035.[5]
  • Morningstar sees narrow-moat fair value at $92/share (~$1.7 trillion equity), with bull case $188 on rapid adoption and share gains vs. OpenAI.[6]
  • Ashika views even $78 as reasonable vs. AI infrastructure peers at ~10x EV/S; Dan Ives (Yorkville) highlighted strong institutional demand as a sector-positive signal.[7]

This implies that for competitors or entrants, matching Anthropic’s data moat (enterprise usage patterns driving NRR) and efficiency claims (roadshow charts on intelligence per compute) is essential; otherwise, they risk being priced out of the public market’s growth narrative.

Bear arguments focus on fixed ~$900 billion+ compute commitments against consumption-based revenue, persistent losses, and structural dependencies that could cap margins.[2]

  • Anthropic disclosed ~$900.9 billion in aggregate purchase/lease commitments (mostly compute/infrastructure through 2045 or later), with tens of billions due in the near term; H1 2026 compute spend ~$9 billion (nearly 4x YoY), Q3 another ~$5 billion, contributing to large operating/net losses (e.g., Q3 operating loss $567 million amid volatility).[2][8]
  • Aswath Damodaran valued shares at ~$77 (below $83 offer); Seeking Alpha analyses peg fair value at $46 or lower, arguing even capturing the entire addressable market or halving compute costs as % of revenue would not justify pricing.[9][10]
  • New Constructs (pre-IPO) called it potentially the “most ridiculous IPO of 2026,” estimating ~$144 billion intrinsic value due to lack of pricing power vs. open-source and unsustainable margins.[11]
  • Gary Marcus noted AI economics “don’t add up” despite utility.[12]

For market entrants, this highlights the risk of over-committing to capex without locked-in demand; any slowdown in enterprise budgets could turn take-or-pay deals into cash-flow drains that public markets penalize harshly.

Customer concentration has eased but remains a flagged risk alongside deep cloud-partner dependence (47% of sales routed via Amazon/Google in 2025, who are also investors, suppliers, and competitors).[13]

  • Top five customers ~25% of 9M 2026 revenue (no single customer >10%); two unnamed customers each ~12% of 2025 revenue; many large customers lack long-term contracts and can reduce spending.[8]
  • Prospectus warns of conflicts from reliance on a limited number of partners for distribution, billing, and capacity, plus potential financing/delivery issues from those partners.[13]

Competitors can exploit this by offering more diversified or direct channels, or by targeting segments less tied to the hyperscalers’ ecosystems.

Valuation faces skepticism tied to the broader AI bubble narrative, regulatory/safety risks, and governance concerns, with historical tech IPO analogues suggesting sharp drawdowns.[14]

  • At IPO pricing, multiples implied 76–100x+ trailing/near-term sales in some analyses—well above the >30x P/S threshold associated with prior bubbles (e.g., dot-com era names like Cisco/Amazon); average year-1 drawdown for recent tech IPOs ~55%.[14]
  • Safety/regulatory risks highlighted in founders’ letter and external critiques (e.g., “MisAnthropic Statement” coalition); governance includes significant founder influence.[15]
  • SpaceX (prior largest IPO) saw initial pop followed by pullback; broader sector jitters (Nvidia-related issues) preceded the debut but demand overrode them short-term.[12]

New entrants or investors should watch lock-up expirations (many around May 2027) and early earnings for proof of sustainable margins; historical patterns indicate that hype-driven pricing often leads to multi-year underperformance unless fundamentals accelerate dramatically.

Overall, recent analyst commentary (post-filing) shows a split: several Wall Street firms see the priced valuation as attractive or cheap on growth, while independent voices (Damodaran, Seeking Alpha contributors) view it as stretched or unjustified given commitments and losses. Claims of overpricing or unreliable figures (e.g., revenue sustainability vs. fixed costs) are backed by named sources and S-1 data rather than pure speculation. The IPO’s success or failure will likely set a benchmark for the AI sector’s public-market viability.

Report