Anthropic IPO: What the S‑1 Shows and How Buying Would Work

Anthropic has filed publicly, but the price and date are not set. The S‑1 of Oct 26 carried results to Jun 30. The amendment of Nov 2 added the third quarter, so older first‑half headlines ($16.3B revenue, $80.7B loss) are now out of date. Everything below comes from the Nov 2 filing unless stated.

Dates and valuations in the news are press reports, not filing facts. For example, Techloy, citing Fortune, reported on Oct 26 a listing in the week of Nov 23 and a target of $1.5 trillion to $2 trillion. Neither figure appears in the filing as an offer term. The price range should appear in a later amendment.

Revenue rose quickly, the operating result was close to break‑even, and the reported net loss was very large. Nine‑month and quarterly figures are unaudited; full years are audited.

Remaining performance obligations count only non‑cancelable commitments, which helps explain why the figure is small next to quarterly revenue. The filing says many top customers have no long‑term contract and can reduce spending at any time. How these contract measures differ is covered in backlog, RPO and ARR.

Split it into what the business did and what accounting revaluations did. The nine‑month loss adds up this way: operating loss $1,260M, plus interest income of $2,359M, less the $94,324M fair‑value charge, less $10,397M of other expense, less $417M of tax.

The largest line moved against the company because its value rose. Anthropic sold convertible notes that turn into stock. The filing says the charge was driven mainly by a higher fair value of those notes as its enterprise valuation increased. No cash left the company for that line.

Non‑cash does not mean costless. Three items still matter to a shareholder:

• Dilution. Convertible notes become shares, so the same company is split among more holders. The filing allows up to $77.0B more of these notes, subject to conditions.
• Stock pay. Stock‑based compensation was $3.3B in the nine months, with another $24.1B not yet expensed.
• Direction of costs. Excluding fair‑value lines does not show a steady profit. The operating result went from +$974M in Q2 to −$567M in Q3. Compute costs rose 77.5% that quarter, against 46.7% for revenue.

Operating cash flow also needs context. The +$6,569M does not count spending on equipment or prepayments for leased assets, and the filing says it expects significant net cash used in operating and investing activities for the foreseeable future.

About $895.4B in aggregate, substantially all for compute and infrastructure, spread over many years: about $447.5B is expected before Dec 31, 2030, and substantially all the remainder from 2031 to 2045. Some of it is not firm: about $213.8B is subject to termination rights or conditions before it starts. On top of the total, the filing estimates another $150B to $175B under arrangements with no fixed dollar amount. Background on who supplies this capacity is in the AI compute supply chain.

About $69.8B falls due in the 12 months after Sep 30, against $131.8B of cash and short‑term investments before any IPO proceeds. The filing also says that some compute suppliers and chip partners provide capital or credit support and also compete with Anthropic, and that it may not receive the full amount of compute it seeks, even under contract.

One vote per share, in a structure where outside holders generally cannot decide outcomes. The filing says so directly: Class A holders generally will not be able to affect the result of a stockholder vote or the election of directors, regardless of how they vote, unless a majority of the continuing founders cannot agree on how to vote Class F. In that case Class F generally votes in proportion to Class A.

Anthropic is also a public benefit corporation, so its directors must balance shareholders’ financial interests against those affected by its conduct and its stated purpose. It will be a Nasdaq “controlled company”, and says it expects to use that exemption only for its nominating committee. Class F lasts until fewer than three continuing founders remain, followed by a transition period, or until it is redeemed.

There are two routes once the deal is live, and neither exists today.

• An IPO allocation, at the IPO price. The filing says a portion of shares is expected to reach retail investors through Charles Schwab, Fidelity, Robinhood, Trade‑PMR, SoFi and E*TRADE from Morgan Stanley, on top of underwriters’ own retail allocations. Each platform sets its own conditions. The filing gives no retail share size, and it describes a separate retail offer in parts of the European Economic Area.
• The open market, from the first trade. Investors whose broker supports ANT and permits trading in their jurisdiction could buy once it opens, at the market price; IPO allocations are not guaranteed. That can be far from the IPO price: Cerebras priced at $185 in May 2026, opened at $350 and closed its first day at $311.07.

Private shares and tokens. Anthropic’s help page (shown as dated Jun 29, 2026) says stock sales its board has not approved are “void”, that it does not permit special purpose vehicles to acquire its stock, and that public offers through “tokenized securities, or other mechanisms” are likely an investment with no value. Some sellers dispute how far such restrictions reach, and we take no position. The risks a buyer of any such product carries are that it may not own what it thinks it owns, the price may not track the real company, and the exit may be limited. Listed funds that say they hold private AI shares are a separate case: their holdings and any premium to underlying value are in their own documents. Tokens more broadly are covered in can US investors buy tokenized stocks.

Lock‑ups. Insiders’ shares would not all stay off the market for a single 180‑day period. The filing’s schedule starts releases on the first trading day, with more on the 27th, 48th, 62nd and 76th calendar days after the prospectus date and after the 2026 annual results, subject to the filing’s conditions. Share counts are still blank. How to read a schedule like this is in the IPO lock‑up plan.

A Nasdaq‑100 fund could own it within weeks if it qualifies. An S&P 500 fund could not for at least a year, and profitability is a further test. The rules below are as published in 2026, and we found nothing from either provider specific to Anthropic.

Sources: Invesco’s summary of the May 2026 Nasdaq‑100 changes, the S&P Dow Jones Indices decision of Jun 4, 2026 and a report on SpaceX’s entry. So a plain S&P 500 index fund holder would have no ANT at first, while a Nasdaq‑100 fund holder might within about three weeks of trading.

Most of these 13 stocks closed lower a year later than on their first day. Nine of 13 did, 7 fell 50% or more below that close at some point, and only Arm and CoreWeave beat SPY from that close. Buying at the IPO price looked better: 6 of the 12 with an IPO price were above it after a year, and 10 of 12 closed day one above it. Daily prices are Exent market data.

Selection rule. Every IPO from 2012 on in Renaissance Capital’s all‑time top 25 US IPOs that it labels Technology, with a full first year (Alibaba, Facebook, Uber, Arm, Coupang, DiDi). We added Rivian, which it labels consumer discretionary; that was a judgment call made knowing its result, so we report the test: without Rivian, CoreWeave and Coinbase, the median first‑year return is −23.8% rather than −26.4%. Then tech and internet IPOs of at least $3B below that cutoff that we could identify (Airbnb, Snap, DoorDash, Snowflake), CoreWeave as the only Technology name in its 2025 top three, and Coinbase as the largest direct listing we found by first‑day value. Reddit and other smaller listings fall outside the rule.

Method. IPO prices come from each final prospectus, linked below. “1 year later” is the last close on or before the first anniversary. Stock figures are split‑adjusted prices; none of these paid a dividend in its first year. SPY includes dividends. Arm’s first‑day close comes from a public report. DiDi left the NYSE in June 2022, so its later closes are over‑the‑counter. Medians for the IPO‑price columns use 12 names.

Limits. Thirteen listings is a small sample. Seven came between Sep 2020 and Nov 2021, just before the 2022 bear market, which weighs on the median. None of them had Anthropic’s size or structure, and this is not a forecast for any stock. The 2026 mega listings have under a year of data so far, to the Nov 4 close:

Analysts at the underwriting banks must wait at least 10 days, and in recent large deals they waited about 25. FINRA Rule 2241 requires firms that took part in an IPO as underwriter or dealer to set a quiet period of “a minimum of 10 days”, with exceptions for emerging growth companies and significant news. Anthropic’s filing does not mark it as an emerging growth company.

The day 25 pattern shows up in practice: banks began CoreWeave coverage on Apr 22, 2025 and SpaceX coverage on Jul 7, 2026, each 25 days after the first trade. Firms outside the deal are not bound the same way and can start earlier.

With 22 banks on this cover, a wave of first ratings could arrive on one day, from banks that earned fees on the deal. That matters for reading them. Across all US stocks, Exent’s study of 13,834 initiations found a small average move around the call, and 68.4% of 661 recent initiations were Buy; see what happens when an analyst initiates coverage. A rating says little until you know that analyst’s past record.