America 2.0 · Live tracker
Read America 2.0 →

Anthropic IPO

Risk to the US Economy

Opened October 20 · Day 20
The Nasdaq MarketSite tower in Times Square, with the Nasdaq logo on its curved video screen.
Nasdaq MarketSite in Times Square, New York (illustrative). Reports say Anthropic plans to list on Nasdaq. Photo: ajay_suresh / Wikimedia Commons (CC BY 2.0)

Anthropic, the maker of Claude, is preparing what could be the largest IPO on record. Reports last month put its target as high as $2 trillion; after this week’s tech selloff, prospective buyers are discussing $1.1 trillion to $1.3 trillion (Bloomberg) and investors call even $1.5 trillion “a stretch” (CNBC). The stakes reach beyond one company: Amazon, Google and Microsoft are at once its investors, its suppliers and its sales channels, and its listing will be a public test of the AI spending boom. It made its IPO filing public on Oct. 26. This page tracks the filing, the valuation and the risks, and what each possible outcome would mean for the US economy.

Latest·
“Investors brace for lower Anthropic valuation, calling $1.5 trillion a 'stretch'”CNBC ↓
Risk to the US EconomyUpdated
ELEVATEDThe FT reported Nov. 6 that Anthropic will press ahead with an IPO targeting $1.5 trillion; some buyers are discussing $1.2 trillion to $1.3 trillion (Bloomberg).

Anthropic made its IPO filing public on Oct. 26, after filing confidentially in June. It could list as soon as mid-November. The FT reports it is still targeting $1.5 trillion, but after a tech selloff prospective buyers have discussed $1.1 trillion to $1.3 trillion (Bloomberg, CNBC). The filing shows fast revenue growth alongside large losses, most of them accounting charges, and hundreds of billions of dollars in long-term computing commitments to Google, Amazon, SpaceX, Microsoft and others. The price range and share count are still blank.

ContainedLow risk to the economy: the IPO lists near its target and trades steadily.
ElevatedRisk is building: the IPO is on track but untested, its valuation disputed, its filing heavy with losses and fixed commitments.
HighHigh risk: the IPO falls short (delayed, downsized, priced below range or trading well under its IPO price), with spillover to its partners.
SevereSevere risk: the IPO is pulled or fails, with spillover into the wider economy: partners cut commitments, other AI listings are shelved, credit tightens.

What would change this rating ↓

anthropicipo.america2.news

Get America 2.0 in your inbox

Analysis of Russian influence, information warfare and the networks behind them. Free; unsubscribe anytime.

Latest reports

New reporting from major outlets and official sources, listed as it appears. Our monitoring adds headlines from major outlets automatically. "Checked" means our software fetched the full report and confirmed it supports its headline. "Reviewed" means an editor checked it against its source. "Auto" items have not been checked yet. Tap a label for our note. None of this is an endorsement. Our assessment is the rating above.

All 174 sources →

IPO clock and pricing signals

When the IPO could happen and at what value, from the filing rules and attributed reporting. We update it as filings land. These are signals for our risk rating, not a prediction.

When could it list?
NovDecJanReported target (Bloomberg)Reported target (Bloomberg): Nov 9–Nov 25Earliest pricing (filing rules)Nov 16Q3 figures filed (due Nov 11)Nov 2Holiday lullOur slip windowOur slip window: Jan 5–Jan 12Today

Public S-1 filed Oct 26. Earliest pricing is about three weeks after it appears. As of Nov 8.

At what value? (trillions of dollars)
HighElevatedContained$0T$0.5T$1T$1.5T$2T$2.5TNew Constructs estimate$150BLast private round (May)$965BInvestors pushing for$1.5TEarly backer: pricing$1.8TReported target$2TEarly backer: first trades$2.3T–2.4T

Shading shows where a pricing would move our rating. The per-share price range is not set yet; it will appear in an amended S-1.

  • Public S-1 filed: Oct. 26 (SEC EDGAR). Anthropic has applied to list Class A shares on Nasdaq under the symbol “ANT”; the price range and share count are not yet set.
  • Earliest possible pricing: mid-November. The law requires the public filing at least 15 days before the roadshow, so marketing could start Nov. 10 at the earliest (Yahoo Finance). SpaceX priced 23 days after its public S-1 this year (SEC EDGAR); the same pace would mean pricing around Nov. 18.
  • Reported target: marketing as soon as the week of Nov. 9 and trading before Thanksgiving, according to Bloomberg (via Yahoo Finance). The Wall Street Journal reported the offering could come “as early as the week before Thanksgiving” (WSJ). Anthropic is also seeking up to $3.5 billion from Japanese investors as part of the offering (Nikkei Asia).
  • Key date, Nov. 11 (now met): financial statements in an IPO filing go stale after 134 days (Deloitte). The S-1’s financial statements run through June 30, and it leaves blanks where third-quarter figures will go (S-1). A listing after Nov. 11 needed those figures; Anthropic added them in an amended filing on Nov. 2 (S-1/A).
  • Our estimate: a listing before Thanksgiving is still possible. It needs an amended S-1 with a price range and third-quarter figures. If neither appears by about Nov. 11, a listing before Thanksgiving is unlikely, December is squeezed by the holidays, and the earliest realistic window becomes Jan. 5–12.

Pricing signals

  • Reported target: about $2 trillion (Reuters, via The Star). Some investors are pushing for about $1.5 trillion (Reuters, via CP24).
  • One early backer, a venture investor who spoke anonymously, expects pricing “at around $1.8 trillion” and first trades at $2.3 trillion to $2.4 trillion (Fortune).
  • Last private valuation: $965 billion in May, when Anthropic raised $65.2 billion (S-1; Orrick).
  • An internal mark in August: the filing says Anthropic booked $2.0 billion to match employees’ share donations made in August and issued about 28 million shares to settle it, about $71 a share. On the share count below, that implies roughly $1.2 trillion, FT Alphaville calculates, though the valuation experts it consulted called the math “very noisy”: the inputs are rounded and such internal marks can be discounted 10% to 50% (FT Alphaville).
  • What share price each valuation implies (our estimate, not the company’s):
    ValuationPrice per share
    $965 billion (May, Series H)~$57
    $1 trillion~$59
    $1.5 trillion~$89
    $1.75 trillion~$103
    $2 trillion~$118
    $2.5 trillion~$148
    Based on about 16.9 billion fully diluted shares as of June 30: the S-1 says 846,840,960 shares given to its Public Benefit Fund are 5% of the total (S-1). Series H investors paid $589.01 per preferred share, which converts 10 to 1, or about $58.90 a share on May’s smaller share count. New shares sold in the IPO would lower each price slightly (about $4 a share for a $75 billion raise). The actual price range will come in an amended filing.
  • Per-share price range: still blank in the S-1. It will appear in an amended filing; SpaceX's came about 10 days before pricing.

How pricing would move our rating

  • Priced at or above about $1.8 trillion, with trading that holds: toward Contained.
  • Priced between about $1.5 trillion and $1.8 trillion: stays Elevated.
  • Priced below about $1.5 trillion, downsized, or trading below the IPO price: toward High.

What would change this assessment

The rating measures the risk Anthropic's IPO poses to the US economy, judged on events we can document, not on our view of what the company is worth. Any one of these would move it.

  • seenThe public filing: Anthropic filed its S-1 on Oct. 26. It confirms the main figures first reported from leaked drafts; the price range and share count are still blank.
  • not seenDown to Contained: the IPO prices within or above its range, near its target valuation, and trades steadily through its first weeks.
  • not seenToward Contained: third-quarter figures in an amended filing (needed for a listing after Nov. 11) showing growth holding up, followed by clear signs of demand at the target valuation, such as a raised price range or a heavily oversubscribed order book.
  • not seenUp to High: a delay past year-end, a smaller offering, pricing below the range, or shares trading well below the IPO price.
  • not seenUp to High: stress at the companies on the other side of Anthropic's commitments, such as share or bond selloffs at computing partners tied to the IPO.
  • not seenUp to High: SpaceX ends, cuts or reprices Anthropic’s compute contract.
  • not seenUp to Severe: the IPO is withdrawn or postponed indefinitely.
  • not seenUp to Severe: computing partners cut or renegotiate Anthropic's commitments, or another major AI listing is shelved, citing Anthropic's offering.

Three ways this can go

The outcomes we are watching for, what each would look like, and what we have seen so far.

OutcomeWhat we would seeRisk to the economy
1. The IPO doesn't happenAn amended S-1 with a price range never appears, the roadshow is postponed without a new date, or the offering is withdrawn. So far: the schedule has slipped from October to November (WSJ, via Crypto Briefing), but Bloomberg reports marketing could start the week of Nov. 9 (Yahoo Finance).High
2. It happens but falls short or slumpsA smaller raise, a price range below the reported target, pricing at the bottom of the range, or shares falling after listing. So far: some existing and prospective investors have argued for about $1.5 trillion instead of $2 trillion after third-quarter results fell short of expectations (Reuters, via CP24).High
3. It happens at the hoped-for valuationPricing at or above about $2 trillion and steady trading. So far: Anthropic made its S-1 public on Oct. 26, and attendees at its Oct. 14 investor day came away optimistic, Reuters reported (via KFGO).Lower near term

Potential impacts of an IPO shortfall or failure

No central bank or the IMF models an Anthropic IPO directly. Several have published scenarios for the wider AI correction that a failed or sharply repriced listing could feed into. The ranges below come from those reports.

ImpactWhat the reports describe
Low
Contained repricing		A listing below the roughly $2 trillion target that markets absorb. The Bank of England judges that “an equity shock in isolation would be unlikely to present a direct risk” to UK financial stability (Bank of England, July). In the European Central Bank’s private-credit stress test, euro-area banks’ losses do not exceed “1.3% of total equity” (ECB, May). JPMorgan Asset Management expects heavy AI borrowing to cause “some indigestion,” not “a credit crisis” (JPMorgan).
Medium
The financing window narrows		Share sales stop funding AI growth, pushing more of it onto debt. The Bank of England, noting that AI companies are turning to IPOs, says equity financing “could be positive for financial stability” because “equity capital can preserve debt capacity” (Bank of England). The Financial Stability Board expects $800 billion of AI spending through 2028 to come from private credit (FSB). The IMF counts about $650 billion of debt in circular deals among AI firms and warns that rising leverage could bring “forced selling and surges in equity volatility and credit spreads” if AI profits disappoint (IMF, October).
High
Trigger for a broad AI correction		In the Bank of England’s near-term scenario, US stocks fall 45% over six quarters, corporate credit spreads widen 350 basis points, the dollar weakens persistently and UK output falls 2.2 percentage points (Bank of England). In the IMF’s “market correction” scenario, US stocks fall 20% and world output is 1.1% lower by 2028, 1.8% in the US (IMF). The Bank of Japan’s Oct. 23 stress test assumes AI-related US stocks fall about 50% and BBB corporate spreads widen about 1,100 basis points (Bank of Japan). The IMF’s global bank stress test, which pairs an AI correction with higher interest rates and an energy shock, finds banks holding 22% of global banking assets falling below a 7% capital ratio (IMF).

What is actually known

  • The filing is public. Anthropic filed its S-1 on Oct. 26 (SEC EDGAR), after saying on June 1 that it had “confidentially submitted a draft registration statement on Form S-1” (Anthropic). It has applied to list Class A shares on the Nasdaq Global Select Market as “ANT.” Goldman Sachs and Morgan Stanley lead 22 underwriters. The price range and share count are still blank.
  • The S-1 confirms the numbers first reported from leaked drafts. Revenue rose from $386 million in 2024 to $4.6 billion in 2025, and to $16.3 billion in the first half of 2026, up from $1.2 billion a year earlier. The net loss was $42.0 billion in 2025 and $80.7 billion in the first half of 2026 (S-1). Sales through cloud providers’ marketplaces were 47% of revenue in 2025 and 48% in the first half of 2026; the filing does not break that out by company (S-1). Reuters first reported the 2025 figures from a confidential draft (via The Star). The Wall Street Journal’s Oct. 20 report of up to $750 billion in possible computing spending, including deals still being negotiated (WSJ), is not a figure in the S-1.
  • Growth is fast but slowing. Third-quarter revenue was $17.0 billion, up 47% from the second quarter, after 144% growth the quarter before. Computing and infrastructure costs rose 77%. Compared with a year earlier, revenue grew 1,127% and computing costs 382%, a gap the filing presents as evidence of compute efficiency (S-1/A). Price reductions on new model versions and contractual discounts took about $600 million out of revenue between the second and third quarters, the filing says (S-1/A). The filing says $5.0 billion of short-term computing “currently costs significantly more” than its long-term deals and “reduced revenue per dollar of compute compared to the preceding quarter.” The company swung to a $567 million operating loss from $974 million of operating income. The filing attributes that mainly to the costlier short-term computing and to $2.0 billion of non-cash charges for its employee donation-matching program. Its own adjusted margin, which excludes those charges, fell to 15% from 19%, mainly on the computing costs. The $23.3 billion net loss is mostly a $21.9 billion accounting charge on convertible notes and other instruments. The filing repeats that it expects its growth rate “to decline over time” (S-1/A). On Oct. 14 Anthropic had told prospective investors that third-quarter revenue was $16.8 billion, up 45% from $11.6 billion in the second quarter (the S-1 shows $11.555 billion for the second quarter and leaves the third quarter blank), and that it was profitable on an adjusted basis, Reuters reported (via KFGO). Slowing month-to-month growth has raised investor concerns about the valuation (Reuters, via CP24). Model versions released in the six months ended Sept. 30 brought in more than 80% of third-quarter revenue; the filing calls the “continuous and overlapping cadence of new model version releases” inherent to staying at the frontier (S-1). The price range is still blank.
  • The commitments are huge and mostly fixed. As of June 30, Anthropic had about $265.6 billion in fixed or minimum purchase commitments, plus $321.4 billion in leases that had not yet started. After June 30 it signed about $245 billion more, plus about $21 billion in short-term commitments (S-1). The filing leaves its own grand total blank, and the categories may overlap, so we do not add them up. Against that, Anthropic held $112.4 billion in cash and $19.3 billion in short-term investments on June 30, after raising $141.1 billion since its founding (S-1). CNBC calculated that among non-bank U.S. companies only Berkshire Hathaway and Alphabet held more (CNBC). Named deals include Google (at least about $114.5 billion through 2033), Amazon (at least $110.0 billion through 2036, “regardless of actual usage”), Microsoft (at least $44.9 billion through 2033), SpaceX (up to $120.5 billion), Nscale (up to about $44.6 billion), Lambda (about $41.3 billion) and data-center leases tied to Broadcom (about $161.2 billion in rent) (S-1). Reuters had reported at least $518 billion over a decade (via KFGO).
  • Its partners sit on every side of the deal. Amazon and Google “have invested tens of billions of dollars in Anthropic,” sell its products and supply its computing; Reuters calls it “an increasingly circular financial relationship.” The S-1 warns that relying on a few partners “creates complex dynamics that could give rise to conflicts of interest.” Amazon and Google will each own more than 5%, in non-voting shares; in May, Google put in $10.0 billion and Amazon $5.0 billion, and Google has an option to buy up to $10.0 billion more at the IPO price (S-1). No single customer was more than 10% of revenue in the first nine months of 2026; the top five were about 25% (S-1). The rise in Anthropic’s valuation that produced its $72.4 billion charge on convertible notes and similar instruments also lifted its backers’ profits. Amazon’s second-quarter net income included $53.4 billion of non-operating income, “primarily from our investments in Anthropic” (Amazon). Alphabet booked a $98.0 billion gain, mostly unrealized gains on equity securities, which reporting ties to Anthropic and SpaceX (Fortune). Anthropic’s own “Adjusted Net Income” excludes those charges, turning its $80.7 billion first-half loss into a $2.7 billion profit (S-1; Jonathan Weil, WSJ).
  • One supplier is controlled by Elon Musk. Since February, SpaceX has owned Musk’s AI company xAI, a direct Anthropic rival. Anthropic says in its S-1 that it agreed to spend up to $120.5 billion on computing from SpaceX’s AI unit through 2029, about $3.1 billion a month once all capacity is delivered, which it expects in December. After initial periods ending in January and March 2027, either side can cancel on 90 days’ notice (S-1). SpaceX’s own filing earlier described a smaller deal: about 325,000 Nvidia chips for $1.25 billion a month (SpaceX S-1). Musk says his company is “providing them with compute which they… really needed” (The Builders). This month, the Wall Street Journal reports, Musk personally ordered a slowdown of Ukraine’s military Starshield service (WSJ). On Oct. 20 Musk said SpaceX and Tesla have each set up committees of independent directors to study a merger; no terms are agreed (NBC News). The same day SpaceX reported that its AI unit is now its largest business, with $7.6 billion in quarterly revenue, and AFP reported that Google is paying SpaceX $920 million a month for computing through June 2029 (AFP). Bloomberg reports, citing a person familiar with the matter, that Anthropic is the unnamed customer behind about 34.5% of SpaceX’s third-quarter revenue, more than the U.S. government (via The Next Web).
  • Anthropic has set aside shares for a “public benefit” fund. The S-1 says it gave a wholly owned subsidiary, Anthropic Public Benefit Fund, LLC, 846,840,960 non-voting shares, 5% of its fully diluted shares as of June 30, after stockholders approved it. After the IPO the board may add up to 2% more each time the share price rises 50% above the last milestone, starting at the IPO price. The fund’s shares may be donated, transferred or sold; its first managers are Dario Amodei, Daniela Amodei, Jack Clark and Neil “Buddy” Shah (S-1). Each co-founder has pledged 80% of personal equity to philanthropy; the S-1 calls these pledges “voluntary, revocable” and not legally binding (S-1). Anthropic said on Oct. 18 that about 20% of its shares would be “dedicated to the public good” and that the fund’s shares would not be used for lobbying or political giving (Anthropic; AP); those two points are not in the S-1.
  • The market around it is already strained. The Philadelphia Semiconductor Index shed $1.6 trillion in less than two weeks; CoreWeave's shares and debt tumbled as investors scrutinized nearly $50 billion in borrowings; and several AI-linked listings were pulled or paused, Reuters reported on Oct. 19–20 (via CP24).
  • CoreWeave as a gauge. CoreWeave is not among the computing partners named in Anthropic’s filing, but as the biggest of the debt-funded “neocloud” providers it is a gauge of how investors feel about AI infrastructure. Its bonds fell to new lows on Nov. 4 after candidates favoring slower data-center construction won governors’ races, including in Texas: its notes due 2031 traded as low as 79.5 cents on the dollar, and its 2030 and 2032 debt yielded more than 15% (Bloomberg). Its shares fell 36% in October, and its five-year credit default swaps reached 1,145 basis points (CNBC).
  • Credit markets are pricing AI risk. Weekly trading in credit default swaps tied to Oracle, SpaceX, Alphabet, Amazon, Meta, Broadcom and Nvidia averaged more than $9 billion in September, up from under $2 billion in July, according to Morgan Stanley Research. This month the cost of that protection hit records at Meta, Nvidia, Alphabet and Amazon. Investors are buying protection on companies that supply computing to OpenAI and Anthropic; insurance on SpaceX’s debt reached a record 226 basis points. AI-linked borrowing is an estimated $466 billion this year, up from $101 billion in the same period of 2025 (FT, via Financial Post).
  • Its credit may land below investment grade. Blackstone, an anchor lender in Anthropic’s $63 billion “Project North Star” chip financing, has told investors it now expects a “high” junk rating after the IPO, in the double B range or above, rather than the investment grade it had predicted by year-end, people familiar with the matter told the Financial Times. A junk rating would shrink the pool of lenders for Anthropic’s plans to spend more than $800 billion on computing through 2045. Fitch, Moody’s and S&P have issued no rating; Kroll rates the deal’s $18 billion of junior debt double B plus. The Nasdaq 100 closed down 3.5% and briefly entered correction territory, in a chip-led rout: Intel fell 29%, its worst day since 1974, and CoreWeave 9.7%. The same day OPay’s $4 billion New York listing never began trading (Financial Times).
  • OpenAI is racing to raise first. OpenAI is seeking about $30 billion at a valuation above $1.4 trillion before Anthropic lists, partly to avoid comparison with Anthropic's disclosures (Reuters, via CP24). On Oct. 31 it cut the target to up to $20 billion at a $1.25 trillion valuation, after some investors reduced their commitments amid a tech selloff; the round has not closed (CNBC). OpenAI has ruled out going public in 2026 (CNBC).
  • A major computing partner sells off. On Oct. 29 Alphabet, one of Anthropic’s largest computing suppliers and an investor, lost about 13% of its value in early trading, roughly $500 billion, after reporting negative cash flow on more than $50 billion of quarterly AI spending and saying “AI mode” searches were cutting into ad profits. Its cloud sales doubled, helped by selling computing capacity to Anthropic (FT). The selloff spread on Oct. 30: Alphabet had closed down 13%, futures on the Nasdaq 100 fell 2.2% before the U.S. open, memory-chip makers SK Hynix and Samsung Electronics dropped 8.4% and 5.8%, and SoftBank, an OpenAI investor, fell 6% (FT).
  • A financing delay in Texas. Banks are delaying plans to sell part of a $15 billion loan for a data center tied to Anthropic in Texas, after the state put a hold on new projects while assessing risks to its power grid, leaving the debt on their books longer than planned (Bloomberg). Anthropic’s filing warns that Texas can curtail or disconnect large loads during grid emergencies, and that such actions “could delay or prevent the construction or expansion of facilities on which we depend for compute” (S-1).
  • A security test exposed internal systems. Cathedral, a start-up founded by former DOGE staffers, told Anthropic that within 12 hours its AI agents found paths to meeting transcripts, salary and job-offer details, and Slack messages about its models, using flaws in third-party business software, under Anthropic’s own testing program. Anthropic confirmed access to “a limited set” of sales, HR and IT systems, said no customer conversations, code or model weights were reached, briefed federal officials, and said Cathedral overstated the national-security implications and did not fully follow the program’s rules (Washington Post).
  • Much of the build-out is not yet on balance sheets. Microsoft, Meta, Oracle, Amazon and Alphabet had committed about $1.09 trillion in future payments under leases that had not yet begun, mostly for AI data centers, nearly four times the lease liabilities already on their books. Oracle is the most concentrated, with $260 billion of such commitments against $37.89 billion recognized (Reuters, Aug. 4, via Yahoo Finance).

What the S-1 says

Notable statements in Anthropic’s public filing of Oct. 26 (S-1). They are the company’s own statements, made under securities law; the 2023–2025 financial statements are audited, and the 2026 figures are not. Quotes are from the filing.

What the filing saysWhy it mattersKind
The founders and a trust will control the company. One share of Class F stock, held by a company of the seven co-founders, carries 50.1% of shareholder votes. Anthropic’s Long-Term Benefit Trust elects a majority of the board through Class T stock, which has no economic rights. Together they can elect every director.Buyers of the public Class A shares will have little say. Anthropic calls the structure “novel” with “no direct public company precedent,” and it will be a “controlled company” under Nasdaq rules. It says it will use that status only to keep a nominating committee that is not fully independent. The founders’ special vote winds down mainly as founders leave, through a 450-day transition.Governance
Insiders’ shares are locked up for up to 180 days, and the seven co-founders’ for up to 365 days, but the restrictions lift in about two dozen scheduled steps: the first on the first trading day for some directors, officers and employees, and later ones tied to the release of 2026 annual and early-2027 quarterly results.Insider selling is spread over months rather than one date. The number of shares freed at each step is blank until the amended filing, and Goldman Sachs and Morgan Stanley can release shares early.Disclosure
Revenue was $16.3 billion in the first half of 2026, up from $1.2 billion a year earlier. The net loss was $80.7 billion, but $72.4 billion of it was accounting charges on convertible notes and other instruments, and the operating loss was $693 million. In the second quarter alone it had operating income of $974 million, an 8% margin, its first quarterly operating profit in the filing’s tables.The headline loss mostly reflects Anthropic’s rising valuation, not cash spent; operations brought in $3.9 billion in cash in the first half, after using about $10.0 billion in 2023–2025. The filing says growth “may not be sustainable” and that it expects the growth rate to decline.Audited, 2026 unaudited
Third-quarter revenue is left blank (“$ for the three months ended September 30, 2026”), though the filing cites other measures as of Sept. 30: net revenue retention of about 900% and revenue per employee over $6.6 million.Anthropic told investors on Oct. 14 that third-quarter revenue rose 45% to $16.8 billion (Reuters, via KFGO), far slower than the second quarter’s 144% jump, and some investors said the results fell short of expectations (Reuters, via CP24). The figures must be added in an amended filing for a listing after Nov. 11.Disclosure gap
Net revenue retention of “approximately 900%” and revenue per employee of “over $6.6 million,” both as of Sept. 30, with about 5,300 employees.These are the company’s own operating measures, not audited figures. Retention of 900% means existing customers spent about nine times more than a year earlier.Company metric
Huge, mostly fixed computing deals: at least about $114.5 billion with Google, at least $110.0 billion with Amazon “regardless of actual usage,” up to $120.5 billion with SpaceX’s AI unit, at least $44.9 billion with Microsoft, plus Nscale, Lambda, AMD and leases tied to Broadcom.These bills come due whatever revenue does, which is the core of our risk rating. The SpaceX deal ties Anthropic’s computing to a company Elon Musk controls; it can be cancelled on 90 days’ notice after early 2027.Commitment
At June 30, Anthropic had about $587 billion in future compute obligations not yet on its income statement: $321.4 billion in leases that had not yet begun (starting 2026–2028 and running up to 20 years) and $265.6 billion in minimum purchase commitments, up from $54.6 billion at the end of 2025. It signed about $245 billion more after June 30. Other agreements with no fixed amount are excluded; the filing expects those payments to be “significant” (S-1).The filing says building out its own compute is “changing the cash flow profile” of the business and will be a significant use of cash “for the foreseeable future.” Capital spending so far is small ($482 million in the first half). At Sept. 30 the total was about $895.4 billion: $447.5 billion before 2031 and $69.8 billion due within 12 months. About $213.8 billion of that total is in deals subject to termination rights or to conditions before they start, such as the counterparty securing financing, facilities or regulatory approvals, and can be terminated if those conditions aren’t met; the filing ties its right to cancel mainly to counterparties failing to deliver (S-1/A). Anthropic had $146.6 billion of liquidity (S-1/A). The New York Times reports that Anthropic expects more than 3.5 gigawatts of computing power this year and aims to double that next year, citing two people with knowledge of its plans. It is also working with former crypto miners such as TeraWulf and Hut 8 to build data centers, and is buying chips and cooling gear directly to speed construction (NYT).Commitment
Its suppliers are also its lenders. All $6.2 billion of its debt at June 30 is convertible notes held by Amazon, one of its main cloud suppliers and sales channels. After the IPO, part of the computing bill can be paid in stock. Amazon’s $15.0 billion facility is settled in new non-voting shares at the 30-day average price as computing milestones are reached. Google can take its $20.0 billion facility in shares on the same terms, and AMD has committed to buy up to $5.0 billion of shares as its chips are delivered. Anthropic can also issue up to $42.0 billion of notes to Broadcom, or a financial partner it names, to pay for Broadcom-tied leases. Those notes convert at 90% of the 30-day average price, with interest paid in more notes. These partners can require Anthropic to register the shares for resale.The filing warns that a default under certain commercial agreements could trigger cross-defaults and the acceleration of “substantially all” of its debt, so a dispute with a supplier could become a debt problem. The notes convert into stock, which would dilute shareholders.Commitment
On Feb. 27, 2026, the President ordered the government to stop using Anthropic’s models, and the Department of War designated it a supply-chain risk. A federal court in California largely ruled for Anthropic, calling the actions “retaliatory”; on Sept. 25 a federal appeals court in Washington upheld the designation, 2–1. The department also “suggested it could invoke the Defense Production Act” to force Anthropic to drop its limits on autonomous weapons and mass surveillance.An unusual open conflict with the US government, disclosed as a business risk. Anthropic says it “may experience material revenue losses” from these events and has asked the appeals court to rehear the case.Legal risk
In June 2026 the Commerce Department imposed worldwide export restrictions on Anthropic’s Fable 5 and Mythos 5 models, which Anthropic disabled for all customers until the restrictions were lifted.A government action that briefly took its most advanced models off the market.Legal risk
Models, “including our own,” have shown “unexpected strategic or self-preserving behaviors, including attempts to resist shutdown.” During outside cybersecurity tests, a misconfigured setup left models connected to the internet, and “four of our models gained unauthorized access to real third-party systems.” In one case an early model used a password it found to get administrator access and read one person’s personal information. Anthropic says the affected parties were notified and it found no evidence that any model “attempted to evade oversight.”Unusually specific disclosure of AI misbehavior in a securities filing. The S-1 also warns that advanced AI could pose “catastrophic or existential risks” and that “no single insurer, and potentially the insurance industry in the aggregate,” may be able to cover a catastrophic AI incident.Safety risk
In September 2025 Anthropic assessed “with high confidence” that a Chinese state-sponsored hacking group used its Claude Code tool to try to break into about 30 targets, with “a limited number of successful intrusions.”A state-linked misuse case disclosed as a security risk.Safety risk
Revenue has included, and will include, money from customers Anthropic later finds are breaking its usage policy, revenue it would not expect to recur once their access is cut.Some reported revenue may not last. Anthropic says this is common in the industry and does not give an amount.Disclosure
A class action filed in September 2026 accuses Anthropic and other frontier AI developers of agreeing to slow the pace of model development, in violation of antitrust law.The same safety-pacing that Anthropic presents as responsible is being challenged as collusion. The case is at an early stage.Legal risk
Anthropic has “publicly committed to covering the cost of electricity price increases to consumers caused by our data centers.”An open-ended cost few companies take on; the filing does not estimate it.Commitment

The valuation gap

The competing numbers, and who put them forward. Anthropic has not published a valuation target or a price range.

  • About $2 trillion: the figure Anthropic is “seeking,” according to Reuters (via The Star). It first surfaced from investors quoted by the Financial Times in August (via Fortune).
  • About $1.5 trillion: what some existing and prospective investors have argued for since the third-quarter results (Reuters, via CP24).
  • $965 billion: the company's last private valuation, set in its May funding round (Orrick).
  • About $150 billion: the estimate of New Constructs, an independent research firm that calls the offering the “most ridiculous IPO of 2026” (CNBC).

Anthropic’s two biggest backers are now spending more on AI infrastructure than their businesses bring in. Amazon’s free cash flow was an outflow of $42.4 billion for the 12 months to Sept. 30, against an inflow of $14.8 billion a year earlier, driven by a $94.8 billion rise in equipment spending that “primarily reflects investments in artificial intelligence” (Amazon). Alphabet’s was negative $9.1 billion in the third quarter, after $51.4 billion of capital spending (Alphabet). Bloomberg reports the pessimism now hits even tech companies with stable cash flow (Bloomberg).

Claims in circulation

The most-shared claims, and what public sources can confirm.

Confirmed

Anthropic's draft S-1 leaked before it was public

Origin. Posts and newsletters since Sept. 28 describing a “leaked prospectus.”

What the evidence shows. Reuters and the Wall Street Journal have reported from copies of the confidential drafts (The Star; WSJ). New Constructs says the S-1 was “selectively disclosed” to a few parties (CNBC). Anthropic filed the S-1 publicly on Oct. 26 (S-1).

Bottom line. Drafts reached reporters; the text is now public.

Disputed

Anthropic's revenue is inflated

Origin. OpenAI, which has told investors and employees that Anthropic's accounting “inflates” its revenue by billions.

What the evidence shows. Anthropic counts the full value of sales made through cloud marketplaces as revenue and records the platforms' cut as a cost; OpenAI records revenue net of partners' shares. Anthropic says it follows established accounting practice (Reuters, via The Star and CP24). In its S-1, Anthropic says it is “the principal” in these sales and “recognizes revenue on a gross basis”; fees paid to channel partners are booked as sales, marketing and partnerships expense (S-1). The S-1 does not mention OpenAI's accusation.

Bottom line. Anthropic's filing now states its method; the dispute is over whether that method is right.

Disputed

Anthropic is worth $2 trillion

Origin. Investors quoted by the Financial Times in August, then banker and investor talk reported by Reuters, Bloomberg and the WSJ.

What the evidence shows. No price has been set; Anthropic's last private valuation was $965 billion in May (Orrick), and some investors now argue for about $1.5 trillion (Reuters, via CP24).

Bottom line. A target in talks, not a market price.

Needs context

Anthropic lost $80 billion in six months

Origin. The Wall Street Journal's Oct. 20 report on the new draft, and posts relaying it.

What the evidence shows. The S-1 shows an $80.7 billion net loss for the first half of 2026 (unaudited). Most of it is accounting: $72.4 billion is a change in the fair value of convertible notes, derivatives and SAFEs, which Anthropic says came mainly from its rising valuation, and $8.7 billion more is a change in the value of other financial instruments. The operating loss was $693 million, and operations brought in $3.9 billion in cash (S-1).

Bottom line. The loss is real on paper, but most of it is not cash spent.

False

Anthropic hid its expenses through $108 billion of convertible notes

Origin. Posts on X after the Oct. 26 filing comparing it to dot-com companies that paid suppliers in stock (X).

What the evidence shows. $108.6 billion is what the notes were worth on June 30 at Anthropic’s rising valuation; the amount still owed was $6.2 billion, to Amazon. The filing lists the notes as sold for cash ($11.25 billion in 2023–2025, none in 2026). Operating expenses of $17.0 billion for the half are on the income statement, and the operating loss was $693 million (S-1).

Bottom line. The notes raised cash and will become shares at the IPO; they did not pay or hide expenses.

False

You can buy Anthropic shares now through “$ANT” tokens or pre-IPO contracts

Origin. Crypto posts since the Oct. 26 filing promoting “$ANT” memecoins and pre-IPO perpetual contracts.

What the evidence shows. No Anthropic shares are on sale yet. The S-1 says the sale to the public begins “as soon as practicable after this registration statement becomes effective” (S-1). Pre-IPO perpetuals on Hyperliquid and other venues are derivatives that bet on a price; they carry no ownership, and regulators may examine them (Decrypt). Tokens using the $ANT name have no known tie to Anthropic.

Bottom line. Shares can be bought only once the offering prices: through the underwriters’ allocation, or on Nasdaq after trading starts.

What independent experts say

  • Michael Burry, Scion founder (“The Big Short”): adds up about $537 billion of computing deals with Amazon, Alphabet and Microsoft, plus preliminary deals worth hundreds of billions more, putting total commitments at about 12 times Anthropic’s annual revenue run rate. The S-1 gives no total; the figure is his own. The web of deals, he wrote, leaves “the entire organism catastrophically vulnerable either on a pause or on the way down” (Business Insider, via Yahoo Finance).
  • Ed Zitron, AI-industry critic: argues the second-quarter profit reflects lower SpaceX fees during the ramp-up, which the S-1 says differ from the full monthly rate, and tallies more than $430 billion of non-cancellable commitments still to come. Both figures are his own readings of the filing (Where’s Your Ed At).
  • PitchBook, via the FT: estimates Anthropic’s hard-to-exit commitments at $450 billion to $600 billion before 2031, as cited by FT columnist John Thornhill; the S-1 leaves those totals blank (FT).
  • David Trainer, New Constructs: values Anthropic at about $150 billion and calls the offering “an unprecedented test of investor gullibility.” His firm called WeWork's 2019 IPO the most ridiculous of its year and was right; it was wrong about DoorDash (CNBC).
  • Dauvin Peterson, 22V Research: “the revised IPO valuation range starting at $1.5T vs the prior expectation of at least $2T was viewed as disappointing and added fuel to the fire for the AI-related equities that were already under pressure” (CNBC).
  • Robert Dafforn, Polus Capital: because of the circularity in AI financing, “whatever the weakest link is will actually start to deteriorate the system” (FT, via Financial Post).
  • James Mackintosh, WSJ columnist: the AI trade rests on a “worrisome growth in interconnections between AI companies, which increasingly means problems at one firm will create trouble elsewhere.” He compares chip makers financing weaker data-center operators to the telecom-equipment makers that wrote off billions in 2001 when dot-com customers failed to pay, and says shareholders “should pay a bit less to cover these growing risks” (WSJ).
  • Edward Best, Willkie Farr & Gallagher (not working on the deal): “It's really a two-horse race in terms of being public” (Reuters, via CP24).

What to watch next

  • Whether the Texas loan sale resumes before pricing.
  • Whether the Alphabet selloff holds and spreads to other Anthropic computing partners or AI-linked debt before the IPO is priced.
  • An amended S-1 with the price range, share count and third-quarter figures, which a listing after Nov. 11 requires.
  • The start of the roadshow, which Bloomberg reports could come the week of Nov. 9, and the price range (Yahoo Finance).
  • OpenAI's roughly $30 billion funding round, and whether other companies keep delaying listings until Anthropic's is done.
  • The health of AI-infrastructure borrowers, with CoreWeave’s shares and bonds as a gauge, and of the companies on the other side of Anthropic’s commitments.

Timeline

  • MayAnthropic raises $65 billion at a $965 billion valuation (Orrick).
  • Jun 1Anthropic says it has confidentially submitted a draft S-1 to the SEC (Anthropic).
  • Aug 13The Financial Times reports investors expect a $2 trillion valuation (via Fortune).
  • Sep 18The planned listing moves from October to November (WSJ, via Crypto Briefing).
  • Sep 28–29Reuters reports from a leaked draft prospectus (via The Star).
  • Oct 1Bloomberg reports marketing could start the week of Nov. 9, with trading before Thanksgiving (via Yahoo Finance).
  • Oct 7New Constructs calls it the “most ridiculous IPO of 2026” and values the company at about $150 billion (CNBC).
  • Oct 14Investor day: third-quarter revenue of $16.8 billion, up 45% (Reuters, via KFGO).
  • Oct 18Anthropic says it will put 5% of its shares in a new Public Benefit Fund (Anthropic). The S-1 says stockholders approved it.
  • Oct 20The Wall Street Journal reports from a new draft: an $80.7 billion first-half net loss and up to $750 billion in possible computing spending (WSJ).
  • Oct 20Musk says SpaceX and Tesla have set up committees to study a merger; SpaceX reports its AI unit is now its largest business (NBC News).
  • Oct 26Anthropic files its S-1 publicly and applies to list on Nasdaq as “ANT”; the price range and share count are blank (S-1).

Key sources

All 174 sources →Every report we have listed, every source cited on this page, background reading, and the sources we cite only with caution.