Anthropic Valuation History: Every Round From Series A to the 2026 IPO

Anthropic has eight named priced equity rounds from Series A through Series H. The company itself published post-money valuations only from Series E onward. Earlier prices come from press, and Amazon/Google’s large checks were mostly convertibles that converted into later priced rounds rather than standalone priced financings. As of October 3, 2026, Series H (May 28, 2026) is the last priced equity round the company has announced. It confidentially filed a draft S-1 on June 1, 2026; that is not a priced round.

Anthropic PBC has not filed its own Form Ds for these rounds. Terms below are from Anthropic blog posts where noted, otherwise from press. Simple pre-money equals stated post-money minus stated amount raised, which is valid only if the full amount is new primary capital included in the post-money figure. Series G and H explicitly fold in previously committed strategic money, so that subtraction overstates “new” cash.

Priced equity rounds

Round	Announced	Close	Amount	Pre-money	Post-money	Lead / participants	Source quality
Series A	May 28, 2021	Not separately disclosed; company announced the raise that day	$124M	Not disclosed	Not disclosed by the company	Lead: Jaan Tallinn. Named: James McClave, Dustin Moskovitz, Center for Emerging Risk Research, Eric Schmidt, “and others”	Anthropic press release. No valuation in the release.
Series B	Apr 29, 2022	Not separately disclosed	$580M	Not disclosed by the company. Court-record math implies roughly ~$3.1B if SBF’s $500M was 13.56% post-round (see discrepancies)	Not disclosed by the company. Implied ~$3.7B from that stake, not a company figure	Lead: Sam Bankman-Fried (FTX). Named: Caroline Ellison, Jim McClave, Nishad Singh, Jaan Tallinn, Center for Emerging Risk Research	Anthropic press release for amount and investors. Valuation is not company-disclosed.
Series C	May 23, 2023	Announced as raised that day	$450M	Not disclosed	Not disclosed by the company. Press put the round at above $4B / about $4.1B	Lead: Spark Capital (Yasmin Razavi joined the board). Company-named: Google, Salesforce Ventures, Sound Ventures, Zoom Ventures, “and others.” VentureBeat also named Menlo Ventures	Anthropic press release for amount and named investors. Valuation is press only; company declined to confirm.
Series D	In market from Dec 20–21, 2023	No company announcement. NYT (Feb 20, 2024) said Menlo closed $750M “this month.” FT (Mar 27, 2024) still described a VC close expected in April	$750M (press)	Press conflict: $15B excluding the round (The Information, Dec 20, 2023) vs $18.4B pre-round (Reuters, Dec 21, 2023, citing its own source)	Press conflict: ~$15B after the round (NYT, three people) vs ~$18–18.4B in Forbes and later retrospectives. Treat as unresolved, roughly $15–19B	Lead: Menlo Ventures, largely via SPV “Menlo Inflection AI Partners” (Forbes: ~$500M via the SPV, ~$250M from Menlo funds and insiders). Lightspeed has said it invested in the Series D. Forbes reported Google’s convertible tranche was set to convert into this round; Spark had pro-rata	Press only (The Information, Reuters, Forbes, NYT, FT). No Anthropic release found.
Series E	Mar 3, 2025	Company and Bloomberg described it as closed that day	$3.5B	$58.0B (implied: $61.5B − $3.5B)	$61.5B	Lead: Lightspeed Venture Partners ($1B, per Bloomberg, citing Anthropic). Company-named: Bessemer, Cisco Investments, D1 Capital Partners, Fidelity Management & Research, General Catalyst, Jane Street, Menlo Ventures, Salesforce Ventures, “among other new and existing investors”	Anthropic press release for amount, post-money, and investors. Lightspeed check size is Bloomberg. Upsized: originally aimed at about $2B (Bloomberg; SiliconANGLE).
Series F	Sep 2, 2025 (“has completed”)	WSJ (Sep 10, 2025) placed Iconiq’s lead investment in August 2025; public announcement was Sep 2	$13B	$170B (implied, and the figure FT used in its headline)	$183B	Lead: ICONIQ. Co-leads: Fidelity Management & Research, Lightspeed. Company-named: Altimeter, Baillie Gifford, affiliated BlackRock funds, Blackstone, Coatue, D1, General Atlantic, General Catalyst, GIC, Growth Equity at Goldman Sachs Alternatives, Insight Partners, Jane Street, Ontario Teachers’ Pension Plan, Qatar Investment Authority, TPG, T. Rowe Price Associates, T. Rowe Price Investment Management, WCM, XN	Anthropic press release. Upsized: FT said Anthropic originally sought about $5B; TechCrunch cited earlier reports of $3–5B at a $170B valuation.
Series G	Feb 12, 2026	Company announced the raise that day. Process ran from early January; FT (Jan 27) said an initial $10–15B would lock in around Jan 27, with the rest in following weeks	$30B post-money headline	$350B (FT explicitly; also $380B − $30B)	$380B	Leads: GIC and Coatue. Co-leads: D. E. Shaw Ventures, Dragoneer, Founders Fund, ICONIQ, MGX. Company-named: Accel, Addition, Alpha Wave Global, Altimeter, AMP PBC, Appaloosa, Baillie Gifford, Bessemer, BlackRock-affiliated funds, Blackstone, D1, Fidelity, General Catalyst, Greenoaks, Goldman Sachs Alternatives growth equity, Insight, Jane Street, JPMorganChase (Security and Resiliency Initiative and Growth Equity Partners), Lightspeed, Menlo, Morgan Stanley Investment Management, NX1, QIA, Sands Capital, Sequoia, Temasek, TowerBrook, TPG, Whale Rock, XN. Includes a portion of previously announced Microsoft and NVIDIA investments	Anthropic press release for final terms. Path to those terms is WSJ, FT, Bloomberg. Marked upsize / extension (below).
Series H	May 28, 2026	Company said it “has raised” the round that day. Last priced round as of Oct 3, 2026	$65B	$900B (implied: $965B − $65B). Matches Bloomberg’s May 13 report of talks at more than $900B excluding new capital	$965B	Leads: Altimeter, Dragoneer, Greenoaks, Sequoia. Co-leads: Capital Group, Coatue, D1, GIC, ICONIQ, XN. Company-named: AMP PBC, Baillie Gifford, Blackstone, Brookfield, D. E. Shaw Ventures, DST Global, Fidelity, General Catalyst, Insight, Jane Street, Lightspeed, MGX, NTTVC, NX1, Situational Awareness LP, T. Rowe Price Associates, T. Rowe Price Investment Management, Temasek, plus Micron, Samsung, and SK hynix. Includes $15B of previously committed hyperscaler money, including $5B from Amazon	Anthropic press release. Upsize path is TechCrunch and Bloomberg.

Where sources disagree, and which closes were upsized

Series A valuation is not a company number. Stock-data sites put post-money anywhere from about $550M to about $845M (Dealroom has cited $845M; Forge-linked secondary data has been cited around $623M). Anthropic never confirmed a price. Do not use those figures as round terms.

Series B price is inferential. Business Insider, citing court records, reported that Bankman-Fried bought $500M for 13.56% at the time of the 2022 investment. That arithmetic implies roughly $3.7B post-money, not the $4.1B some databases assign to both Series B and Series C. The company never published a Series B valuation. The same records are the basis for later reports that the FTX estate sold that stake in 2024 for about $1.3B.

Series C vs a March 2023 Spark report. The Information, via Axios and others in March 2023, reported a $300M Spark-led raise at $4.1B. On May 23, 2023 Anthropic announced a $450M Series C led by Spark and did not disclose valuation. TechCrunch said the company would not give a price and that The Information had earlier said Anthropic was seeking capital at an over-$4.1B valuation. Axios/PitchBook said the May round valued the company at more than $4B. Whether the March figure was an earlier close, a different check, or a report that was later superseded is not resolved in company materials.

Series D is the weakest priced round in the set.
- The Information (Dec 20, 2023), via Reuters: talks for $750M, $15B not including the investment, final price could go above $18B.
- Reuters the next day, citing its own source: pre-round valuation of $18.4B.
- Forbes (Jan 11, 2024): round still in progress; “nearly quadruple” the valuation to $18.4B; Menlo SPV filing aimed at $500M, with another $250M from Menlo and insiders. Forbes also said the round was partly a way to price Amazon’s and Google’s earlier bets, and that Google’s investment converted into this round.
- NYT (Feb 20, 2024): Menlo closed $750M that month; three people said valuation had tripled to $15B.
- FT (Mar 27, 2024): VC investors expected to commit at least $750M, close expected in April, combined deals at a valuation of more than $18B.

There is no Anthropic blog post. $15B vs $18.4B should be left as a discrepancy, not averaged. Later write-ups (including a 2026 TechCrunch account of Menlo’s bet) repeat ~$18–18.4B and a “quadrupling,” which tracks Forbes/Reuters more than the NYT’s post-close $15B.

Series E was upsized, not extended after announcement. Bloomberg reported Anthropic had planned to raise about $2B and closed $3.5B because the round was oversubscribed. Company post-money of $61.5B is consistent across Anthropic, Reuters, Bloomberg, and the NYT. Some data vendors later print a different Series E valuation (for example about $59.8B); that conflicts with the company release.

Series F was upsized from a much smaller target. FT: originally sought about $5B, closed $13B. TechCrunch: prior reports of $3–5B at $170B, versus the completed $13B at $183B post-money. The $170B figure is the pre-money, not a rival post-money. WSJ’s “August” versus the September 2 announcement is a close-versus-announce gap, not a second round.

Series G is the clearest multi-step upsize.
- WSJ (Jan 7, 2026): plans to raise $10B at $350B before the new investment, led by GIC and Coatue, in addition to up to $15B Nvidia and Microsoft had planned to invest.
- FT (Jan 27, 2026): target doubled to about $20B at a $350B valuation after demand of five to six times the original target; initial $10–15B expected imminently, rest over the following weeks.
- Bloomberg (Feb 6, 2026): on track for more than $20B at $350B, close as soon as the next week; checks north of $1B each from Coatue, GIC, and Iconiq, plus as much as $15B from Nvidia and Microsoft.
- Anthropic (Feb 12, 2026): $30B at $380B post-money, which is $350B pre-money. FT the same day said the company raised its funding target by $10B during the process.

The $30B is not all incremental venture capital. Anthropic said the round includes a portion of previously announced Microsoft and NVIDIA investments. It did not say how much of the $30B that portion was.

Series H was also upsized and partly pre-committed. TechCrunch said that in April 2026 Anthropic was close to a $50B round. Bloomberg (May 13, 2026) reported early talks for at least $30B at more than $900B excluding new money. The May 28 release was $65B at $965B post-money ($900B pre-money), of which $15B was previously committed hyperscaler capital, including $5B from Amazon. Stripping that $15B leaves about $50B of other capital, in line with the earlier TechCrunch figure. Treat $65B / $965B as the company-stated round, and flag that headline size includes prior commitments.

Not separate priced equity rounds

These show up in funding databases as “rounds” but were not announced as independently priced primary equity.

• Google, late 2022, reported Feb 3, 2023 (FT): about $300M for roughly a 10% stake, which implies about $3B post-money, plus a cloud contract. Some outlets (VentureBeat, and a New York Times figure repeated by Fortune) said the deal valued Anthropic around $5B, which does not match a 10% stake for $300M. FT is the primary report; Google confirmed an investment and a cloud contract but not the stake size. This predates Series C and is best treated as a strategic equity investment, not a named series.
• Amazon, September 2023 ($1.25B) and March 2024 ($2.75B): up to $4B, structured as convertible notes, not a priced round. UK CMA decision (September 2024) describes the $4B as notes convertible into non-voting equity. NYT and Reuters reported the same structure.
• Google, October 27, 2023: up to $2B ($500M upfront, $1.5B over time). Bloomberg: Anthropic confirmed this was a convertible note that would convert at the next funding round. WSJ broke the commitment. Not a priced round on its own. Forbes later tied conversion of Google’s tranche to the Menlo Series D.
• Amazon, November 22, 2024: another $4B, bringing the commitment to $8B. Reuters: like the first $4B, this was convertible notes in phases, first tranche $1.3B. Anthropic said Amazon remained a minority investor. Not a disclosed priced equity round. A portion later converted into preferred stock around the March 2025 Series E, producing a multibillion-dollar accounting gain in Amazon’s Q1 2025 results (GeekWire, citing Amazon’s 10-Q). That conversion used the Series E price; it was not a new Anthropic round.
• Google, January 2025: press reported an additional about $1B. Not announced by Anthropic as a priced series, and no company valuation was attached in the sources retrieved here.
• Microsoft / NVIDIA commitments (announced before Series G) and Amazon’s further 2026 commitment: Anthropic’s Series G and Series H releases say those rounds include portions of previously committed strategic investments. April 2026 reports of Amazon putting in $5B (with more contingent) and Google agreeing to invest at a $350B valuation describe commitments that were later rolled into Series H’s $965B price, not a separate close at $350B.
• Debt, secondaries, and the IPO process are outside this table. Press has described a 2025 credit facility and, in 2026, a much larger debt package; those are not equity. Secondary sales (including the FTX estate sale) do not set a primary round price. The June 1, 2026 confidential draft S-1 states that share count and price have not been set.

Database tranche labels such as “Series H-1 / H-4” or “Series E-4,” and vendor valuations that reprint the same price for Series B and Series C, are not in Anthropic’s disclosures and should not be treated as additional priced rounds.

---

Recent Findings Supplement (October 2026)

The most recent priced equity round for Anthropic is the Series H (May 28, 2026), which closed the same day as the announcement per the company’s press release and contemporaneous reporting.[1][2][3]

This round raised $65 billion at a $965 billion post-money valuation (pre-money ~$900 billion), making Anthropic the most valuable private AI company at the time and surpassing OpenAI’s then-reported valuation. It was led by Altimeter Capital, Dragoneer, Greenoaks, and Sequoia Capital, with co-leads including Capital Group, Coatue, D1 Capital Partners, GIC, ICONIQ, and XN. Significant participants included Baillie Gifford, Blackstone, Brookfield, D.E. Shaw Ventures, DST Global, Fidelity Management & Research, General Catalyst, Insight Partners, Jane Street, Lightspeed Venture Partners, MGX, Temasek, and others, plus strategic infrastructure partners Micron, Samsung, and SK Hynix. It incorporated $15 billion in previously committed hyperscaler investments, including $5 billion from Amazon.[1][2]

Figures are directly from Anthropic’s official press release (anthropic.com/news/series-h), corroborated without material discrepancies by Reuters, TechCrunch, Financial Times, and others. The round was not described as an extension or upsized close in primary sources, though it exceeded an initial ~$30 billion target from financial institutions thanks to infrastructure partner participation and hyperscaler commitments.[4]

No new priced equity rounds have been announced or closed after the May 28, 2026 Series H as of October 3, 2026. Subsequent developments (e.g., confidential IPO filing in June 2026, ongoing IPO preparations targeting a potential November 2026 listing, revenue run-rate updates, and infrastructure commitments) do not involve new priced equity raises.[5][6]

Prior rounds referenced in recent 2026 reporting (for chronological context) include:
- Series G (announced/closed February 12, 2026): $30 billion at $380 billion post-money valuation. Co-led by GIC and Coatue Management, with participation from D.E. Shaw Ventures, ICONIQ, MGX, Founders Fund, and others (including portions of prior Microsoft/Nvidia commitments). Reported consistently in Reuters and Bloomberg timelines.[7][8]
- Series F (September 2025): $13 billion at $183 billion post-money. Co-led by Iconiq, Fidelity, and Lightspeed.[5]
- March 2025 round: $3.5 billion at $61.5 billion post-money, led by Lightspeed Venture Partners.[5]

Earlier rounds (e.g., Series A in 2021 and intermediates through 2024) predate the post-April 2026 publication focus and lack fresh primary details or discrepancies in the retrieved sources; they are not updated here as no new data emerged. All figures above derive from press reports and company announcements in the sources; no direct Anthropic SEC filings for these rounds were identified in results. No source conflicts on the Series H or G terms were noted.[9]

For anyone tracking or competing in this space, the Series H details underscore Anthropic’s ability to attract strategic (chipmakers) and hyperscaler capital alongside traditional VCs, tied to compute scaling and Claude demand—key differentiators versus pure financial rounds. IPO-related activity (confidential filing, investor day, revenue projections) signals a shift toward public markets rather than further private equity raises in the near term.

The secondary summaries disagree on cash versus contingent amounts. I’m checking primary announcements and the Amazon and Alphabet filings next.
Anthropic’s strategic cap table is a set of staged cash instruments sitting next to much larger purchase obligations. Cash invested by Amazon, Google, Microsoft, and others is real equity or convertible debt. The $100 billion-plus cloud contracts are Anthropic spending money on those same partners, not capital those partners put into Anthropic.

How the instruments actually work

The pattern, repeated from Amazon in 2023 through AMD in 2026, is a minority, usually non-voting security, often a convertible note or non-voting preferred, paired with a multi-year commitment by Anthropic to buy that investor’s chips or cloud. The cash check is small relative to the purchase obligation, and later tranches are gated on milestones. Amazon has said its original $4 billion was cash, not AWS credits. Google’s 2023 cloud contract was described as larger than the equity check and separate from it. Microsoft’s $30 billion Azure commitment is a purchase obligation, not a credit that substitutes for the equity investment.

Accounting then splits the economics. Amazon carries convertible notes at fair value with unrealized gains in other comprehensive income, and marks non-voting preferred through earnings when a new funding round supplies an observable price. Alphabet uses the measurement alternative: private stakes stay at cost until an observable transaction, then jump. Anthropic books the mirror image. A leaked draft prospectus shows a 2025 net loss of about $42 billion, of which roughly $34 billion is a non-cash remeasurement of convertible notes as the valuation rose, against an operating loss of more than $8 billion on about $4.6 billion of revenue.

Priced-round anchors used below: Series E, March 2025, $3.5 billion at a $61.5 billion post-money; Series F, September 2, 2025, $13 billion at $183 billion; Series G, February 12, 2026, $30 billion at $380 billion; Series H, May 28, 2026, $65 billion at $965 billion.

Amazon: $18 billion funded, $15 billion still gated on compute delivery

Amazon is the only strategic investor whose cash, instrument, and carrying value are laid out in its own quarterly filings. The position is minority, non-voting, and subject to an ownership cap Amazon can waive. Amazon has not published a percentage.

Date	Cash funded	Instrument	What was still contingent
Sep 25, 2023	$1.25 billion	Convertible note; option for a second note	Up to $2.75 billion more, exercisable by end of Q1 2024. AWS named primary cloud; Trainium and Inferentia for future models. Minority stake. Cash, not cloud credits.
Mar 27, 2024	$2.75 billion	Second convertible note, at the September terms	Completed the original $4 billion. TechCrunch noted the option let Amazon invest at the lower September valuation.
Nov 22, 2024, through Q4 2025	$4 billion more, staged	Additional convertible notes	Amazon announced another $4 billion and named AWS primary training partner. By year-end 2024 it had put in about $1.3 billion of that tranche and agreed to invest $2.7 billion more by the end of 2025. Filings: $8.0 billion of notes invested from Q3 2023 through Q4 2025. Some notes later converted into non-voting preferred (preferred carrying value was $14.8 billion at Dec 31, 2025).
Apr 20, 2026, recorded in Q2	$5 billion	Series G non-voting preferred	Announced as $5 billion now plus up to $20 billion more. Anthropic committed to spend more than $100 billion on AWS technologies over ten years and to secure up to 5 GW of Trainium capacity, including nearly 1 GW by the end of 2026.
Q2 2026, Series H	$5 billion	Series H non-voting preferred, via an option	Exercising the option cut a new financing facility from $20 billion to $15 billion.

Funded cash through June 30, 2026: $18 billion ($8 billion notes plus $10 billion preferred). Still available: up to $15 billion, not as a lump sum. Amazon’s Q1 2026 filing describes a facility of up to $20 billion that expires 30 months after a liquidity event. Nothing is drawable at inception. Amounts open only as Amazon hits compute-delivery milestones under the amended AWS agreement. Draws are new convertible notes or, after an IPO, common stock, issued for cash. That is equity financing tied to capacity delivery, not a cloud-credit investment.

Carrying value at June 30, 2026: $190.4 billion — notes at an estimated fair value of $97.9 billion (unrealized gain in AOCI of $92.0 billion) and non-voting preferred at $92.5 billion. In Q2 alone, Amazon recorded about $50.5 billion of upward adjustments on the preferred in other income, inside $53.4 billion of other income, and net income of $62.6 billion. Preferred marks hit earnings; note marks mostly sit in AOCI until conversion or sale. Earlier marks: about $13.8 billion at the end of 2024 on roughly $5.3 billion invested, and $60.6 billion at the end of 2025 ($45.8 billion notes plus $14.8 billion preferred).

The prospectus obligation is separate and larger: at least $110 billion to Amazon from May 2026 through April 2036, payable even if usage falls short. That matches the commercial shape of the April “more than $100 billion over ten years” pledge, not the equity check.

Google / Alphabet: early convertibles, a 15% cap, then a $40 billion ceiling

Google’s early stake is the best-documented ownership term in the cap table, and the later dollars are the least itemized.

• April 2023: $300 million for about a 10% stake, per CNBC. A separate report put an earlier 2023 check near $400 million. Treat $300 million / ~10% as the figure Google’s later comments were anchored to.
• October 27, 2023: up to $2 billion as a convertible note — $500 million upfront, $1.5 billion over time — confirmed by an Anthropic spokesperson. Bloomberg reported the note was set to convert at the next funding round, and that a multi-year Google Cloud agreement signed before the financing was larger than the investment itself. That cloud contract is a purchase commitment, not part of the $2 billion.
• January 2025: more than $1 billion, taking Google’s total investment to around $3 billion, on top of the prior $2 billion and a reported ~10% stake at that time.
• March 11, 2025: Anthropic court filings, reported by The New York Times, put Google at 14%, capped at 15%, with no voting rights, no board seat, and no board observer. The same filings said Google had invested more than $3 billion and was set to invest another $750 million in September 2025 via convertible debt agreed in 2023.
• April 24, 2026: up to $40 billion. Anthropic said Google committed $10 billion in cash now at a $350 billion valuation, plus $30 billion more if performance targets are met. The Financial Times described the $350 billion figure as pre-money; Reuters described it as the valuation. That is below the February Series G post-money of $380 billion, so this was a negotiated strategic price, not the last venture round. Alongside the cash, Google Cloud was to provide 5 GW of capacity over five years. The FT said a five-year capacity deal could be worth about $200 billion. An April 6 agreement with Google and Broadcom, separate from the equity, covers multiple gigawatts of next-generation TPUs from 2027; a Broadcom filing put that at 3.5 GW.
• Series H, May 28, 2026: Bloomberg reported Google contributed several billion dollars as part of the up-to-$40 billion commitment. Anthropic’s own release itemizes $15 billion of previously committed hyperscaler money in the round, of which $5 billion is Amazon, and does not break out Google.

A secondary Dealroom note put cash invested at about $13.3 billion plus up to $30 billion still contingent. That is consistent with roughly $3 billion through early 2025 plus the $10 billion April tranche, but it is not an Alphabet disclosure. Alphabet’s non-marketable cost basis rose from $28.4 billion at December 31, 2025 to $47.6 billion at June 30, 2026 — a $19.2 billion increase across all private holdings, an upper bound, not an Anthropic-only number.

Compute, not equity: the draft prospectus obligates Anthropic to spend at least $111.1 billion with Google from April 2026 through July 2033, and to pay the difference if actual spend falls short.

Microsoft and Nvidia: “up to” checks that partially closed inside Series G

On November 18, 2025, Microsoft, Nvidia, and Anthropic announced a three-way partnership. Microsoft committed to invest up to $5 billion and Nvidia up to $10 billion. Anthropic committed to purchase $30 billion of Azure compute and to contract up to 1 GW more. Its Nvidia compute commitment was initially up to 1 GW on Grace Blackwell and Vera Rubin systems. Amazon remained the primary cloud and training partner. A person familiar told Reuters both investments were aimed at the next funding round, then expected above $300 billion.

Anthropic’s February 12, 2026 Series G release says the $30 billion round “includes a portion of the previously announced investments from Microsoft and NVIDIA.” It does not say the full $15 billion closed there. Series H’s $15 billion of previously committed hyperscaler money is identified only as including Amazon’s $5 billion.

Subsequent earnings coverage is firmer on Microsoft than on Nvidia. TechCrunch and TipRanks reported that Microsoft invested $5 billion in November 2025 and recorded a $3.2 billion gain on the stake in the quarter ended June 30, 2026, adding $0.33 to diluted EPS. Microsoft does not routinely mark the stake every quarter, so that gain is a point-in-time observable-price adjustment, not a full carrying-value disclosure. Nvidia’s funded amount is still only “a portion” of “up to $10 billion” in primary company language. In March 2026, Jensen Huang reportedly described that $10 billion commitment as likely Nvidia’s last direct investment before an IPO. Separate September 2026 reports said Nvidia was in talks to anchor up to $10 billion of an IPO; that is unconfirmed and is not closed capital.

The prospectus later sizes the Microsoft purchase obligation at $31.4 billion from November 2026 through May 2033, non-cancelable except for Microsoft’s uncured material breach — slightly above the $30 billion announced in November 2025, consistent with the extra gigawatt option being partly firmed up. Nvidia-based capacity also shows up through xAI: up to $84.5 billion through 2029, largely cancelable on 90 days’ notice. That is an Anthropic spending ceiling, not an Nvidia equity investment.

AMD, Salesforce, Zoom, SK Telecom, and the memory suppliers

AMD (July 22, 2026). AMD committed to a strategic equity investment of up to $5 billion in the future, tied to deployment milestones, not cash funded at announcement. Anthropic will deploy up to 2 GW of Instinct MI450 GPUs in Helios racks, first gigawatt in the first half of 2027. Reuters described the hardware sale as tens of billions of dollars. The draft prospectus says AMD agreed to buy up to $5 billion of stock and to supply computing capacity expected to exceed $20 billion. Direction matters: unlike AMD’s OpenAI warrant deal, AMD is buying Anthropic equity rather than issuing AMD stock to win the order.

Salesforce Ventures. First check was about $50 million in the May 2023 Series C, when Anthropic was valued a little above $4 billion. Salesforce Ventures says it participated in every subsequent round through at least Series G. Cumulative cash is not in a Salesforce filing. Marc Benioff has said the company spent hundreds of millions, and one report put the total above $300 million. Bloomberg, after Series H, put the stake at about $5 billion. Benzinga reported a $2.6 billion gain on strategic investments in a later quarter, tied to the Anthropic mark. A Dealroom estimate of a 4% stake worth $38.6 billion conflicts with the Bloomberg mark and should not be used; Dealroom multiplies a disclosed percentage by the headline valuation, and that percentage is not in a company filing. Benioff has said he expects the position to be worth tens of billions at an IPO and has talked about selling to retire debt. Salesforce is also a large customer — Benioff has projected about $300 million of token spend in 2026 — so the strategic return is distribution plus equity, not a cloud-capacity swap.

Zoom Ventures. About $51 million in the May 2023 Series C. A later report says Zoom added about $46 million of preferred stock between late January and April 2026, for roughly $97 million total. The same report marked the stake at about $1.27 billion at the start of 2026, when Anthropic was at $380 billion, and about $3.13 billion by the end of July 2026 at the $965 billion mark. An earlier June 2026 figure of about $1.3 billion predates the full Series H revaluation. No Zoom filing with a precise share count was in the sources reviewed.

SK Telecom. $100 million in August 2023 for roughly a 2% stake, since diluted to about 0.3%. Korean regulatory disclosures carried the stake at about 1.376 trillion won, reported as roughly $900 million, at the end of 2025, versus a 132.1 billion won book value at the first investment. SK Telecom made a further, undisclosed investment in a later round. After that, analysts estimated the position above 4 trillion won, reported as about $2.6 billion. SK Telecom has not disclosed the latest check size. This is equity in a model lab, not a compute take-or-pay.

Samsung, SK hynix, and Micron joined Series H as strategic infrastructure partners. Amounts were not disclosed. Anthropic framed them as memory, storage, and logic suppliers whose equity aligns them with Claude’s capacity ramp, not as cloud landlords. SK hynix is a different company from SK Telecom.

Cisco is named in secondary round recaps as a Series E participant. Check size was not disclosed in the sources used here.

Sovereign and other strategic money in the priced rounds

These are mostly straight equity at the round price, not convertible notes tied to chips. Individual check sizes are generally undisclosed except where noted.

Series F, September 2, 2025 — $13 billion at $183 billion post-money. Led by ICONIQ, co-led by Fidelity and Lightspeed. Named participants include GIC, Qatar Investment Authority, Ontario Teachers’ Pension Plan, Blackstone, BlackRock-affiliated funds, Goldman Sachs Alternatives, General Atlantic, Coatue, TPG, Altimeter, Insight, Jane Street, and others. MGX was reported in talks and was not on the final list.

Series G, February 12, 2026 — $30 billion at $380 billion post-money. Led by GIC and Coatue. Co-leads: D.E. Shaw Ventures, Dragoneer, Founders Fund, ICONIQ, and Abu Dhabi’s MGX. Participants include QIA, Temasek, JPMorganChase through its Security and Resiliency Initiative, Blackstone, BlackRock-affiliated funds, and a portion of the Microsoft and Nvidia commitments. Run-rate revenue at announcement was $14 billion. GIC’s step from Series F participant to Series G lead is the clearest sovereign escalation. Dario Amodei had earlier expressed reservations about Gulf capital; MGX’s co-lead shows that constraint had loosened by early 2026.

Series H, May 28, 2026 — $65 billion at $965 billion post-money. Led by Altimeter, Dragoneer, Greenoaks, and Sequoia, each reportedly above $2 billion. Co-leads include Capital Group, Coatue, D1, GIC, ICONIQ, and XN. Significant investors include MGX, Temasek, Blackstone, Brookfield, Fidelity, Baillie Gifford, DST, Jane Street, and NTT’s venture arm NTTVC. Of the $65 billion, $15 billion was previously committed hyperscaler money, including Amazon’s $5 billion. Run-rate revenue had crossed $47 billion earlier that month. The round is the last fully disclosed private mark before a confidential S-1 filed around June 1, 2026. IPO price talk near $2 trillion is banker and press reporting, not a closed financing.

What the accounting marks imply — and what they do not

Amazon’s $190.4 billion at June 30, 2026, against a $965 billion Series H, is about 19.7% if you divide carrying value by post-money. That is an upper-bound sketch, not an ownership percentage. Three adjustments cut against a clean division. Preferred is marked from observable funding prices but with a Level 3 process that includes discounts for lack of marketability, security rights, and time to liquidity. Notes are fair-valued separately, and part of the original $8 billion has already converted, so note fair value is not “$8 billion times the same multiple.” An ownership cap can force conversion into non-voting preferred rather than common, which changes both control and the discount. One secondary analysis claimed about 21%; another derived roughly 9% and 20% by two methods. Until the S-1 cap table, high teens is the right order of magnitude, not a precise stake.

The income-statement mark is narrower than the balance-sheet mark. The $50.5 billion Q2 earnings boost is the preferred revaluation. The $92 billion note gain is in AOCI and does not flow through net income the same way. A $2 trillion IPO would not create $190 billion of new earnings; a large piece is already marked. Scaling the June carrying value from $965 billion to $2 trillion implies roughly $395 billion, about $200 billion above the June mark, only if the stake percentage and discounts stay constant.

Alphabet does not name Anthropic. As of June 30, 2026, non-marketable equity securities under the measurement alternative had a carrying value of $124.3 billion, up from $64.1 billion at December 31, 2025. The filing says those investments “primarily consist of our investment in a private company,” and $87.9 billion of the bucket was remeasured in the second quarter, mostly as Level 2, meaning an observable transaction rather than a pure model. Bloomberg identified that investment as Anthropic and put the stake at about $124 billion. SpaceX is not in this bucket: after its June 2026 listing, restricted SpaceX shares sit in marketable securities (about $80 billion restricted, with the stake reported around $94 billion). Q2 other income of about $98 billion, including $99 billion of net equity-securities gains, is SpaceX plus the private company, not an Anthropic-only gain.

If nearly all of the $124.3 billion is Anthropic at the $965 billion round, the implied stake is about 13%. That sits just under the 14% disclosed in early 2025 and the 15% cap. Dilution from the $30 billion Series G and $65 billion Series H would have pushed a static 14% lower; the April $10 billion and the Series H follow-on pulled it back up. A measurement-alternative mark can also embed a small discount to the headline price. Roughly 13%, capped at 15%, non-voting, is the defensible inference. It is not a second source of a precise share count.

Anthropic’s $34 billion 2025 note charge is the same valuation event seen from the issuer. As Amazon’s and Google’s conversion value rose, Anthropic recorded a loss. That charge confirms the notes were still outstanding and in the money relative to their conversion terms. It does not, by itself, give a share count. Amazon marking notes from $42.2 billion on March 31, 2026 to $97.9 billion on June 30 is the investor-side version of that remeasurement across the Series H jump from $380 billion to $965 billion.

Microsoft’s $3.2 billion quarterly gain on a reported $5 billion cost, over a period when the post-money mark went from the Series G area to $965 billion, implies the funded stake was small — on the order of 1% or less if the gain is a partial mark toward the Series H price. That is an inference, not a disclosed percentage. Nvidia has not published a comparable mark.

What is cash, what is a promise, what is Anthropic’s bill

Partner	Cash funded (best disclosed)	Still contingent equity	Anthropic’s purchase obligation	Instrument / control
Amazon	$18B through Jun 30, 2026	Up to $15B, drawable only as compute milestones are met; notes or post-IPO stock	≥$110B, May 2026–Apr 2036, pay-regardless; up to 5 GW Trainium	Convertible notes + non-voting preferred; ownership cap waivable; no disclosed %
Google / Alphabet	~$3B by early 2025 + $10B in Apr 2026; several billion more inside Series H. ~$13B is a secondary estimate, not a filing	Up to $30B more, performance-gated, inside a $40B ceiling	≥$111.1B, Apr 2026–Jul 2033, shortfall payment; 5 GW over five years plus a separate multi-GW Broadcom/TPU deal from 2027	Early convertibles; 14% in Mar 2025, cap 15%, no vote, no board
Microsoft	Announced up to $5B; earnings coverage says $5B was invested; a portion closed in Series G. $3.2B gain in quarter ended Jun 30, 2026	Any unfunded remainder of the “up to” not itemized	$30B Azure announced; prospectus $31.4B, Nov 2026–May 2033, cancelable only for material breach; option for ~1 GW more	Equity; % undisclosed
Nvidia	Up to $10B; a portion inside Series G. Funded amount not itemized	Remainder of the “up to”; reported IPO-anchor talks of up to $10B are not closed	Up to 1 GW Grace Blackwell / Vera Rubin. Separate xAI path: up to $84.5B through 2029, mostly 90-day cancelable	Equity commitment, not a disclosed carrying value
AMD	$0 at announcement	Up to $5B, milestone-gated, “in the future”	>$20B of capacity in the prospectus; up to 2 GW MI450 from H1 2027; Reuters: tens of billions of servers	Future equity, not warrants in AMD stock
Salesforce Ventures	~$50M in 2023; hundreds of millions cumulative (one report: >$300M)	Undisclosed pro rata	None disclosed. Customer, ~$300M token spend projected for 2026	Preferred / venture equity. Bloomberg mark ~$5B after Series H
Zoom	~$51M in 2023; ~$97M if the 2026 follow-on is included	Undisclosed	None disclosed	Preferred. Secondary mark ~$3.1B at Jul 2026
SK Telecom	$100M in Aug 2023; further undisclosed round	Undisclosed	None disclosed	~2% at entry, diluted to ~0.3%; carried ~$900M at end-2025; later estimate ~$2.6B
GIC, QIA, MGX, Temasek, OTPP	Inside $13B / $30B / $65B rounds. Check sizes not disclosed. GIC led Series G; MGX co-led Series G and joined Series H	Standard pro rata, not chip-gated	None as sovereigns	Priced equity at $183B, $380B, $965B
Samsung, SK hynix, Micron	Undisclosed Series H checks	Undisclosed	Supply relationship, not a dollar take-or-pay in the sources	Strategic equity alongside memory supply

Broadcom’s $161.2 billion of largely non-cancelable equipment leases, and the xAI/Nvidia capacity, belong in the infrastructure stack, not the investor stack. Together with Google, Amazon, and Microsoft they are most of the at least $518 billion decade-long buildout in the draft prospectus, about 80% non-cancelable or payable regardless of use.

The competitive implication is structural. A new model lab cannot replicate “get Amazon and Google to fund you” without also signing pay-regardless capacity that now exceeds $100 billion at each of those two clouds. The equity is the smaller side of the trade, and it is deliberately non-controlling: Google is capped at 15% with no vote, and Amazon holds notes and non-voting preferred. The investors get a marked-up stake and a locked customer. Anthropic gets chips it cannot currently buy elsewhere, and a GAAP loss every time those notes are worth more. The S-1, once public, is the document that replaces the 13% and high-teens inferences with a share count.

---

Recent Findings Supplement (October 2026)

Anthropic’s April–May 2026 deals with Amazon and Google layered new equity commitments on top of massive, largely non-cancelable compute purchase obligations, while the May Series H round brought in memory-chip makers as strategic equity investors.[1][2]

These arrangements separate cash/equity investments (which create ownership stakes marked to fair value) from long-term cloud/compute purchase commitments (which are contractual spending obligations, often with “must-pay” or non-cancelable features). The September 2026 reporting on Anthropic’s confidential IPO prospectus (filed June 2026) provides the first detailed breakdown of these obligations.[3]

Amazon’s April 2026 Expansion: Cash + Compute Tie-Up

Amazon announced it would invest up to $25 billion more in Anthropic ($5 billion immediate + up to $20 billion future, subject to commercial milestones), on top of its prior ~$8 billion. This is structured as convertible notes and nonvoting preferred stock.[1][4]

In exchange, Anthropic committed to spend more than $100 billion over 10 years on AWS technologies (including Trainium chips), with access to up to 5 GW of capacity. The prospectus later quantified Amazon’s share at $110 billion (May 2026–April 2036), with a must-pay shortfall clause.[3]

• This deepens Amazon’s dual role as investor and primary cloud/training partner.
• At June 30, 2026, Amazon carried its Anthropic position at $190.4 billion ($97.9 billion notes + $92.5 billion preferred), after ~$18 billion total invested; it recorded a $50.5 billion Q2 upward adjustment on the preferred shares (part of $53.4 billion other income).[5]

Implication for competitors: Pure compute providers without equity upside face harder economics; Amazon’s structure creates a self-reinforcing loop where usage drives both revenue and stake value.

Google’s April 2026 Parallel Commitment

Days after Amazon’s announcement, Google committed up to $40 billion ($10 billion cash at a $350 billion valuation + up to $30 billion contingent on performance targets).[2][6]

This supports expanded TPU capacity (up to 5 GW via a Google-Broadcom-Anthropic arrangement). The prospectus quantified Google’s infrastructure obligation at $111.1 billion (April 2026–July 2033), also with a must-pay shortfall provision.[3]

Alphabet’s Q2 2026 filings showed non-marketable equity securities at $124.3 billion (primarily one unnamed private company, widely reported as Anthropic), with $87.9 billion remeasured and $77.5 billion gross unrealized gains in the quarter.[7]

Implication: Google’s stake (~14% in some analyses) provides similar mark-to-market upside as Amazon’s, but the compute commitments lock in usage regardless of short-term demand fluctuations.

May 2026 Series H Round: $65 Billion at $965 Billion Valuation

Anthropic closed a $65 billion Series H round at a $965 billion post-money valuation (co-led by Altimeter Capital, Dragoneer, Greenoaks, Sequoia; participants included Capital Group, Coatue, and others).[8][9]

This included $15 billion from previously committed hyperscaler tranches (e.g., $5 billion from Amazon). Strategic infrastructure partners Micron, Samsung, and SK hynix took equity stakes alongside their memory/supply roles.[10]

• Run-rate revenue reportedly exceeded $47 billion earlier in May.
• This round more than doubled the February 2026 valuation (~$380 billion post-money) and positioned Anthropic above OpenAI’s then-reported valuation.

Implication: The inclusion of memory suppliers as equity holders signals vertical integration in the supply chain; new entrants must compete on both capital and hardware ecosystem access.

September 2026 Prospectus Disclosures: $518 Billion Total Commitments

Reporting on Anthropic’s confidential IPO prospectus (June 2026 filing) revealed ~$518 billion in decade-long infrastructure commitments across six partners, with ~80% non-cancelable or payable regardless of usage.[3][11]

Key breakdowns (compute/infrastructure obligations, distinct from equity cash):
- Google: ≥$111.1 billion.
- Amazon: $110 billion.
- Microsoft: $31.4 billion (Nov 2026–May 2033; non-cancelable except material breach).
- Broadcom: ~$161.2 billion equipment leases (largely non-cancelable).
- xAI arrangement: Up to $84.5 billion Nvidia-based capacity through 2029 (mostly cancelable with 90-day notice).
- AMD: >$20 billion capacity; AMD to purchase up to $5 billion in Anthropic shares.

Anthropic also highlighted Micron/Samsung/SK hynix memory partnerships. Revenue routed through Amazon and Google reached 47% in the prior year.[12]

Implication: These fixed obligations de-risk suppliers’ capacity investments but create substantial balance-sheet pressure on Anthropic; competitors without similar locked-in demand visibility face higher risk in building out infrastructure.

Accounting Marks Enable Valuation Inference

Amazon and Alphabet’s Q2 2026 filings (reported July–September 2026) provide observable transaction-based marks that allow back-calculation of implied ownership and fair value.[5]

• Amazon’s $190.4 billion carrying value at June 30, 2026 (post ~$18 billion invested) against the $965 billion Series H valuation implies roughly 19–20% effective exposure (exact percentage not disclosed; instruments include notes and nonvoting preferred).
• Alphabet’s $124.3 billion primarily-Anthropic bucket (with large Q2 remeasurements) aligns with ~14% historical stake estimates.
• At a hypothetical $2 trillion IPO valuation, Amazon’s position could approach $395 billion and Alphabet’s ~$280 billion (illustrative; subject to dilution, instrument terms, and final cap table).[13]

Anthropic’s own 2025 results showed a ~$42 billion net loss, of which ~$34 billion was a non-cash remeasurement charge on convertible notes tied to rising valuations.[14]

Implication: These marks turn private-company gains into reported earnings for investors, creating volatility upon IPO or further rounds; entrants must model similar accounting effects if using convertible or preferred structures.

No material new public details emerged after April 3, 2026, on Microsoft equity stakes, Nvidia direct investments, Salesforce Ventures, Zoom, SK Telecom, or sovereign backers. Earlier commitments (e.g., Microsoft Azure compute) appear reflected in the prospectus figures. All figures are in USD.

Anthropic has run two company-facilitated employee liquidity events, both priced at or just below the nearest primary round, while the open secondary market has repeatedly marked the stock far above those company-set prices—and then Anthropic tried to shut most of that market down. Completed employee sales are small relative to demand. Almost every “market” price after mid-2026 is an indication, not a deep clearing price.

Company-run employee liquidity

May 2025 buyback — par to Series E, company as the buyer. Anthropic’s first employee liquidity program was a company repurchase, not a sale to outside funds. The Information reported on May 2, 2025 that Anthropic would buy shares from current and former employees at a $61.5 billion valuation, matching the March 3, 2025 Series E led by Lightspeed ($3.5 billion raised, $61.5 billion post-money). Eligible sellers were people who had worked at the company at least two years; they could sell up to 20% of equity, capped at $2 million per person, at $56.09 per share—the Series E issue price. The program was expected to close by the end of May 2025. Total dollars bought back were not disclosed; contemporaneous coverage described “hundreds of millions” as a plausible outcome, not a confirmed figure. Versus the nearest priced round, the premium/discount was zero on the preferred price the company itself had just set. Reliability is medium-high on price, valuation, tenure, percentage, and cap (multiple outlets citing The Information); low on aggregate size. Liquidity was real for eligible employees but tightly rationed.

February–early April 2026 tender — $350 billion pre-money, undersubscribed because employees would not sell. Bloomberg reported on February 4, 2026 that Anthropic was planning an employee tender at a valuation of at least $350 billion, the same pre-money figure then being discussed for a primary round. On February 23–24, Bloomberg reported the company had opened a sale for some current and former employees at about $350 billion, with outside investors—not Anthropic—lined up to buy $5–6 billion of stock, final size depending on how many employees opted in. Eligibility: current and former employees with at least 12 months of tenure. No public per-person dollar cap or percentage cap was reported. The primary closed February 12, 2026: $30 billion Series G at a $380 billion post-money valuation, led by GIC and Coatue (Anthropic’s own announcement). The tender therefore priced at the Series G pre-money mark—about an 8% discount to the $380 billion post-money figure, and at par to the pre-money. Bloomberg reported on April 8, 2026 that the tender had closed the prior week at that same $350 billion pre-money level, but employees sold less than investors had capital for. Some buyers got full allocations; others did not. Exact dollars transacted were not disclosed. Forge’s funding table, as reproduced by Yahoo Finance, lists a “Tender Offer 1” dated April 8, 2026 against a $380 billion valuation with no issue price shown—treat that as a platform label of the same event, not a second price. A per-share tender price was not publicly reported. Reliability: high that the event happened at the Series G pre-money mark and fell short of $5–6 billion of demand; low on the exact dollars that cleared. The mechanism that matters: this was a curated syndicate price, not an auction. Employees who declined were, within weeks, looking at secondary indications and inbound primary interest far above $350 billion.

Not yet a transaction: IPO secondary component. In late August 2026, The Information reported (via Reuters and others) that Anthropic was considering letting existing shareholders sell stock in the IPO itself, possibly with lockups longer than 180 days, and possibly requiring rank-and-file sales through 10b5-1 plans. Bankers had discussed valuations around $1.5 trillion; later Reuters reporting on the draft prospectus described an expected valuation that could exceed $2 trillion. Size, who can sell, and eligibility were not set. This is a contemplated structure, not a completed secondary. Reliability: medium on the fact of the discussion; low on terms.

Forced investor secondary: the FTX block

The largest disclosed secondary before the employee programs was not employee stock. It was the FTX bankruptcy estate selling the stake Sam Bankman-Fried’s entities bought in 2021.

• March 2024: court filings showed a deal to sell about 29.5 million shares for roughly $884 million, about $30 per share. Buyers included ATIC Third International (an Abu Dhabi/Mubadala-linked vehicle, about $500 million / 16.6 million shares), Jane Street (about $100 million), Fidelity-managed funds (about $50 million), and a long list of others. Hiive-related SPVs and Anthropic-affiliated vehicles also appeared among buyers in court documents summarized by Decrypt. Subject to bankruptcy-court approval.
• May–June 2024: the estate sold the remaining about 15 million shares at the same about $30 per share, for more than $450 million. G Squared took about 4.5 million shares for about $135 million. Combined proceeds from the original $500 million investment were reported around $1.3 billion.

Implied equity value on the $30 prints was in the mid-teens of billions, in the neighborhood of Anthropic’s then-latest primary (Menlo-led financing around $18.4 billion in early 2024, depending on the source). These were court-supervised block sales of a forced seller, so they are more reliable as executed prices than platform indications, but they are not a continuous market and they reflect a distressed holder, not employee optionality. Eligibility was “whoever the estate and Anthropic’s transfer process would accept,” not an employee program.

Secondary-market indications (not company tenders)

Treat every figure below as publicly reported or estimated. None is a continuous, deep market. Anthropic’s own May 2026 support-page warning said transfers not approved by the board are void and will not be recognized, that SPVs are not permitted to acquire Anthropic stock, and that forward contracts and unauthorized platforms do not convey recognized ownership. Named in contemporaneous coverage: Open Doors Partners, Unicorns Exchange, Pachamama Capital, Lionheart Ventures, Hiive (new offerings), Forge Global (new offerings), Sydecar, and Upmarket. Forge later said it had been included erroneously. Hiive acknowledged the company’s concerns. That policy is the binding constraint on liquidity: a print that never receives board approval is a contract with a counterparty, not a share.

Platform and broker marks, with the nearest primary:

• Late 2024, Notice. A Notice.co subscriber report dated December 23, 2024 showed a 60-day consensus around $44.56 and stated Anthropic did not allow direct transfers—only indirect exposure via SPVs or forwards. No company valuation was attached in that extract. Nearest primary context was the path from the early-2024 ~$18 billion area toward the March 2025 $61.5 billion Series E. Reliability: low; a single delayed consensus screen, and the structure was explicitly not direct stock.
• January 2026, Caplight via Augment. Before the Series G, Caplight-tracked secondary pricing was described as about $150–175 per share, roughly in line with the September 2, 2025 Series F at $183 billion post-money ($13 billion raised, led by ICONIQ, co-led by Fidelity and Lightspeed; Series F-1 issue price $140.97 on Forge’s table). By the time the $350 billion round was in market, the same write-up put secondary pricing near $300 per share—a large premium to the $183 billion round and still below the incoming $350 billion pre-money. Reliability: medium as a directional Caplight read; not a published trade blotter.
• April 2026, Forge. Augment, citing Forge, said Anthropic’s implied mark had crossed $1 trillion within weeks of the tender close—versus the $350 billion tender and the $380 billion February post-money. That is roughly a 2.6–2.9x premium to the company-set marks. Bloomberg was also reported as saying the company was fielding investor interest around $800 billion outside the tender syndicate. Reliability: medium for “the screen moved above $1 trillion”; low for executable size. Scarcity, not a thick book, drove the print.
• May 28, 2026 primary anchor. Anthropic announced a $65 billion Series H at a $965 billion post-money valuation, led by Altimeter, Dragoneer, Greenoaks, and Sequoia. Forge’s table shows Series H-1 and H-2 at $589.01 per share. A later secondary write-up that called a “Series H-1 secondary” at $589 was describing the round price, not a separate discount or premium trade. Versus the April tender, this primary was about 2.8x the $350 billion pre-money mark in under two months.
• Early June 2026, Notice. A secondary consensus near $625.26, about 6% above the $589.01 Series H price, with an implied market cap just over $1 trillion, and reported buy interest about 2.9x sell interest. Reliability: low-to-medium; algorithmic consensus, not a guaranteed fill, and published in a promotional investor note.
• June 2026, Caplight (90-day delayed). Caplight’s public page showed a MarketPrice of $848.86 per share “as of June 2026,” with recent months withheld for clients. At the ~1.64 billion share count implied by $965 billion / $589.01, that price would imply well above $1 trillion. Caplight itself labels MarketPrice as an estimate that may be incomplete and is not a price at which you can necessarily transact. An April 2026 Caplight commentary (buyer vs. seller interest, not a price) said Anthropic had about $2 billion of buy interest versus $765 million of sell interest over the prior year, and that Q1 2026 was heavily buyer-dominated.
• July 9–10, 2026. Business Insider, Quartz, and others reported shares changing hands at an implied $1.2 trillion on Caplight, with Caplight CEO Javier Avalos calling Anthropic the most sought-after name the venture secondary market had seen, and Rainmaker Securities’ Glen Anderson saying completed trades were rare because almost no one was selling. That is about a 24% premium to the $965 billion Series H, six weeks later. A separate anecdote in the same news cycle: one shareholder offered stock at $1.15 trillion; a growth fund was said to have bid $1.05 trillion; another buyer saw a $960 billion offer disappear within a day. Reliability: medium that indications clustered around $1.0–1.2 trillion; low that any single anecdote cleared in size. The mechanism is a one-sided book: price is set by the marginal desperate buyer, not by volume.
• September 2, 2026. Benzinga, citing OpenVC’s NYSE OpenVC Unicorn Index via David Shapiro, put an implied valuation at $1.17 trillion, about 21% above the $965 billion primary. Same scarcity story. Reliability: medium as an index indication; not a trade print.
• October 1–3, 2026 screens (conflicting). Moonberg cited a Nasdaq Private Market mark of $1.36 trillion, about $829 per share, 41% above Series H. Stock Analysis, sourcing a price “from Clarity” (formerly Hiive), showed $1,075.57 as of October 3, 2026, a 52-week range of $166.24 to $1,225.88, and an “implied valuation” of about $910 billion—below the last primary—which does not reconcile cleanly with the $1,075 price if the Series H share count is the right denominator. Forge’s own IPO page, updated around October 3, 2026, listed Anthropic Forge Price as “not available” and market activity as “medium.” These three screens cannot all be the clearing price. Reliability: low. Use them as evidence that vendors disagree, not as a single mark.

SPVs and forwards. Anthropic has said it does not permit SPVs to acquire its stock and that transfers into an SPV are void under its transfer restrictions. Forwards—a promise to deliver shares or cash value later—do not put the buyer on the cap table. DefiLlama’s pre-IPO page, as of October 3, 2026, still listed tokenized or SPV-style wrappers (for example Colb and PreStocks) with quoted prices above the Series H share price. Those quotes are not evidence of recognized ownership. Fee stacks on layered SPVs (management fee plus carry at each layer) mean the economic entry price can be materially worse than the headline valuation. Reliability of any SPV or forward “price”: low, and legal recognition is the company’s explicit position that unapproved transfers are void.

Premiums and discounts versus the nearest priced round

Event	Date	Price / valuation	vs. nearest primary	Size	Who	Reliability
FTX block 1	Mar 2024	~$30/share; ~$884M	Roughly in line with ~$15–18B primary area	~29.5M shares	Estate to funds (ATIC, Jane Street, Fidelity, others, some Hiive vehicles)	High on filings; forced seller
FTX block 2	May–Jun 2024	~$30/share; ~$452M+	Same	~15M shares	Estate; G Squared largest named buyer	High on filings
Employee buyback	May 2025	$56.09; $61.5B	Par to Mar 2025 Series E	Undisclosed; $2M cap, 20% max, ≥2 years	Company buys from current/former employees	Medium-high on terms; size unknown
Employee tender	Opened late Feb 2026; closed ~early Apr 2026	$350B pre-money; per-share not disclosed	Par to Series G pre-money; ~8% below $380B post-money	Demand $5–6B; filled amount lower, undisclosed	Outside investors buy from current/former employees, ≥12 months	High on valuation and shortfall; size unknown
Forge indication	~Apr 2026	Implied >$1T	~2.6x+ vs. $380B post / $350B tender	Thin	Anonymous secondary	Medium as a screen
Series H (primary, for anchor)	May 28, 2026	$589.01; $965B post	—	$65B primary	Led by Altimeter, Dragoneer, Greenoaks, Sequoia	High (company)
Notice consensus	Early Jun 2026	~$625; ~6% above $589; ~$1T+	Small premium to Series H	Indication	Notice algorithmic consensus	Low-medium
Caplight / broker	Jul 2026	Implied $1.2T	~24% above $965B	Rare closes	Caplight, Rainmaker anecdotes	Medium on level; low on volume
OpenVC index	Sep 2, 2026	Implied $1.17T	~21% above $965B	Index	OpenVC via Benzinga	Medium as index
NPM / Clarity screens	Oct 2026	$1.36T (~$829) vs. $1,075.57 last price vs. Forge “not available”	Screens disagree	Indications	NPM, Clarity/Hiive, Forge	Low; do not reconcile

What the gap actually means

Company tenders cleared at the last primary (or a small discount to post-money) because Anthropic and its syndicate set the price and employees were not the marginal sellers in size. The April 2026 shortfall is the cleanest evidence: at $350 billion, buyer capital exceeded employee supply. Within weeks, revenue run-rate disclosures and outside bids repriced the story, and by July the only shares that moved were moving near $1.2 trillion because holders would not sell into an expected IPO. That is not proof the company is “worth” $1.2 trillion. It is proof that a one-sided book with transfer restrictions, ROFR, and a company that has declared many SPV and forward structures void will print optimistic indications on tiny volume.

For anyone trying to buy or underwrite this:

• The only prices with real eligibility rules and a company on the other side are the May 2025 buyback (par, tight caps, company cash) and the spring 2026 tender (par to pre-money, employees rationed supply). Both are closed.
• Platform marks from Forge, Caplight, Hiive/Clarity, EquityZen, Notice, and NPM are delayed, algorithmic, or anecdotal. They routinely disagree by hundreds of billions. Minimums on the major venues are roughly $25,000–$100,000+ for hot names, with fees that can run several percent per side, and ROFR or board consent can kill the trade after you have committed.
• SPV and forward “access” is the product Anthropic has specifically said it will not recognize. A buyer in those structures is underwriting counterparty and legal risk, not Claude equity.
• The next real liquidity print, if the August 2026 discussions hold, is an IPO secondary whose size, sellers, and lockups are still unset—against banker talk that has ranged from about $1.5 trillion to above $2 trillion, far above every completed employee sale.

---

Recent Findings Supplement (October 2026)

No new employee tender offers, company buybacks, or large-scale secondary sales have been reported since the April 2026 tender (published details after April 3, 2026, primarily reference that earlier event or ongoing platform activity).[1][2]

The April tender (launched around February, closed early April) was priced at a $350 billion pre-money valuation (matching the February 2026 Series G primary round; ~$380 billion post-money including the $30 billion raised). Outside investors bought shares from eligible current/former employees (minimum 12 months tenure); up to $5–6 billion was lined up in demand, but actual volume fell short due to limited employee participation as holders retained shares ahead of a potential IPO. The total transacted amount was not publicly disclosed but was below the targeted range.[3][4]

This represented a significant discount to subsequent primary and secondary pricing amid rapid revenue growth and IPO anticipation. No comparable new tender or buyback activity appears in July–October 2026 reporting.

Secondary-market indications on platforms (Forge, Caplight, Hiive/Clarity, EquityZen, etc.) show strong premiums to the May 28, 2026 Series H primary round ($965 billion post-money, ~$589/share), driven by scarcity and IPO expectations rather than high transaction volume.[5][6]

• July 2026 (primarily Caplight data reported July 9): Shares trading at an implied ~$1.2 trillion valuation (roughly 24% premium to the $965 billion primary). Caplight CEO described Anthropic as “the most sought-after company the venture secondary market has ever seen.” Actual completed trades remained rare due to very limited seller supply; Rainmaker Securities reported similar ~$1.2 trillion levels in sparse transactions. This marked a ~550% year-over-year increase and a flip ahead of OpenAI (~$908 billion on the same platform).[7][8][9]
• September 2026: OpenVC Unicorn Index implied ~$1.17 trillion as of September 2 (~21% premium to the May primary). High buyer demand met with holder reluctance ahead of IPO.[10]
• As of early October 2026 (Clarity/Hiive platform via stockanalysis.com, October 3 data): Last price ~$1,075.57/share (implied valuation in the ~$910 billion range on some calculations, though 52-week high reached $1,225.88/share). Forge and other platforms note medium activity but limited or no direct facilitation without company approval; indicative prices unavailable or sparse in some reports.[11][6]

These platform marks are indicative or based on limited matched trades/SPV interests and carry low liquidity/transferability risk—Anthropic has explicitly warned that unauthorized sales/transfers (including via certain platforms, SPVs, or forwards) are void and not recognized on its books.[12]

SPV, forward-contract, and synthetic pricing (e.g., crypto perpetuals) implies even higher valuations but carries substantial reliability and legal risks. September 2026 reports noted crypto perpetual futures (Hyperliquid and others, e.g., Binance ANTHROPIC/USDT) pricing the company at ~$2 trillion (or higher in some prints), more than double the May primary. These are purely synthetic derivatives with no actual share ownership or delivery rights. Company policy prohibits SPVs and unauthorized forwards; any such structures risk being void.[13][14]

IPO-related developments (August–September 2026) include discussions of potential insider/secondary sales in the offering itself, though no execution details have emerged. Reports indicated consideration of allowing existing shareholders (employees/early investors) to sell alongside new shares in a potential $1.5 trillion (or higher, up to $2 trillion target) IPO, paired with extended lockups for some holders. The company confidentially filed for IPO in June 2026, with targets shifting toward October/November 2026 (or pre-Thanksgiving). No priced secondary component has been confirmed.[15][16]

Key implications and caveats: Secondary prices reflect extreme scarcity and forward-looking IPO hype (revenue run-rate cited up to $65 billion in some reports) more than cleared market volume. Liquidity remains very low; most holders appear to be waiting for the IPO. All secondary data should be treated as estimated/indicative with limited reliability for large positions. No updates contradict the April tender details or introduce new priced employee liquidity events after mid-2026. For the most current platform quotes, direct verification on approved channels (subject to company approval) would be required, as unauthorized routes carry high risk of invalidation.

Anthropic’s private marks have not gotten more expensive. From the March 2025 Series E to the May 2026 Series H, the post-money price rose about 16x ($61.5 billion to $965 billion) while the company-stated run rate rose about 47x (roughly $1 billion to $47 billion). The implied valuation-to-run-rate multiple compressed from the high-40s/50s into the low-20s. Later investor updates pushed the run rate to $65 billion by the end of July 2026 and, per the New York Times, past a $100 billion pace by mid-September — which would put the stale $965 billion mark under 10x, and a discussed $2 trillion IPO price back around 20–31x depending on which revenue snapshot you use.

The mechanism is simple and easy to misuse: in this market the price has roughly tracked the latest monthly annualization, not trailing GAAP sales. That is why a 16x valuation jump can still look like “multiple compression.”

1. Marks matched to run rate

Figures below use post-money primary prices unless noted. Multiples are valuation divided by the run rate disclosed at or immediately around the mark. They are not enterprise-value / next-twelve-months multiples.

Date	Transaction	Price	Run rate used	Implied multiple	Confidence
Mar 3, 2025	Series E, $3.5B, Lightspeed-led	$61.5B post	~$1B at the start of 2025; press at the time used a figure that implied ~58x	~47–62x; contemporaneous press 58x	Valuation high. Run rate medium: company later said ~$1B at the start of 2025; VentureBeat’s 58x implies they used a bit above the December 2024 $1B print after a reported ~30% rise in the first two months
Sep 2, 2025	Series F, $13B, ICONIQ-led	$183B post	Over $5B in August 2025 (company)	≤36.6x	High. Both numbers are in Anthropic’s own announcement
Feb 12, 2026	Series G, $30B, GIC/Coatue-led	$380B post; $350B pre (FT)	$14B (company)	27.1x post / 25.0x pre	High
Late Feb 2026	Employee tender, $5–6B lined up	~$350B (pre-money level of the Series G)	$14B	~25x	Valuation medium (Bloomberg sources; terms not company-confirmed). Run rate high
Apr 6–20, 2026	No closed round. Company confirmed the run rate in compute-deal posts. Press reported investor offers near $800B	Offers, not a clearing price	Surpassed $30B, up from ~$9B at end-2025 (company)	~27x if an $800B offer is taken at face value	Run rate high. Price low — offers, not a close
May 28, 2026	Series H, $65B	$965B post	Crossed $47B earlier that month (company)	20.5x	High
End-Jul 2026	Investor update, no new primary	Last primary still $965B. Separate IPO talk around $2T	$65B	14.8x on the stale primary; ~30.8x on a $2T talk price	Run rate high (Bloomberg, CNBC, Reuters, all citing the investor update). $2T is a reported target, not a transaction
Mid-Sep 2026	No new primary. NYT, via Bloomberg/Axios	Same stale marks	More than $100B annualized pace expected this year	~9.7x on $965B; ~20x on $2T	Medium. “People familiar,” company declined comment. “This year” is ambiguous between exit run rate and a full-year projection

Company posts that anchor the run-rate ladder: Series F (start-2025 ~$1B, August 2025 over $5B), Series G ($14B), the April 6 and April 20 compute announcements (over $30B, versus ~$9B at end-2025), and Series H ($47B). The July $65B figure is an investor update, not a blog post, but three outlets confirmed it independently. The September “over $100B” figure is one step weaker.

A cross-check that the May multiple is not just a one-month spike: preliminary Q2 2026 revenue was more than $11.5 billion, so annualizing the quarter gives more than $46 billion. $965 billion / $46 billion is about 21x — essentially the same as 20.5x on the May monthly run rate. By late spring, the monthly and quarterly pictures had converged. They had not converged in 2025.

2. Why “run rate” and “revenue” are different numbers

Anthropic’s own metric, as Reuters and Bloomberg describe it, annualizes a short recent period — typically the latest month times 12. It is not contracted SaaS ARR, and it is not GAAP revenue. During this ramp the gap is large enough to change the multiple by a factor of two.

• Draft IPO filing, reported by Reuters on September 29, 2026: 2024 recognized revenue about $386 million; 2025 recognized revenue nearly $4.6 billion (about 12x). PitchBook/Morningstar, using the same filing, has the same pair and adds Q1 2026 revenue of $4.73 billion and preliminary Q2 above $11.5 billion.
• Exit-2025 run rate was about $9 billion (company, restated in April 2026 posts; Bloomberg, January 21, 2026). So the year-end run rate was roughly 2x what the company actually booked in 2025.
• CNBC’s August 17 piece said the company generated “roughly $10 billion” for all of 2025. That conflicts with the prospectus figure of ~$4.6 billion and looks like a run-rate/booked mix-up. The filing wins.
• Q1 2026 booked revenue ($4.73 billion) already exceeded all of 2025. That is the signature of a curve so steep that any trailing-twelve-month multiple is a historical artifact.

There is a second definitional fight, and it is not resolved. In an April 13, 2026 internal memo reported by The Verge and covered by PitchBook and others, OpenAI’s then chief revenue officer, Denise Dresser, argued Anthropic grosses up Amazon and Google revenue-share, and that this overstated a then-cited ~$30 billion run rate by about $8 billion (implying ~$22 billion on OpenAI’s net basis). The New York Times later noted that OpenAI has disputed Anthropic’s revenue calculations. Bloomberg has also said the two firms may not measure run rate the same way.

The prospectus cuts against a claim that gross-up is most of the story, at least for 2025. Reuters’ reading of the filing: about $3.8 billion of 2025 revenue was usage-based and $789 million was subscription; 47% of sales ($2.16 billion) went through Amazon and Google cloud marketplaces; distribution fees were about $351 million (Reuters’ analysis, roughly 16 cents per marketplace dollar, or ~8% of total 2025 revenue). Both gross and net treatments can be GAAP-compliant. They are not comparable across labs, and an $8 billion gap on a $30 billion headline is an adversary’s analysis, not an audited adjustment. Treat cross-lab “who is bigger” rankings as medium confidence until both S-1s use the same basis.

A third trap: much of this revenue is consumption, not recurring seats. The filing says Anthropic expects consumption-based revenue to remain “the substantial majority.” Calling the metric ARR imports SaaS stickiness the contracts do not have. Two unnamed customers each supplied about 12% of 2025 revenue, and the company warned that many large customers are not on long-term commitments.

3. API versus consumer, and what Claude Code actually is

This is an enterprise token business with a coding wedge, not a ChatGPT-style consumer company.

• At the February 2026 Series G, the Financial Times reported that about 80% of the $14 billion run rate came from enterprise customers. Reuters had the same 80% business mix in October 2025, when the run rate was approaching $7 billion and the internal year-end target was $9 billion.
• The 2025 filing split — usage-based ~$3.8 billion versus subscription $789 million — is about 83% / 17%. That is recognized revenue, not run rate, and it predates the 2026 acceleration. Directionally it matches the 80% enterprise figure.
• Claude Code, generally available in May 2025, is the product the company cites as the breakout, but company-confirmed dollars show it as a mid-teens share, not the majority:
  • September 2, 2025: over $500 million run rate (company), against a total above $5 billion → about 10%.
  • October 2025: nearly $1 billion, per a Reuters source, not a company blog line.
  • February 12, 2026: over $2.5 billion, more than doubled since the start of 2026; business subscriptions had quadrupled; enterprise was more than half of Claude Code revenue (company and Reuters). Against a $14 billion total, that is about 18%.
• Later Claude Code figures in the $8–15 billion range (May–August 2026) come from analyst and alt-data writeups, not Anthropic. Do not use them in a multiple. The economically important point is the one the company does state: Claude Code pulls engineering teams in, and the bulk of dollars still clear as API and enterprise usage, including through Bedrock and Vertex. Marketplace distribution was 11% of revenue in 2023, 32% in 2024, and 47% in 2025.

4. OpenAI, xAI, and public software

OpenAI has been marked at a higher multiple on a slower, more consumer-heavy curve — and that gap has closed as its own run rate caught up.

• March 31, 2026: $122 billion raised at an $852 billion post-money valuation. The company said it was generating $2 billion a month (about $24 billion annualized); the New York Times later described a March implication of about $24 billion. That is roughly 35.5x. At the same moment Anthropic’s February mark was 27x on $14 billion, and six weeks later Anthropic’s Series H was 20.5x on $47 billion.
• August 2026: Bloomberg reported OpenAI’s run rate above $40 billion, while an August employee tender held the price at $852 billion. Implied multiple about 21x — in line with Anthropic’s May primary, and richer than Anthropic’s stale $965 billion mark on the July $65 billion update (~15x).
• September 29, 2026: Reuters, confirming Axios, said OpenAI’s annualized revenue was approaching $70 billion, up more than 70% since the start of the third quarter, with enterprise sales more than doubling since July and Q3 consumer revenue exceeding all of 2025 consumer revenue. Early-October reports had OpenAI in talks to raise at least $30 billion at about a $1.4 trillion pre-money valuation. If both figures held, that would be about 20x — the same neighborhood as a $2 trillion Anthropic IPO on a $100 billion pace. Neither round has closed.
• Mix contrast: Anthropic is ~80% enterprise/API. OpenAI’s historical mix was consumer-led (ChatGPT); by mid-2026 enterprise was described as more than half of revenue in some estimates, but the September Reuters note still shows consumer as a large incremental engine. Weekly users remain an order of magnitude apart. Revenue leadership and user leadership are different businesses.

xAI is not a useful run-rate comp unless you specify which revenue you mean.

• January 2026 Series E: $20 billion at a $230 billion valuation; the February SpaceX acquisition marked xAI at $250 billion.
• SpaceX’s IPO filing, reported by TechCrunch on May 20, 2026: the AI segment (Grok, X, and compute/infrastructure, not a pure model API) did $3.2 billion of revenue in 2025 and lost $6.4 billion from operations. On that segment number, $230 billion is about 72x trailing segment revenue. On narrower estimates of standalone Grok product revenue near $500 million, the multiple is several hundred times. Those are different businesses stuffed into one label. Q2 2026 AI-segment revenue of $2.56 billion (later SpaceX reporting) annualizes near $10 billion — still a different object from Anthropic’s Claude run rate.

Public software, as of late summer 2026, does not underwrite a 20x multiple on trailing sales. It can underwrite something in that range only for the fastest growers on forward revenue — and even they are not growing like this.

• Scalar’s September 2026 software index, EV / next-twelve-months revenue: Palantir 43.3x (64.7% NTM growth), CrowdStrike 35.0x (23%), Cloudflare 33.2x (29.7%), Snowflake 17.5x (28.3%), Datadog 16.5x (24.4%). The high-growth cohort (NTM growth above 20%) averaged 10.3x. In August the same high-growth cohort was 9.6x, with Palantir at 33.6x NTM.
• Windsor Drake’s August 21, 2026 public AI set: median 15.0x EV / trailing revenue (interquartile 9.3x–23.8x); AI application software 11.1x; Palantir 68.7x trailing on ~93% growth and ~85% gross margin. A broader SaaS set the same date had a median near 4.9x trailing.
• SaaSDB’s Q2 2026 public SaaS screen: median 4.8x EV / trailing revenue; Palantir 75.3x trailing; Snowflake 21.5x.

Palantir is the only large public software name whose forward multiple sits above Anthropic’s May run-rate multiple, and it earns that with software gross margins and positive operating margin. Anthropic’s Q2 2025 gross margin is not in the draft filing excerpts that have been reported; compute was $7.33 billion in 2025 against $4.6 billion of revenue, and operating loss was about $8 billion. A 20x run-rate mark is a bet that consumption growth continues and that gross margin inflects — PitchBook’s October 2, 2026 note puts a $2 trillion price at about 435x 2025 revenue, under 44x annualized Q2 2026 revenue, and just under 31x the July run rate, and says that price still needs margin improvement the filing has not established.

5. How to use this if you are pricing the IPO or competing

The stable object in this tape is not the multiple on trailing sales. It is a willingness to pay roughly 20–30x the latest monthly annualization, then let the next quarter re-rate the denominator. Series G (27x), the April offer chatter (~27x), Series H (20.5x), a $2 trillion talk price on the July print (~31x), and a $2 trillion talk price on a $100 billion pace (~20x) all sit in that band. OpenAI’s March 2026 mark was the outlier on the high side (35x) and has since compressed toward the same band as its run rate caught up.

Three adjustments matter more than another headline:

1. Haircut the run rate to a booked-revenue equivalent if growth slows. In 2025, exit run rate was ~2x full-year GAAP. If 2026 growth decelerates, a $65–100 billion July/September pace will not become $65–100 billion of recognized 2026 revenue. Q1 plus Q2 already imply a second-half hurdle: $4.73 billion + $11.5 billion = $16.2 billion through June, so a $65 billion year would require ~$49 billion in the second half.
2. Do not compare Anthropic gross marketplace revenue to OpenAI net Azure revenue. The 2025 fee drag was hundreds of millions, not the majority of sales, but at a $65 billion pace even an 8% marketplace haircut is about $5 billion. OpenAI’s $8 billion claim was larger than the 2025 fee ratio and should be treated as a negotiating position until the S-1 accounting policy is public.
3. Claude Code is the wedge, not the P&L. Company-confirmed share went from ~10% (September 2025) to ~18% (February 2026). The dollars that reprice the company are still API and enterprise consumption, concentrated in a small number of buyers and in two cloud partners who are also investors, suppliers, and competitors. That concentration is the risk a 20x run-rate multiple does not price, and it is why a coding-tool competitor can move Anthropic’s growth rate without matching its headline revenue.

Confidence summary: primary-round valuations and the company-stated run rates at Series F, G, the April compute posts, and Series H are high. The July $65 billion update is high for a private figure (three independent outlets, investor documents). The September $100 billion pace and any $2 trillion IPO price are medium. Claude Code above $2.5 billion is estimate-grade. Any multiple that uses 2025 GAAP revenue against a 2026 valuation, or that ranks Anthropic against OpenAI without a gross/net bridge, is not a like-for-like sales multiple.

---

Recent Findings Supplement (October 2026)

Anthropic’s revenue run-rate accelerated sharply after its May 2026 Series H round, reaching ~$65B annualized by late July 2026 (up from $47B in May), driven by enterprise/API usage and Claude Code, while its valuation climbed from $380B (Feb) to $965B (May) with IPO targets exceeding $2T.[1][2]

This growth outpaced OpenAI’s reported trajectory in the same period, though both firms’ run-rates rely on similar forward-looking annualization methods that differ from booked GAAP revenue.[3]

Funding/Valuation Timeline and Matched Run-Rates (Post-April 2026 Updates)

• February 2026 Series G: $30B raise at $380B post-money valuation. Run-rate reported around $14B (consistent with progression to later figures). Claude Code alone exceeded $2.5B run-rate.[4]
• May 28–29, 2026 Series H: $65B raise (including $15B from prior hyperscaler commitments like Amazon) at $965B post-money valuation. Company-announced run-rate crossed $47B earlier in May (one of the last self-reported figures). This more than doubled the prior valuation amid surging enterprise demand.[1][5]
• Subsequent developments (June–October 2026): Confidential S-1 filed in June 2026 targeting potential $2T+ IPO valuation (possibly November timing). Run-rate reached >$65B by end-July (Bloomberg sources, shared in investor updates). Projections cited: $100–120B+ by end-2026 and $190–200B in 2028. Later rumors of additional ~$30B+ raises at >$900B pre-money.[6][7]

Implication for competitors: The speed of these step-ups (roughly doubling valuation while run-rate grew ~3–4x from Feb to July) signals that private AI valuations are increasingly tied to monthly momentum rather than annual results, raising the bar for new entrants needing similar hyperscaler backing or product traction to match pace.

Revenue Run-Rate Progression (Detailed Monthly/Quarterly Snapshots)

Multiple independent trackers (Bloomberg, Sacra, First Page Sage models, Reuters) align on the trajectory, with run-rate defined as annualizing the most recent month’s performance:[8][9]

• End-2025: ~$9B.
• Feb 2026: ~$14–15.3B.
• March–April 2026: ~$19–30.8B (steepest early acceleration).
• May 2026: ~$47–49.3B (Series H timing).
• June–July 2026: ~$58.9–65B+ (late July confirmed >$65B).
• August 2026 models: Up to ~$69.7B.
• Q2 2026 actual (preliminary, not run-rate): $11.5–11.6B (14x YoY from $787M in Q2 2025; exceeded full 2025 revenue of ~$4.6B). Q1 2026 actual: ~$4.2–4.73B.[10]

2025 full-year actual revenue reached ~$4.6B (12x growth from ~$386M in 2024).[9]

Note on definitions: Run-rate (or ARR in some reports) extrapolates current pace and can exceed recognized revenue due to lumpy deals or usage spikes; it is the metric Anthropic and peers emphasize in updates. Some reports flag gross vs. net recognition differences (e.g., cloud partner revenue booked gross).[11]

Implication: Entrants must demonstrate verifiable monthly scaling (not just annual guidance) to attract comparable capital; the gap between run-rate and actuals widens scrutiny on sustainability.

Revenue Breakdowns and Claude Code Role

• Overall mix (recent estimates): ~75–85% from enterprise/API/business (pay-per-token dominant); consumer/Pro/Max/Team seats smaller (~15–25% combined). Cloud partners (AWS, Google, Microsoft) contribute significantly but at lower margins due to revenue share.[3][11]
• Claude Code contribution: 15–21% of total run-rate across 2025–2026 months (e.g., 18.7–20.9% in mid-2026). Q2 2026 revenue share: ~19.4% ($2.25B of $11.6B total) at 48% gross margin. It hit $1B ARR by Nov 2025, $2.5B+ by Feb 2026, and continued tracking overall growth (not outsized). Enterprise use >50% of Claude Code revenue.[8][8]
• Other lines (Q2 2026 example): Direct API ~34% (higher margins), cloud partner API ~23% (lower margins), seats/subscriptions balance the rest.[8]

Implication: Pure consumer plays or undifferentiated API offerings face headwinds; coding/agent tools with enterprise lock-in (like Claude Code) provide a repeatable growth lever, but concentration risks persist (top customers, including coding platforms, can drive outsized shares).

Valuation Multiples, Comparisons, and Context

• Anthropic implied multiples: At May 2026 $965B valuation on $47B run-rate ≈ 20.5x. At $2T IPO target on $65B July run-rate ≈ 31x (vs. 435x on 2025 actual revenue, compressing with growth). Vs. 2028 forecasts, far lower.[12][9]
• OpenAI comparables (recent): End-2025 >$20B run-rate; Feb/Mar 2026 ~$25B; July 2026 >$40B (with >20% MoM July growth); later reports of nearing $70B annualized recurring revenue. Valuation context ~$852B post-raise. Higher consumer/ChatGPT mix (~85% in some older snapshots) vs. Anthropic’s enterprise tilt. Multiples at similar stages often 25–40x on run-rate.[13][14]
• xAI: Much smaller product ARR (~$500M mid-2026 for Grok standalone); broader AI/compute segment higher (Q2 2026 revenue examples in billions including leasing). Valuation ~$230B pre-SpaceX integration (or part of larger entity). Multiples 400x+ on low base.[15]
• Public software/AI peers: High-growth names like Palantir ~53x forward revenue; SpaceX/Cloudflare examples in 40x+ range on expected sales. Traditional SaaS 6–15x (up to 20x+ for fastest growers). AI infra commitments (Anthropic’s ~$518B over decade) add risk not always reflected in pure software multiples.[16]

Definitional/confidence flags: Figures are primarily from Bloomberg/Reuters reporting on company updates or leaked filings (high confidence on headline run-rates at funding dates); monthly breakdowns often from analyst models (medium confidence, directionally consistent across sources); breakdowns are estimates (Sacra, internal models) with no official granular public splits. Run-rate methodologies may differ slightly between firms (e.g., inclusion of certain contracts or gross/net). No single source provides exhaustive audited breakdowns.[11]

Implication for market entrants: Multiples remain elevated vs. public software but compress rapidly with scale; success hinges on matching Anthropic/OpenAI’s enterprise/API + vertical tool (coding/agent) flywheel while managing compute commitments and customer concentration. New data continues to emerge rapidly ahead of potential IPOs.

The process is well documented; next I’ll pin down counsel, the exchange, analyst multiples, and whether a public S-1 is out as of today.
Anthropic is in confidential registration, not a priced IPO. As of October 3, 2026, the live plan is a Nasdaq debut with marketing as soon as the week of November 9 and trading before Thanksgiving, aimed at roughly $2 trillion and as much as $100 billion of proceeds. That price is a banker and investor target, not a company-set offer price, and independent research does not treat $2 trillion as earned until margins and the $518 billion compute stack are clearer.

The process is real, late, and still not public

Anthropic, PBC, confidentially submitted a draft Form S-1 on June 1, 2026, four days after closing a $65 billion Series H at a $965 billion post-money valuation. The company said the filing only preserves the option to list after SEC review, and that share count and price were not set. [1] [2] Reuters reviewed that confidential prospectus in late September. It was still not on EDGAR as of September 26, and as of October 3 there is no company announcement of a public S-1. [3] [4] [5]

The calendar has slipped twice. Early September reporting had a public prospectus in late September and marketing from mid-October, with a listing days before the November 3 midterms. By October 1, Bloomberg’s sources had formal marketing as soon as the week of November 9, trading before Thanksgiving on November 26, and a year-end deadline if that window moves again. IFR’s bankers, the same day, said November after the midterms is the working plan. Renaissance Capital’s October 2 week-ahead note uses the same November 9 marketing date and tracks the company as pre-IPO under an internal ticker ANTH.RC, with the public ticker still TBA. [6] [7] [8] [9]

Venue is the firmer fact. Business Insider reported on September 13 that Anthropic had selected Nasdaq; Bloomberg and Reuters carried the same account. A Nasdaq listing would also make the stock eligible for the Nasdaq-100, which is why the venue fight mattered after Nasdaq won SpaceX. [10] [11] [12]

Banks are inside the deal but the tombstone is not final. The Financial Times reported September 4 that Morgan Stanley was in pole position for “lead left” and Goldman Sachs was expected to be stabilization agent, with the lead-left choice not yet locked. JPMorgan, Citigroup, and Barclays were expected in leading roles after providing debt. Reuters named the same four—Morgan Stanley, Goldman, JPMorgan, and Citi—as working on the IPO. In parallel, Anthropic has been finalizing a $15 billion revolving credit facility, up from a prior aim around $10 billion, with Morgan Stanley leading the syndicate and the IPO banks committing at the top tier so credit roles convert into underwriting roles. [13] [6] [14]

Counsel is thinner in the public record. Anthropic retained Wilson Sonsini for IPO preparation, first reported by the Financial Times in December 2025; the firm has advised the company since 2022. No 2026 report retrieved here names underwriters’ counsel or confirms a second issuer firm. [15] [16]

The raise is a banker number, not a filed amount. The New York Times reported August 21 that bankers told potential investors Anthropic could seek more than $100 billion at a $2 trillion valuation. IFR, citing a leaked draft prospectus on October 1, said bankers leading the deal were targeting $100 billion of proceeds at $2 trillion. Anthropic has not set either figure. For scale, SpaceX listed in June at $1.77 trillion and raised $85.7 billion, per the Times; the FT put that deal at $1.78 trillion and $86 billion. [17] [8] [13]

What that means for anyone trying to get allocated: the book is being built around a credit syndicate and a still-unfinalized lead-left bank, and the public document that would let buyside set its own price is not out. Missing the pre-midterm window did not kill the 2026 listing; it moved the marketing into a compressed post-election, pre-Thanksgiving slot.

The leaked prospectus is a growth story strapped to a utility bill

The numbers Reuters extracted from the confidential filing are why the multiple debate is not abstract. Revenue rose about twelvefold in 2025 to nearly $4.6 billion, from roughly $386 million. Operating loss widened to $8.06 billion from $2.98 billion. Compute and infrastructure spending was $7.33 billion. Net loss was about $42 billion, of which roughly $34 billion was an accounting charge on financing that could convert into shares, not cash spent running the business. Cash, equivalents, and short-term investments were $20.28 billion at December 31, 2025. [3] [18]

2026 is a different company than 2025. First-quarter revenue of $4.73 billion already exceeded all of 2025. Preliminary second-quarter revenue topped $11.5 billion; the Times separately cited $11.6 billion for the quarter. By late July the annualized run rate was more than $65 billion, versus about $9 billion at the end of 2025 and the $47 billion run rate Anthropic disclosed with the May round. The Financial Times has also reported that the company told a small group of investors it expects a second consecutive quarter of positive adjusted operating income, with gross margins above 80% before partner revenue-share and training costs. [18] [19] [17] [20]

The binding number is $518 billion of cloud, compute, and infrastructure obligations over roughly the next decade, about 80% non-cancelable or payable even if usage falls short. Reported slices include at least $111.1 billion to Google, $110 billion to Amazon, $31.4 billion to Microsoft, and about $161.2 billion of Broadcom-related equipment leases. PitchBook’s reading of the same leak: many large customers can cut spend at any time, so revenue is flexible and a large share of cost is not. Two customers supplied nearly a quarter of 2025 revenue. [21] [22] [23] [18]

Governance will also be in the public filing. Reuters’ review describes a Founder LLC through which Anthropic’s seven co-founders would hold 50.1% of voting power via Class F shares, including on board elections. [5]

The mechanism that makes $2 trillion arguable is the run-rate jump, not the audited year. A buyer paying 435 times 2025 sales is really underwriting that July’s pace holds, that 2028 revenue of roughly $190–200 billion—which Reuters reported in August from people familiar with the company’s financials—shows up, and that the non-cancelable compute bill is filled by customers who are not locked in. [24]

Each valuation camp is using a different clock

There is no single sell-side price target, because the banks inside the deal are restricted and the public S-1 is not out. What exists is a stack of reported targets, each tied to a different revenue year.

Banker and venture target, about $2 trillion, sometimes above. The Times and IFR both describe a $2 trillion valuation with up to or about $100 billion raised. Half a dozen Anthropic investors told the FT they expected $2 trillion or more, largely off the May $47 billion run rate and a year-end 2026 pace of $100–120 billion. One of those investors said that if growth stayed near 800%, the company could command at least 30 times revenue and be worth about $3 trillion. Bloomberg’s October 1 sources put the “fair” ballpark that prospective investors are using at $1.8–2.0 trillion, and said the company expects to match or beat SpaceX’s deal size. [17] [8] [25] [26]

The multiple only looks ordinary if the denominator is forward revenue. On the July run rate above $65 billion, $2 trillion is just under 31 times sales (PitchBook’s calculation). On a year-end pace above $110 billion, which the Motley Fool attributes to updated company expectations, it is about 18 times. Morningstar’s Michael Field, speaking to Reuters, said a $2 trillion price “sounds outlandish” until it is restated as roughly 18–20 times sales, which he called more reasonable than SpaceX. That framing implies a sales base around $100–111 billion, not the $4.6 billion audited year. On the Reuters 2028 forecast of $190–200 billion, $2 trillion is about 10–10.5 times sales—below where SpaceX and Cloudflare were trading at 41.6 times expected 2026 revenue, and well below Palantir at 53 times, on LSEG figures Reuters cited in August. [23] [27] [28] [24]

Public-market institutions, closer to $1.5 trillion. The Information, as summarized on October 1, reported that some banks had discussed about $2 trillion in early talks, while two major institutional investors thought closer to $1.5 trillion was appropriate, citing rates, compute capex, and funding pressure. An earlier August account of The Information had bankers discussing as much as $1.5 trillion, so that figure has functioned as both a banker ceiling and a buyside floor depending on the week. At $1.5 trillion the July run rate is about 23 times sales, and the 2028 forecast is roughly 7.5–8 times. [29] [30]

PitchBook will not underwrite $2 trillion on the leak alone. Harrison Rolfes’s September 30 note, republished by Morningstar, says the filing supports a value well above $1 trillion and does not justify $2 trillion until the public document shows gross margin, the payment schedule on the $518 billion, and a path to free cash flow. The test he sets: $150 billion of annual revenue at a 30% free-cash-flow margin would produce $45 billion of cash, and $2 trillion would be about 44 times that cash flow. That is a mechanism argument, not a point target. [23] [18]

Renaissance Capital has not published an IPO valuation range in the materials retrieved. Its public AI tracker, updated around October 3, lists Anthropic as pre-IPO, AI-centric, ticker TBA, at $965 billion—the last private round, not an estimated offer price. Its October 2 note confirms the November 9 marketing rumor and does not add a multiple. [31] [9]

Secondary markets are a scarcity premium, not a clearing price. Prints disagree by platform and date, which is the point: almost no one wants to sell into a rumored $2 trillion IPO.

• July: Caplight and Rainmaker described trades around $1.2 trillion, above OpenAI on the same screens. [32] [33]
• August 14: Business Insider, citing three secondary traders including Rainmaker’s Glen Anderson, said the few sellers were around $1.5 trillion. [34]
• September 2: the NYSE OpenVC Unicorn Index implied $1.17 trillion, about 21% above the May round. [35]
• October 1: Moonberg attributes a Nasdaq Private Market mark of $1.36 trillion, about $829 a share, 41% above Series H. [36]
• October 3: StockAnalysis, sourcing Clarity, shows a last trade of $1,075.57 and an “implied valuation” of $910 billion against the $965 billion confirmed round. Those two Clarity figures do not reconcile with each other on the Series H share price of $589.01, so the page should be read as a quote screen, not a fundamental mark. [37] [38]

FinanceFeeds’ arithmetic, using Forge’s Series H price of $589.01 and the $965 billion round, implies about 1.64 billion shares and about $1,221 a share at $2 trillion. Anthropic has not published a share count, so that is a reconstruction, not an offer price. [39]

Forecast and prediction markets sit on top of the banker number. A FutureSearch model re-run August 29, cited by Markets.xyz, put the median first-day market cap at $1.82 trillion, with a 10th percentile of $1.06 trillion and a 90th of $2.77 trillion. Moonberg’s October 3 aggregation of Polymarket and Kalshi shows an 83% implied probability of a listing by year-end and, conditional on an IPO, a median debut cap of about $2.1 trillion, with 66% odds of at least $2.0 trillion. Those are betting markets, not research. [40] [36]

FT Lex on September 21 pushed the bull case further as illustration, not a target: if 2028 sales were valued like SpaceX at 16 times that year’s revenue, the future value could be $5 trillion; discounting a 10-times multiple on a much larger out-year revenue base produced $4.5 trillion or, in an extreme share-of-knowledge-work scenario, $10 trillion today. Lex was showing how total-addressable-market inflation works, including a Journal report that Anthropic may tell investors its opportunity exceeds $30 trillion. [41] [42]

Low, base, and high — only ranges that someone has actually published

These are reported cases, not a model built for this note. “Low” is the cautious public-market and downside-forecast cluster. “Base” is the number bankers and the company are being associated with. “High” is what growth investors and tail forecasts require. Multiples in the last column are arithmetic on published revenue figures, marked as calculated where a source did not print the multiple itself.

Case	Valuation	Who, and when	Methodology	Implied multiple
Last private round (anchor, not an IPO case)	$965 billion post-money	Anthropic Series H, May 28, 2026	Priced primary round; $65 billion raised. Run rate had just crossed $47 billion	~20.5× May run rate (SmartAsset calculation) [1] [43]
Low	About $1.5 trillion; forecast downside $1.06 trillion	Two large institutions via The Information, Sept. 30; FutureSearch p10, Aug. 29	Institutions weighting rates, capex, and funding need over growth. FutureSearch is a forecast distribution, not a DCF	~23× July run rate above $65 billion; ~7.5–8× the $190–200 billion 2028 forecast (calculated) [29] [40]
Research “not yet” band	Well above $1 trillion, short of $2 trillion	PitchBook / Morningstar, Sept. 30–Oct. 2	Will not pay 31× July sales until gross margin, commitment schedule, and free cash flow are in the public filing. Illustrative hurdle: $150 billion revenue at 30% FCF margin = $45 billion cash, and $2 trillion is ~44× that cash	$2 trillion = ~435× 2025 sales, ~31× July run rate [23] [18]
Base	$1.8–2.0 trillion, with “more than $2 trillion” still in circulation	Bloomberg investors, Oct. 1; NYT bankers, Aug. 21; Reuters on the prospectus, Sept. 28; IFR, Oct. 1	Forward sales, not trailing. Morningstar’s Field: 18–20× sales looks less extreme than SpaceX. Motley Fool: ~18× if year-end run rate exceeds $110 billion. On 2028 company forecast, ~10×	~31× July run rate; ~18–20× a ~$100–111 billion sales base; ~10–10.5× 2028 revenue [26] [28] [27] [24]
High	About $2.8–3 trillion, with illustration far above that	FutureSearch p90 $2.77 trillion; one FT-quoted investor ~$3 trillion; FT Lex illustrations of $4.5–10 trillion	30× revenue if ~800% growth holds and year-end run rate is $100 billion-plus. Lex applies 10–16× out-year sales, or a share of a $30–60 trillion knowledge-work TAM, then discounts	30× on $100 billion = $3 trillion (investor’s own math). Lex is scenario math, not a price target [25] [40] [41]
Secondary, not a case	$1.17–1.5 trillion recent prints; NPM-attributed $1.36 trillion on Oct. 1	Caplight, Rainmaker, OpenVC, Moonberg/NPM	Scarcity: holders will not sell into a $2 trillion rumor, so the print is an ask, not a fundamental clearing price	Premium of roughly 20–55% to the $965 billion round, depending on the screen [34] [35] [36]

Renaissance Capital does not appear in this table because its public tracker still carries the $965 billion private mark and does not publish low/base/high IPO values. [31]

A $2 trillion deal would be about 2.1 times the May round. On the July run rate it is a lower multiple than SpaceX’s roughly 94 times sales at its debut, which is the comparison bulls keep making. It is still a bet that revenue compounds into the commitments, not a bet that 2025’s $4.6 billion justifies the price. [27]

OpenAI is choosing a private mark instead of this window

The competitive fact is the sequencing, not the logo. Both companies filed confidentially in June. Anthropic is trying to list in 2026. OpenAI has pushed a debut to 2027. [8] [20]

OpenAI’s last priced round was $122 billion committed in March 2026 at an $852 billion post-money valuation. It is now in early talks to raise at least $30 billion at about $1.4 trillion before the new money, as a bridge in place of an IPO. Bloomberg and The Information both described the talks as early; The Information said no term sheet had been signed. A September 15 Bloomberg report had the company considering about $1.2 trillion. Sam Altman told Bloomberg TV on September 29 he does not want the pressure of being newly public while safety decisions are still in front of the company; CNBC reported OpenAI has ruled out a 2026 IPO. [44] [45] [46]

Revenue is the cleaner comparison. Axios, with the figure separately tied to Reuters in follow-on coverage, put OpenAI’s annualized revenue near $70 billion by DevDay, up more than 70% since the start of the third quarter, with enterprise sales more than doubled since July. Anthropic’s last widely reported run rate is the July figure above $65 billion, with investor and company expectations of $100–120 billion by year-end still unconfirmed in a public filing. [45] [44] [25]

On current run rate, the headline gap is mostly timing of the denominator. OpenAI at $1.4 trillion on about $70 billion is roughly 20 times sales (calculated). Anthropic at $2 trillion on a July pace above $65 billion is about 31 times, and on a $110 billion year-end pace is about 18 times. The companies are asking public or private investors for similar multiples on this year’s pace. Anthropic is asking them to capitalize a higher forward year, and to do it in the public market first, with $518 billion of mostly fixed compute against OpenAI’s choice to raise another private round and wait. [23] [27]

Secondary markets had already flipped the old hierarchy by July, with Anthropic near $1.2 trillion and OpenAI near $908 billion on Caplight. StockAnalysis’s October 3 Clarity screen still shows OpenAI’s last confirmed round at $852 billion and an implied mark around $906 billion. If the $1.4 trillion talks price, OpenAI would leapfrog Anthropic’s last primary round and land near the institutional “low” case for Anthropic’s IPO, while Anthropic’s bankers are still socializing twice OpenAI’s March mark. [33] [47] [44]

What has to be true for the base case to clear

The IPO is a pricing event for frontier AI, but the mechanism is narrower than the $2 trillion headline. Public investors are being asked to pay a high-teens to low-30s multiple of a run rate that did not exist a year ago, while underwriting contracts that force Anthropic to pay cloud and chip landlords whether customers stay. PitchBook’s condition is the right one: the public S-1 has to show that gross margin after partner share survives, that the $518 billion has a payment schedule the $15 billion revolver and the IPO proceeds can bridge, and that the largest customers cannot walk without the cost walking with them. [18] [6]

Three disagreements are still unresolved, and they move the table more than another week of timing rumors.

• Which revenue year is the multiple on. Trailing 2025 makes $2 trillion absurd at ~435 times. July’s run rate makes it a growth-stock price at ~31 times. The 2028 forecast makes it look cheaper than software comps at ~10 times. Bankers are selling the third. PitchBook is refusing to underwrite it without the cash conversion of the second.
• Primary versus secondary versus IPO. The only fully priced mark is $965 billion in May. Secondary asks of $1.2–1.5 trillion, and a reported NPM mark of $1.36 trillion, are what scarce stock costs ahead of a rumored pop. They are not evidence that a $100 billion book can be filled at $2 trillion.
• Anthropic versus OpenAI. OpenAI’s delay removes a same-quarter competing mega-deal, which helps Anthropic’s window. It also means the first public print will set the comp. If Anthropic lists at $1.5 trillion, OpenAI’s $1.4 trillion private talks look full. If it lists at $2 trillion on a similar current run rate, OpenAI’s 2027 process inherits a higher comp—and a public template for how investors haircut non-cancelable compute and dual-class founder control.

Nothing in the reporting says the deal is pulled. Renaissance Capital, Bloomberg, and IFR, all in the first days of October, still have a 2026 listing, with the next hard date the week of November 9. The number that will decide whether that listing prices near $1.5 trillion or $2 trillion is not in a prediction market. It is the gross-margin and commitment schedule that the confidential filing has not yet put on EDGAR.

---

Recent Findings Supplement (October 2026)

Anthropic confidentially filed its draft S-1 with the SEC on June 1, 2026, and has since advanced preparations for a Nasdaq IPO targeting a valuation exceeding $2 trillion—more than double its May 2026 private mark—with the latest reports pointing to a mid-November listing window after repeated delays.[1][2]

This positions Anthropic ahead of OpenAI (which has shifted to a potential 2027 timeline) in the race to become one of the first major AI labs publicly traded, testing whether public markets will pay frontier-lab multiples amid massive compute costs and governance risks highlighted in the filing.[3][4]

Recent Timeline Shifts and Marketing Plans

Anthropic has adjusted its IPO schedule multiple times in recent months due to regulatory reviews, market conditions, and a desire to include fuller Q3 financials.

• Confidential S-1 filed June 1, 2026; public prospectus reporting began late September 2026 (Reuters/FT review of the draft).[5]
• Earlier targets (mid-September prospectus, mid-October marketing) slipped; current plans call for investor meetings (e.g., Oct. 14 in San Francisco), formal marketing as early as the week of Nov. 9, and trading potentially before Thanksgiving (Nov. 26), with a year-end 2026 close at latest.[6][7]
• A $15 billion revolving credit facility is being finalized (with Nvidia reportedly in talks for up to $10 billion as an anchor).[8]

Implication for competitors/entrants: The compressed window and pre-IPO credit line underscore the need for strong balance-sheet optics before roadshow; delays signal sensitivity to quarterly momentum and external factors like midterms.

Banks, Exchange, and Raise Details

Lead underwriters are Morgan Stanley (lead-left) and Goldman Sachs (stabilizing agent), supported by JPMorgan and Citigroup (with possible Barclays or smaller firms in the syndicate).[8][9]

• Exchange: Nasdaq (reported).[8]
• Target raise: Up to ~$100 billion in a deal sized for a >$2 trillion valuation.[8]
• No public details on specific law firms in recent reporting.

Implication: Top-tier bulge-bracket involvement signals institutional confidence but also highlights capacity constraints across mega-deals; anchor interest from Nvidia could provide a valuation floor and signaling effect.

Key New Disclosures from the September 2026 Prospectus Reporting

The leaked/reported draft S-1 provides the first detailed public view of trailing and forward metrics (post-April 2026 developments focus here):

• 2025 revenue: ~$4.6 billion (12x or 1,088% growth from $386 million in 2024); ~25% from two customers.[1][10]
• Operating loss: Widened to $8.06 billion (from $2.98 billion); net loss ~$42 billion (mostly ~$34 billion non-cash accounting charge on financing instruments).[1]
• Compute/infrastructure: $7.33 billion spent in 2025; ~$518 billion in future contractual obligations (~80% non-cancelable).[5]
• Forward guidance (per prospectus/investor updates): 2028 revenue $190–200 billion; run-rate reached ~$47 billion by May, ~$65 billion by July, with end-2026 targets of $100–120 billion or >$110 billion.[8][11]
• Risks emphasized: AI models potentially exhibiting blackmail or shutdown resistance; government actions (e.g., export controls, model bans) causing revenue disruption; heavy customer concentration and flexible spending.[5]

Cash position: $20.28 billion as of Dec. 31 (prior year).[1]

Implication: The filing reframes the narrative around explosive growth offsetting losses and capex commitments, but highlights execution risks (compute availability, customer stickiness) that could pressure multiples if margins or retention disappoint.

Valuation Ranges, Methodologies, and Table

Investor and analyst discussions center on revenue multiples applied to run-rate or forward revenue, benchmarked against SpaceX’s June 2026 IPO (~$1.77 trillion valuation) and growth trajectories. No detailed Renaissance Capital or named sell-side reports appeared in recent coverage; perspectives come from PitchBook/Morningstar analysis, investor models (FT), and commentators.[12][13]

Methodology common themes:
- Trailing 2025 revenue yields extreme multiples (~435x at $2T), so focus shifts to annualized run-rate (better captures momentum) or 2028 projections.
- Assumes continued hyper-growth (e.g., from $65B July run-rate toward $100B+ by year-end and $190–200B in 2028), margin expansion via scale/compute efficiencies, and stable/high-value customers.
- Comparables: SpaceX (AI-adjacent growth story); implied 18–31x on near-term run-rate at upper end (seen as aggressive but plausible by bulls given 12x+ YoY history; bears note need for FCF proof and ~$518B commitments).[12][11]

Valuation Table (Low/Base/High cases, synthesized from recent reports):

• Low case (~$1T or below current private): ~15x on $65B+ run-rate or conservative growth assumptions; PitchBook views $2T as not yet justified without stronger gross margins/FCF visibility, supporting well above $1T but short of peak target.[12]
• Base case ($1.5–1.8T): 20–25x on July run-rate or adjusted for risks; aligns with some investor models bridging private $965B and ambitious targets, or 18x on end-2026 projections.[14]
• High case (>$2T or $1.8–2T+): 30x+ on current run-rate or lower on 2028 revenue (~18x on $110B+); driven by revenue acceleration bets and SpaceX outperformance; most investor commentary clusters here.[1][11]

Sources for ranges: Primarily Reuters (target >$2T), Bloomberg/FT investor interviews ($1.8–2T+), PitchBook/Morningstar analysis (run-rate multiples and skepticism on full $2T), and secondary commentary.[12][6]

Latest private valuation: $965 billion post-money (May 2026 Series H, $65B raise).[3]

OpenAI comparison: OpenAI’s March 2026 round at $852 billion (lower revenue run-rate, e.g., ~$24–70B range reported variably); targeting ~$1–1.4T (bridge or IPO expectations); IPO delayed to 2027 amid higher reported losses and safety/governance focus.[15][16] Anthropic’s higher private valuation and faster-reported growth give it an edge in timing and perceived momentum.

Implication: A successful $2T+ print would reset AI valuation benchmarks upward (potentially pressuring or validating OpenAI’s path) but expose the company to quarterly scrutiny on compute ROI and customer concentration; entrants must demonstrate similar run-rate visibility or risk discounting.

Overall, post-June 2026 developments show accelerating preparations and ambitious targets backed by revenue momentum, tempered by explicit risk disclosures and timing flexibility. Public filing and roadshow demand will be the next key tests.

A $2 trillion IPO ask prices a future software franchise. The strongest bear case is that Anthropic is still a compute reseller with take-or-pay bills, flexible customers, and a price war—and that the numbers used to paper over that gap are not yet public, audited, or even consistent across the press.

The valuation rests on a run rate the prospectus has not yet proven can become cash

The reported target—about $2 trillion, more than double the $965 billion May 2026 Series H mark—is a bet on annualized revenue, not on 2025 results. Reuters’ review of the confidential draft prospectus shows 2025 revenue of nearly $4.6 billion, an operating loss of $8.06 billion (up from $2.98 billion in 2024), and a GAAP net loss of about $42 billion. Roughly $34 billion of that net loss was a non-cash remeasurement of financing that can convert into shares, not cash spent running the business. Compute and infrastructure alone was $7.33 billion—more than revenue, and about 58% of $12.65 billion in operating expenses. Cash and short-term investments at year-end 2025 were $20.28 billion. [1] [2] [3]

PitchBook’s reading of the same leak is the cleanest valuation critique: a $2 trillion price is about 435 times 2025 revenue and just under 31 times a late-July 2026 run rate above $65 billion. That multiple only works if gross margins rise, customers stay, and contracted capacity is used productively. PitchBook’s conclusion is explicit: the leak supports a valuation well above $1 trillion and does not justify $2 trillion until the public filing shows gross margin, the payment schedule on commitments, and a path to free cash flow. [4] [5] Private-market marks have not fully bought the ask either. Nasdaq Private Market put the company near $1.36 trillion as of October 1, 2026—about 41% above Series H, still roughly a third below $2 trillion. [6]

What this means: public investors are being asked to pay a multiple that assumes the July run rate is durable cash revenue. The last full year in the draft does not show that conversion.

Gross margin is the missing number, and the cost structure is the opposite of software

Anthropic does not have software economics yet. In 2025 it spent about $1.60 on compute and infrastructure for every dollar of revenue. [7] Epoch AI’s reconstruction from The Information reporting put 2025 inference compute near $2.7 billion on roughly $4.5 billion of revenue (implying about a 40% gross margin) and training compute near $4.1 billion—so serving customers and training models together dwarf everything else. [8] The Information reported that Anthropic cut its own 2025 gross-margin projection to 40% from an earlier 50% goal because inference costs ran about 23% above plan, even as cloud rental prices fell. OpenAI missed its margin forecast the same way. [9] [10]

The forward bill makes the margin question existential. The draft describes at least $518 billion of cloud, compute, and infrastructure obligations, about 80% non-cancelable or payable regardless of usage. Named pieces include at least $111.1 billion to Google (April 2026–July 2033), $110 billion to Amazon (May 2026–April 2036), $31.4 billion to Microsoft (non-cancelable except for uncured material breach), and about $161.2 billion of Broadcom-related equipment leases that neither side can cancel except on default. The filing language is blunt: if actual spend falls short, Anthropic must pay Google the difference, with similar terms on Amazon. [11] [12] PitchBook notes the draft still does not disclose gross margin—the figure that decides whether those contracts are a moat or a trap. [5]

Modeled tables circulating online (including quarterly gross margins climbing into the mid-50s and a Q2 2026 adjusted operating profit) should be treated as estimates, not prospectus facts. PitchBook’s only hard operating signal is a second-quarter 2026 adjusted operating profit excluding stock-based compensation, which it says does not establish cash generation or gross-margin improvement. [5] [13]

What this means: a competitor or a short does not need Anthropic to “fail.” It needs token prices or utilization to fall while minimum payments stay fixed. That mismatch is already written into the contracts.

Circular financing turns suppliers into underwriters of their own demand

Amazon, Google, and Nvidia are not arm’s-length vendors. They are investors, landlords, distributors, and—in Google and Amazon’s case—model competitors.

• Amazon has put about $18 billion in (convertible notes plus preferred), with more capacity available, and carries the stake near $190 billion. Anthropic owes AWS about $110 billion through 2036 on take-or-pay terms. An April 2026 expansion took Amazon’s potential commitment toward $33 billion in exchange for more than $100 billion of AWS spend over a decade, including Trainium. [14] [15] [16]
• Google is owed at least $111.1 billion on similar shortfall terms. Separate reporting describes Google at the center of a large TPU financing network (private-credit SPVs that buy hardware and lease it to Anthropic), so Google is chip supplier, investor, and credit backstop at once. [11] [17]
• Nvidia committed up to $10 billion of equity in a late-2025 arrangement with Microsoft, and later structures go further: a reported $35 billion Lambda cloud deal in which an Nvidia-backed neocloud serves Anthropic inside a data center whose lease Nvidia holds. Trade press put Nvidia’s contracted value with Anthropic above $180 billion. Nvidia’s CFO has rejected the “circular financing” label; the Financial Times’s cleaner description is old vendor financing—writing checks so customers can buy more product than they could otherwise afford. [18] [19] [20]

The accounting loop feeds the valuation loop. Amazon marked Anthropic notes from $42.2 billion on March 31 to $97.9 billion on June 30—a $55.7 billion one-quarter increase driven by private-round prices, not by cash distributions. [5] Marketplace distribution closes the circle in the income statement: 47% of 2025 sales, about $2.16 billion, ran through Amazon and Google, up from 11% in 2023 and 32% in 2024, with roughly $351 million paid back as distribution fees booked in operating expense rather than as a contra-revenue. [21]

What this means: demand, capex, and private marks are partly the same capital recycled. If public investors haircut the equity value, supplier marks, credit appetite, and “committed” capacity can reprice together.

Customer concentration and run-rate durability are the revenue-side mirror of those fixed bills

Two unnamed customers each produced 12% of 2025 revenue. The company warns that many of its largest customers are not on long-term contracts and can cut or stop spending. Consumption-based usage was about $3.8 billion of 2025 revenue versus $789 million of subscriptions, and management expects consumption to remain the substantial majority. Cloud partners collected 60% of the $909 million in customer bills outstanding at year-end 2025. [21] [21]

That is the bear mechanism in one sentence: revenue can fall inside a quarter; minimum infrastructure payments run into the 2030s. [5] Earlier reporting tied a large slice of the coding boom to Cursor and GitHub Copilot. Those are sophisticated buyers who can multi-home the day a cheaper model is “good enough.” Even the company’s own growth narrative concedes the base is lumpy: high-value API accounts, not a broad installed base of sticky seats, move the top line. [22] [23]

Run-rate headlines have also outrun recognized revenue before. Company and press figures put year-end 2025 annualized revenue near $9 billion and, in at least one CNBC-cited account, “actual” 2025 revenue near $10 billion. The draft prospectus puts recognized 2025 revenue near $4.6 billion. Q1 2026 revenue of $4.73 billion and preliminary Q2 revenue above $11.5 billion show the business did accelerate—but they also show why annualizing a peak month is not the same as a year of GAAP sales. [24] [25] [5]

What this means: the IPO case treats $65 billion, or $100–120 billion by December, as the valuation base. The filing’s own risk factors say the customers behind that curve are not locked in.

Price competition is already attacking the margin expansion the multiple requires

September 2026 made commoditization operational, not theoretical. OpenAI priced GPT-6.1 Sol at one-fifth the token cost of its own most capable model. Anthropic launched Claude Opus 5.5 at token prices 20% below Opus 5 and claimed about 40% lower effective cost than Opus 5 because the model uses fewer tokens. Gartner’s Anushree Verma described general-purpose models as increasingly interchangeable, with vendors grabbing share on price. [26] [27]

Chinese open-weight labs set the floor. Juniper Research, citing OpenRouter, said leading closed providers’ share of work on that platform fell from about 70% to 30% in a year, and that Chinese models typically run 60–90% cheaper. DeepSeek’s V4.1 Flash launch in September hit listed Chinese AI stocks and was framed as fresh pressure on labs including Anthropic. [28] [29]

The non-obvious implication: Anthropic’s scarcity thesis—“limited principally by the availability of compute”—justifies $518 billion of take-or-pay. A price war says the scarce input is being turned into a cheaper output faster than those contracts depreciate. If intelligence per dollar keeps falling, utilization can stay high and revenue per committed megawatt can still disappoint. [11] [26]

What this means: competing labs do not need to beat Claude on benchmarks. They need to be close enough that procurement switches on price, which is exactly when fixed compute bills hurt most.

Legal and government exposure is no longer a footnote, and some of it is already adjudicated

Copyright is a settled cash cost plus an open tail. In Bartz v. Anthropic, the company agreed to a $1.5 billion class settlement—described by plaintiffs’ counsel and the court record as the largest known U.S. copyright recovery—covering roughly 482,000 to 500,000 works at about $3,000 per work. Judge William Alsup had held that training on lawfully acquired books can be fair use, but that downloading and storing pirated copies was not. Final approval came on July 20, 2026. The release covers past acquisition and copying through August 25, 2025. It does not release output claims or future conduct. Funding is in installments through September 2027. [30] [31] [32] $1.5 billion is small next to a $2 trillion ask, but the open output docket is the live risk, and authors and publishers are still fighting over who gets paid. [33]

Government risk is fresher and closer to the IPO. After Anthropic refused to allow military use for autonomous weapons and mass domestic surveillance, the Pentagon designated it a supply-chain risk—the first time that label was applied to a U.S. company. On September 25, 2026, the D.C. Circuit upheld the designation 2-1. Anthropic has said the fight cost billions in lost business and damaged its reputation ahead of the IPO. A separate California ruling limited a broader ban; the Pentagon bar itself stands. [34] [35] [36]

In June 2026 the Commerce Department, using export-control authority, ordered Anthropic to cut foreign-national access to Claude Fable 5 and Mythos 5. The company disabled the models broadly to comply. Restrictions were lifted on June 30 after Anthropic coordinated mitigations with the government. The prospectus warns that government attitudes can hurt commercial customers and partners, not just public-sector revenue, which the company says is under 1% of sales. It also discloses that advanced AI could pose “catastrophic or existential risks to humanity,” and Reuters reported an FTC industry probe that includes Anthropic. [37] [38] [39]

What this means: the copyright settlement is largely a known liability. The regulatory pattern—export “is informed” letters, a supply-chain designation upheld on appeal, and a prospectus that flags contagion into commercial accounts—is a recurring tax on the enterprise growth story the IPO is selling.

Bubble talk, delayed listings, and precedents that rhyme—even without an Anthropic down round

Anthropic itself has not taken a private down round. The mark went from the May $965 billion round to about $1.36 trillion on Nasdaq Private Market. The stress shows up around it. The IPO slipped from an October target to after the November midterms, with marketing discussed for as early as the week of November 9. OpenAI postponed its own listing, with Sam Altman arguing it would be ill-advised to go public now. [40] [6]

Skeptics are arguing from capital structure, not vibes. Ed Zitron has called private-credit financing of the AI buildout a “brewing crisis,” pointing to CoreWeave’s $35.6 billion of debt as of June 30 and Goldman’s count of lower-rated AI borrowing. His June claim that Anthropic could not sustain a $47 billion run rate was later contradicted by the company’s own higher figures—which is a useful reminder that bears have been early, and wrong, on the top line. The $518 billion commitment stack is what remains after that miss. [41] Wealth managers quoted by Reuters have said they cannot underwrite multi-trillion equity values on companies that lose billions and require massive capex. [42] Commentary around the filing has revived Scott McNealy’s post-dot-com warning on paying rich sales multiples for businesses that never earned their cost of capital, and telecom-style vendor financing as the historical rhyme for Nvidia’s balance-sheet support. [43] [20]

There is no clean public precedent of an Anthropic-scale lab repricing in a down round yet. The nearer precedents are adjacent: neocloud leverage, Oracle credit spreads cited in bust scenarios, and the gap between derivative pre-IPO prints above $2 trillion and the last institutional round at $965 billion. Decentralized perpetual futures have implied valuations above $2 trillion; those markets are not price discovery for an S-1. [44] [45]

Leaked and press-reported figures have already been wrong, incomplete, or non-comparable

Treat every number that is not in a public EDGAR filing as provisional. As of late September, searches of SEC filings had not produced a public Anthropic S-1; Reuters reviewed a confidential draft, and Anthropic declined to comment. [1] [46]

Documented reliability problems:

• Gross versus net revenue. In an April 2026 internal memo reported by CNBC and The Verge, OpenAI’s chief revenue officer Denise Dresser told staff Anthropic’s then-$30 billion run rate was inflated by about $8 billion because Anthropic grosses up cloud-marketplace revenue while OpenAI reports Microsoft-channel revenue net. Both treatments can be GAAP-compliant. Anthropic’s defense, given to Reuters, is that it is the principal and sets price and delivery. OpenAI’s claim is adversarial, but Reuters later confirmed the gross-up is real: full marketplace billings hit revenue, partner cuts hit operating expense. [47] [21]
• Run rate versus recognized revenue. End-2025 “run rate” figures near $9 billion, and at least one press account of $10 billion in 2025 “actual” revenue, do not match the prospectus figure of nearly $4.6 billion. Annualizing a strong month is a company-preferred metric, not a GAAP year. [25] [1]
• The $42 billion loss is easy to misuse. About $34 billion is fair-value accounting on convertibles, which rises when the company is marked higher. The operating loss above $8 billion is the economic number. Bull notes that ignore it, and bear notes that lead with $42 billion of “cash burn,” are both wrong. [1]
• Forward revenue is investor talk. The $100–120 billion year-end 2026 range came from Anthropic backers talking to the Financial Times, not from audited guidance. One investor’s “800% growth justifies 30x revenue, so $3 trillion” is a circular multiple, not a forecast. [48]
• Margin path has already slipped once. Internal 2025 gross-margin hopes were cut before the year closed. Any third-party quarterly margin table that is more precise than the draft prospectus should be labeled a model. [9] [5]
• 2026 quarterly revenue in the leak is partly preliminary. PitchBook flags second-quarter revenue above $11.5 billion as preliminary. [5]

What this means: the bull case is built on the least audited number (run rate) and the bear case is sometimes built on the most misleading one (GAAP net loss). The figures that survive both filters—operating loss, compute spend above 2025 revenue, 80% non-cancelable commitments, two customers at 12% each, 47% channel sales—are the ones in Reuters’ reading of the draft.

Bull counterpoints that a serious bear has to answer

The growth is not fake. Revenue rose about twelvefold in 2025. First-quarter 2026 revenue exceeded all of 2025. Preliminary second-quarter revenue topped $11.5 billion. By late July the run rate was above $65 billion. If December annualized revenue really lands at $100–110 billion, $2 trillion is roughly 18–20 times sales, which Morningstar’s Michael Field called less extreme than SpaceX’s debut multiple. [1] [49] [42] FT’s Lex column goes further: if 2028 sales hit the levels bulls imply, $2 trillion could be a mid-single-digit to low-teens multiple of that year’s revenue—and the bull case can be stretched to absurdity ($5–10 trillion) by inflating TAM. That is an argument about narrative elasticity, not about 2025 cash flow. [50]

Other fair counters:

• Inference cost per dollar of revenue has been falling in company projections and in third-party models. A sustained gross margin in the 40–50% range, which some pre-filing analysis expected, would confirm half the private-round underwriting. The prospectus can confirm that half. It cannot confirm 2030 revenue. [51]
• Enterprise mix is a real difference versus a consumer-heavy peer, if it holds. Channel partners are also distribution, not only concentration.
• Training on legally acquired books was held to be fair use; the $1.5 billion deal is finite and partly paid.
• Export controls on Fable and Mythos were lifted within weeks. Direct government revenue is under 1%.
• If compute really is the binding constraint, take-or-pay contracts are capacity insurance, and fixed costs leverage margins up as usage grows inside contracted power. PitchBook states that upside explicitly. [5]
• Ed Zitron’s mid-2026 call that the $47 billion run rate would not hold has already been overtaken by higher company figures. Top-line bears have a losing recent record. [41]

The disagreement is not whether Claude is selling. It is whether a business that in 2025 spent more on compute than it earned, that has locked in hundreds of billions of dollars it must pay even if usage disappoints, and that sells a product whose price is being cut by rivals and by itself, should be capitalized at twice the price sophisticated investors paid four months earlier. PitchBook’s line is the fairest summary of the evidence now in hand: fast growth is established; $2 trillion is not. [5]

---

Recent Findings Supplement (October 2026)

Anthropic’s September 2026 IPO prospectus (reviewed by Reuters) reveals 2025 revenue of ~$4.59 billion (12x growth from $386 million in 2024) against an $8.06 billion operating loss and $7.33 billion in compute/infrastructure spend, with $518 billion in future cloud/compute commitments (roughly 80% non-cancelable or take-or-pay). This structure shows how explosive top-line growth coexists with capital intensity that far exceeds current cash ($20.28 billion at year-end 2025) and revenue, while improving gross margins in some analyses (e.g., quarterly models showing progression toward 50%+) still leave the company dependent on continued hyperscaler financing and usage-based revenue that may not cover fixed obligations.[1][2]

• Compute spend tripled year-over-year and represented ~58% of 2025 operating expenses; Q2 2026 showed early adjusted operating profit signals in some reports, with gross margins cited around 52% or higher before partner shares/training costs in select analyses.[3]
• Projections in the filing and related reporting include 2028 revenue of $190–200 billion, but the $518 billion commitment horizon (e.g., $111.1B to Google through 2033, $110B to Amazon through 2036) dwarfs near-term cash flow.[4]
• For competitors or new entrants: Securing equivalent long-term compute at scale requires either deep Big Tech ties or alternative hardware paths; pure-play model developers face margin pressure unless they achieve materially better utilization or vertical integration.

Anthropic’s relationships with Amazon, Google, and Nvidia illustrate circular financing where investors double as suppliers, revenue collectors, competitors, and (in Nvidia’s case) potential IPO anchors, creating interdependent cash flows that may not reflect arm’s-length demand.[5]

• Amazon and Google (major investors and primary cloud partners) accounted for 47% of 2025 revenue routed through their marketplaces (~$2.16 billion), with Anthropic paying ~$351 million in distribution fees (~16% take); these same firms collect bills and compete directly in AI.[6]
• Nvidia has discussed up to $10 billion as an IPO anchor (in addition to prior commitments) while Anthropic buys Nvidia-powered capacity through partners; similar loops exist with Microsoft and others.[7]
• Implications: New entrants or rivals must navigate or replicate these ecosystems; any disruption (e.g., partner prioritization of their own models or regulatory scrutiny of vertical integration) could cascade to both funding and distribution.

Two unnamed customers each contributed 12% of 2025 revenue (24% combined), with many large customers lacking long-term contracts and able to reduce spending without corresponding cost relief; 47% of revenue flows through Amazon/Google channels whose share has risen sharply.[8]

• Run-rate figures accelerated from ~$9 billion annualized at end-2025 to $47 billion by May 2026 and >$65 billion by end-July 2026, with investor expectations of $100–120 billion by year-end 2026.[9]
• OpenAI has reportedly questioned gross vs. net revenue recognition on some partner deals, highlighting potential inflation in reported run-rates.[10]
• For market participants: High concentration and flexible customer terms create downside asymmetry—revenue can drop faster than costs (especially fixed compute commitments)—favoring diversified or enterprise-locked models over pure usage-based plays.

Reported IPO targets of >$2 trillion (more than double the $965 billion May 2026 private valuation) imply ~30x the July 2026 run-rate or higher multiples assuming continued 800%+ growth, yet the prospectus underscores losses, commitments, and risks that have prompted AI-bubble comparisons and down-round precedents elsewhere in tech.[11]

• Bull case (investor views in FT reporting): Sustained hyper-growth and margin expansion (some quarterly models project positive adjusted operating margins by late 2026) could justify premiums akin to or exceeding SpaceX’s $1.77 trillion IPO.[12]
• Bear case (PitchBook/Morningstar analysis of leaked figures): At $2 trillion, valuation assumes stable customers, productive use of contracted capacity, and rising margins that have not yet produced sustained free cash flow; compute costs per revenue dollar have declined but remain material.[13]
• Implications: Public-market scrutiny may force more conservative pricing or disclosure; entrants betting on similar multiples face execution risk if growth decelerates.

New legal and regulatory developments include final approval of a $1.5 billion copyright settlement (largest known U.S. copyright recovery) in July 2026 over unauthorized book downloads for training, plus a September 2026 appeals court ruling upholding a Pentagon “supply-chain risk” designation restricting DOD use of Anthropic products.[14][15]

• Some authors opted out and continue separate cases; the settlement requires destruction of pirated files and provides ~$3,000 per work.
• The DOD designation (typically for adversary-linked firms) survived one appeal despite a partial earlier court block, with Anthropic considering further review.
• Implications: Copyright exposure remains live for non-settled claims and peers; government relations risks (export controls or procurement bans) could limit addressable markets for frontier labs.

September 2026 model launches by Anthropic (Claude Opus 5.5) and OpenAI featured aggressive price/performance cuts (20–80% token price reductions plus efficiency gains yielding ~40%+ effective savings), intensifying competition with each other and cheaper open-weight models (including Chinese offerings), supporting commoditization concerns.[16]

• Anthropic has resisted deep enterprise discounts in favor of strict metered billing, contrasting with more flexible competitor approaches.
• Past run-rate or valuation reports (e.g., earlier private rounds or leaked figures) have shown variability in recognition and growth assumptions, underscoring the need for prospectus-level verification over press estimates.
• For competitors: Price competition compresses margins and rewards efficiency/segmentation; durable advantages may shift toward distribution, data, or vertical applications rather than raw model capability.

These post-April 2026 disclosures (primarily the late-September prospectus and related reporting) provide the most concrete, company-sourced data yet, tempering optimism around valuation multiples with explicit warnings on costs, concentration, and obligations. Bullish growth narratives persist but rest on execution assumptions that public markets will test directly.